Advertisement

SWAVE 2.0 Imaging of Placental Elasticity and Viscosity: Potential Biomarkers for Placenta-Mediated Disease Detection

Open AccessPublished:September 27, 2022DOI:https://doi.org/10.1016/j.ultrasmedbio.2022.08.001

      Abstract

      Pregnancy complications such as pre-eclampsia (PE) and intrauterine growth restriction (IUGR) are associated with structural and functional changes in the placenta. Different elastography techniques with an ability to assess the mechanical properties of tissue can identify and monitor the pathological state of the placenta. Currently available elastography techniques have been used with promising results to detect placenta abnormalities; however, limitations include inadequate measurement depth and safety concerns from high negative pressure pulses. Previously, we described a shear wave absolute vibro-elastography (SWAVE) method by applying external low-frequency mechanical vibrations to generate shear waves and studied 61 post-delivery clinically normal placentas to explore the feasibility of SWAVE for placental assessment and establish a measurement baseline. This next phase of the study, namely, SWAVE 2.0, improves the previous system and elasticity reconstruction by incorporating a multi-frequency acquisition system and using a 3-D local frequency estimation (LFE) method. Compared with its 2-D counterpart, the proposed system using 3-D LFE was found to reduce the bias and variance in elasticity measurements in tissue-mimicking phantoms. In the aim of investigating the potential of improved SWAVE 2.0 measurements to identify placental abnormalities, we studied 46 post-delivery placentas, including 26 diseased (16 IUGR and 10 PE) and 20 normal control placentas. By use of a 3.33-MHz motorized curved-array transducer, multi-frequency (80,100 and 120 Hz) elasticity measures were obtained with 3-D LFE, and both IUGR (15.30 ± 2.96 kPa, p = 3.35e–5) and PE (12.33 ± 4.88 kPa, p = 0.017) placentas were found to be significantly stiffer compared with the control placentas (8.32 ± 3.67 kPa). A linear discriminant analysis (LDA) classifier was able to classify between healthy and diseased placentas with a sensitivity, specificity and accuracy of 87%, 78% and 83% and an area under the receiver operating curve of 0.90 (95% confidence interval: 0.8–0.99). Further, the pregnancy outcome in terms of neonatal intensive care unit admission was predicted with a sensitivity, specificity and accuracy of 70%, 71%, 71%, respectively, and area under the receiver operating curve of 0.78 (confidence interval: 0.62–0.93). A viscoelastic characterization of placentas using a fractional rheological model revealed that the viscosity measures in terms of viscosity parameter n were significantly higher in IUGR (2.3 ± 0.21) and PE (2.11 ± 0.52) placentas than in normal placentas (1.45 ± 0.65). This work illustrates the potential relevance of elasticity and viscosity imaging using SWAVE 2.0 as a non-invasive technology for detection of placental abnormalities and the prediction of pregnancy outcomes.

      Key Words

      Introduction

      The idea of “elastography,” that is, assessing the mechanical properties of inaccessible organs using "remote palpation," was described well in one of the earliest works by Kit Hill in the book Physical Principles of Medical Ultrasonics (
      • Hill CR.
      Telehistology.
      ). Since its inception, the technology has greatly evolved in the last three decades and has validated and extended the long-established value of manual palpation for the assessment of tissue stiffness in many clinical applications. Besides the common diagnostic and treatment monitoring applications such as breast mass evaluation, focal and diffuse liver disease detection, prostate gland evaluation and thermal therapy monitoring (
      • Sarvazyan A
      • Hall TJ
      • Urban MW
      • Fatemi M
      • Aglyamov SR
      • Garra BS.
      An overview of elastography—An emerging branch of medical imaging.
      ), new areas of applications are steadily emerging. In the field of obstetrics, the utility of elasticity imaging is being explored with a focus on cervix and placenta to predict labor onset and to assess the risk of pregnancy complications, respectively (
      • Feltovich H.
      Elastography applications in pregnancy.
      ).
      After the first exploratory study on placental elasticity evaluation in 2012 (Li et al. 2012), several studies have been conducted with the objectives of establishing baseline elasticity for normal placenta tissue (
      • McAleavey SA
      • Parker KJ
      • Ormachea J
      • Wood RW
      • Stodgell CJ
      • Katzman PJ
      • Pressman EK
      • Miller RK.
      Shear wave elastography in the living, perfused, post-delivery placenta.
      ;
      • Abeysekera JM
      • Ma M
      • Pesteie M
      • Terry J
      • Pugash D
      • Hutcheon JA
      • Mayer C
      • Lampe L
      • Salcudean S
      • Rohling R.
      SWAVE imaging of placental elasticity and viscosity: Proof of concept.
      ;
      • Calle S
      • Simon E
      • Dumoux MC
      • Perrotin F
      • Remenieras JP.
      Shear wave velocity dispersion analysis in placenta using 2-D transient elastography.
      ) and understanding the changes in placental stiffness resulting from pregnancy complications such as pre-eclampsia (PE) (
      • Cimsit C
      • Yoldemir T
      • Akpinar IN.
      Shear wave elastography in placental dysfunction: Comparison of elasticity values in normal and preeclamptic pregnancies in the second trimester.
      ;
      • Kiliç F
      • Kayadibi Y
      • Yüksel MA
      • Adaletli I
      • Ustabaşioğlu FE
      • Öncül M
      • Madazli R
      • Yilmaz MH
      • Mihmanli İ
      • Kantarci F.
      Shear wave elastography of placenta: In vivo quantitation of placental elasticity in preeclampsia.
      ;
      • Alan B
      • Tunc S
      • Agacayak E
      • Bilici A.
      Diagnosis of pre-eclampsia and assessment of severity through examination of the placenta with acoustic radiation force impulse elastography.
      ;
      • Karaman E
      • Arslan H
      • Cetin O
      • Şahin HG
      • Bora A
      • Yavuz A
      • Elasan S
      Akbu-dak I. Comparison of placental elasticity in normal and pre-eclamptic pregnant women by acoustic radiation force impulse elastosonography.
      ;
      • Fujita Y
      • Nakanishi TO
      • Sugitani M
      • Kato K.
      Placental elasticity as a new non-invasive predictive marker of pre-eclampsia.
      ;
      • Spiliopoulos M
      • Kuo CY
      • Eranki A
      • Jacobs M
      • Rossi CT
      • Iqbal SN
      • Fisher JP
      • Fries MH
      • Kim PC.
      Characterizing placental stiffness using ultrasound shear-wave elastography in healthy and preeclamptic pregnancies.
      ), intrauterine growth restriction (IUGR) (
      • Sugitani M
      • Fujita Y
      • Yumoto Y
      • Fukushima K
      • Takeuchi T
      • Shimokawa M
      • Kato K.
      A new method for measurement of placental elasticity: Acoustic radiation force impulse imaging.
      ;
      • Ohmaru T
      • Fujita Y
      • Sugitani M
      • Shimokawa M
      • Fukushima K
      • Kato K.
      Placental elasticity evaluation using Virtual Touch tissue quantification during pregnancy.
      ;
      • Durhan G
      • Unverdi H
      • Deveci C
      • Büyüksireci M
      • Karakaya J
      • Degirmenci T
      • Bayrak A
      • Kosar P
      • Hucumenoglu S
      • Ergün Y.
      Placental elasticity and histopathological findings in normal and intra-uterine growth restriction pregnancies assessed with strain elastography in ex vivo placenta.
      ;
      • Habibi HA
      • Davutoglu EA
      • Kandemirli SG
      • Aslan M
      • Ozel A
      • Ucar AK
      • Zeytun P
      • Madazli R
      • Adaletli I.
      In vivo assessment of placental elasticity in intrauterine growth restriction by shear-wave elastography.
      ;
      • Eroglu H
      • Tolunay HE
      • Tonyali NV
      • Orgul G
      • Şahin D
      • Yücel A.
      Comparison of placental elasticity in normal and intrauterine growth retardation pregnancies by ex vivo strain elastography.
      ), pre-term birth (
      • Albayrak E
      • Dogru HY
      • Ozmen Z
      • Altunkas A
      • Kalayci TO
      • Inci MF
      • Server S
      • Sonmezgoz F
      • Aktas F
      • Demir O.
      Is evaluation of placenta with real-time sonoelastography during the second trimester of pregnancy an effective method for the assessment of spontaneous preterm birth risk?.
      ) and gestational diabetes mellitus (GDM) (
      • Yuksel MA
      • Kilic F
      • Kayadibi Y
      • Alici Davutoglu E
      • Imamoglu M
      • Bakan S
      • Mihmanli I
      • Kantarci F
      • Madazli R
      Shear wave elastography of the placenta in patients with gestational diabetes mellitus.
      ;
      • Bildaci TB
      • Cevik H
      • Desteli GA
      • Tavasli B
      • Ozdogan S.
      Placental elasticity on patients with gestational diabetes: Single institution experience.
      ;
      • Lai HW
      • Lyv GR
      • Wei YT
      • Zhou T.
      The diagnostic value of two-dimensional shear wave elastography in gestational diabetes mellitus.
      ). These complications fall within the spectrum of “great obstetrical syndromes,” a term indicating their association with structural and functional changes of the placenta (
      • Romero R
      • Romero R.
      Prenatal medicine: The child is the father of the man.
      ;
      • Gabbay-Benziv R
      • Baschat AA.
      Gestational diabetes as one of the great obstetrical syndromes—The maternal, placental, and fetal dialog.
      ). The etiopathogenesis of these pregnancy-specific complications, a topic of extensive research, is currently accepted to be multifactorial. The early-onset type of PE (<32 wk), the more severe phenotype, which is often accompanied by fetal growth restriction, is dominated by placental factors. Defects in the placentation process between weeks 8 and 18 cause an incomplete spiral artery remodeling and, therefore, intermittent and high-velocity uteroplacental hypo-perfusion (
      • Staff A
      • Benton S
      • von DP
      • Roberts J
      • Taylor R
      • Powers R
      • Charnock-Jones D
      • Redman C
      Redefining preeclampsia using placenta-derived biomarkers.
      ). The resulting oxidative and hydrodynamic stress triggers the release of placenta-derived factors and, therefore, maternal inflammatory response, inducing de novo onset of hypertension and other signs of PE. Also, the insufficient uteroplacental perfusion can contribute to the associated growth restriction of the fetus. On the contrary, the late-onset types of PE (>32 wk) and gestational diabetes are predominantly “maternal,” where predisposing cardiovascular or metabolic risks for endothelial dysfunction exacerbate the systemic inflammatory response to the physiological burden of pregnancy (
      • Staff A
      • Benton S
      • von DP
      • Roberts J
      • Taylor R
      • Powers R
      • Charnock-Jones D
      • Redman C
      Redefining preeclampsia using placenta-derived biomarkers.
      ). Irrespective of the underlying pathophysiological process, these pregnancy complications affect the placenta and manifest as different types of pathologies depending on the severity and chronicity of the placental injury and placental maladaptation. The changes in placenta pathology will affect its stiffness, which can be quantified using elastographic methods. Previous studies indicated the efficacy of a quantitative elasticity measure as a global parameter to differentiate between placentas from normal pregnancies and pregnancies affected by PE, IUGR and GDM (
      • Sugitani M
      • Fujita Y
      • Yumoto Y
      • Fukushima K
      • Takeuchi T
      • Shimokawa M
      • Kato K.
      A new method for measurement of placental elasticity: Acoustic radiation force impulse imaging.
      ;
      • Cimsit C
      • Yoldemir T
      • Akpinar IN.
      Shear wave elastography in placental dysfunction: Comparison of elasticity values in normal and preeclamptic pregnancies in the second trimester.
      ;
      • Ohmaru T
      • Fujita Y
      • Sugitani M
      • Shimokawa M
      • Fukushima K
      • Kato K.
      Placental elasticity evaluation using Virtual Touch tissue quantification during pregnancy.
      ;
      • Yuksel MA
      • Kilic F
      • Kayadibi Y
      • Alici Davutoglu E
      • Imamoglu M
      • Bakan S
      • Mihmanli I
      • Kantarci F
      • Madazli R
      Shear wave elastography of the placenta in patients with gestational diabetes mellitus.
      ;
      • Durhan G
      • Unverdi H
      • Deveci C
      • Büyüksireci M
      • Karakaya J
      • Degirmenci T
      • Bayrak A
      • Kosar P
      • Hucumenoglu S
      • Ergün Y.
      Placental elasticity and histopathological findings in normal and intra-uterine growth restriction pregnancies assessed with strain elastography in ex vivo placenta.
      ), as well as a local parameter to identify fibrotic lesions, inflammation and infarction from healthy placental tissue (
      • Sugitani M
      • Fujita Y
      • Yumoto Y
      • Fukushima K
      • Takeuchi T
      • Shimokawa M
      • Kato K.
      A new method for measurement of placental elasticity: Acoustic radiation force impulse imaging.
      ;
      • Ohmaru T
      • Fujita Y
      • Sugitani M
      • Shimokawa M
      • Fukushima K
      • Kato K.
      Placental elasticity evaluation using Virtual Touch tissue quantification during pregnancy.
      ).
      Strain elastography and acoustic radiation force impulse (ARFI)-based elastography techniques (shear wave elasticity imaging [SWEI] and ARFI imaging) are the predominant approaches currently being considered for measuring placental stiffness in clinical practice. Elastography techniques generally assess, either directly or indirectly, the Young's modulus (E), the physical parameter corresponding to tissue stiffness, by applying an external force to the studied tissue and tracking the resulting motion. Strain elastography (
      • Ophir J
      • Céspedes E
      • Ponnekanti H
      • Yazdi Y
      • Li X.
      Elastography: A quantitative method for imaging the elasticity of biological tissues.
      ) applies a constant stress (a) (quasi-static) with arbitrary value using transducer palpation (or alternatively using the internal physiological pulsations from cardiac, respiratory or muscle contractions while holding the transducer still [
      • Dietrich CF
      • Barr RG
      • Farrokh A
      • Dighe M
      • Hocke M
      • Jenssen C
      • Dong Y
      • Saftoiu A
      • Havre RF.
      Strain elastography—How to do it?.
      ]) and measures the strain (e) map, where strain gives an indirect measure of Young's modulus via Hooke's law (a = Ee). Strain elastography, despite being simple to implement, has limited clinical uptake because of the operator dependence for the control of stress applied and absence of absolute quantification, which hinders the comparison of stiffness between different tissue types and pathological states across different studies. Additionally, the requirement for application of a stress by the operator excludes pregnancies with posterior placentation, which is more prevalent compared with those with accessible placentation sites (e.g., anterior and lateral) (
      • Magann EF
      • Evans SF
      • Newnham JP.
      Placental implantation at 18 weeks and migration throughout pregnancy.
      ). The alternative approach to use the internal motion would not be applicable for the placenta. Contrary to the quasi-static strain elastography technique, the dynamic approaches apply a time-varying force (transient or steady-state). The ARFI-based techniques apply a focused high-power beam (pushing beam) to displace tissue at the focal region of the pushing beam. The pushing beam generates shear waves with a speed (cs) directly related to the elasticity (E) via E = 3pcs, under the assumption that the propagation medium with density p is linear, isotropic, incompressible and elastic. ARFI imaging computes a relative measure of tissue stiffness by measuring the on-axis displacement and recovery (
      • Nightingale K.
      Acoustic radiation force impulse (ARFI) imaging: A review.
      ), while SWEI measures the shear wave speed by tracking tissue motion at several lateral positions along the shear wave propagation path (
      • Sandrin L
      • Fourquet B
      • Hasquenoph J
      • Yon S
      • Fournier C
      • Mal F
      • Christidis C
      • Ziol M
      • Poulet B
      • Kazemi F
      • Beaugrand M
      • Palau R
      Transient elastography: A new noninvasive method for assessment of hepatic fibrosis.
      ). ARFI-Based techniques are available in several commercial diagnostic ultrasound scanners and have had promising results in measuring placental elasticity. However, these techniques have potential risk of thermal tissue damage because of the long-duration and high-power acoustic push pulse. Transducer face heating and tissue heating become particularly critical during the multi-frame acquisitions and multiple pushes required for 2-D image formation (
      • Nightingale K.
      Acoustic radiation force impulse (ARFI) imaging: A review.
      ). The bio-effects of ARFI-based techniques on fetal and placental tissue is yet to be determined (
      • Simon E
      • Calle S.
      Safety of elastography applied to the placenta: Be careful with ultrasound radiation force.
      ;
      • Issaoui M
      • Debost-Legrand A
      • Skerl K
      • Chauveau B
      • Magnin B
      • Delabaere A
      • Boyer L
      • Sauvant-Rochat MP
      • Lemery D.
      Shear wave elastography safety in fetus: A quantitative health risk assessment.
      ). Alternatively, the elastography techniques that use low-frequency mechanical vibrations externally to generate shear waves, such as transient elastography (TE) (
      • Calle S
      • Simon E
      • Dumoux MC
      • Perrotin F
      • Remenieras JP.
      Shear wave velocity dispersion analysis in placenta using 2-D transient elastography.
      ) and shear wave absolute vibro-elastography (SWAVE) (
      • Abeysekera JM
      • Ma M
      • Pesteie M
      • Terry J
      • Pugash D
      • Hutcheon JA
      • Mayer C
      • Lampe L
      • Salcudean S
      • Rohling R.
      SWAVE imaging of placental elasticity and viscosity: Proof of concept.
      ), offer alternatives to measure the placental stiffness without acoustic radiation force impulses.
      In the first study (
      • Abeysekera JM
      • Ma M
      • Pesteie M
      • Terry J
      • Pugash D
      • Hutcheon JA
      • Mayer C
      • Lampe L
      • Salcudean S
      • Rohling R.
      SWAVE imaging of placental elasticity and viscosity: Proof of concept.
      ), low-frequency steady-state mechanical waves were used to excite tissue and provide quantitative measure of placental stiffness. Low-frequency shear waves have the capability of deep penetration (and, thus, the ability to access both anterior and posterior located placentas) and reduced safety concerns. That study illustrated the application of the SWAVE system on clinically normal placentas ex vivo and established a baseline elasticity. In the current SWAVE 2.0 project, we extend our previous baseline study to a clinical study that includes multimodal image acquisition from ex vivo placenta specimens from clinically normal pregnancies and pregnancies complicated by PE and intrauterine growth restriction.
      SWAVE 2.0 elastography incorporates several improvements over the previous elastography study (
      • Abeysekera JM
      • Ma M
      • Pesteie M
      • Terry J
      • Pugash D
      • Hutcheon JA
      • Mayer C
      • Lampe L
      • Salcudean S
      • Rohling R.
      SWAVE imaging of placental elasticity and viscosity: Proof of concept.
      ). First, the previous SWAVE study used single-frequency excitation and, therefore, required repeated acquisitions for measurement at different frequencies. This involved increased acquisition time, inconsistency between region-of-interest probed at different frequencies, and inability to utilize multi-frequency elasticity reconstruction approach required for improved performance. Second, the previous study utilized a linear array transducer with a depth limit of 9 cm, which would be unable to access the posterior-lying placentas. Third, a bandpass sampling scheme was used (
      • Eskandari H
      • Goksel O
      • Alcudean E
      • Rohling R.
      Bandpass sampling of high-frequency tissue motion.
      ), achieving a frame rate of 72 Hz, which is not sufficient for in vivo applications in the presence of breathing and fetal motions. Finally, a single-frequency 2-D local frequency estimation (LFE) algorithm was used for elasticity reconstruction, potentially leading to less accurate estimation. In the current work, we address the above-mentioned limitations. We perform hardware and software modifications to enable multi-frequency excitation, thereby improving acquisition time and elasticity reconstruction performance. We use a curved array transducer, with an imaging depth up to 24 cm, as a preparation for the future in vivo study. We use a sector-based sampling approach (
      • Baghani A
      • Brant A
      • Salcudean S
      • Rohling R.
      A high-frame-rate ultrasound system for the study of tissue motions.
      ), which resulted in a 11-fold increase with a sampling rate of 800 Hz. Finally, we apply a multifrequency 3-D LFE algorithm for accurate elasticity reconstruction. Novel contributions of this article are (i) validation of the improved elastography system in tissue-mimicking phantoms; (ii) application of the system to acquire data from a study cohort including normal control placentas and placentas from pregnancies affected by PE and IUGR; (iii) establishment of a stiffness threshold to distinguish the diseased from the control placentas; and (iv) viscoelastic characterization of placentas using a fractional rheological model. We compared different elasticity measurement methods (2-D LFE vs. 3-D LFE, single frequency vs. multi-frequency) to identify the most suitable SWAVE parameter measurement approach in terms of improved classification performance, that is, sensitivity, specificity, accuracy and area under the receiver operating characteristic curve (AUROC) measures. Finally, we performed a comprehensive analysis of the diagnostic performance of SWAVE parameters for predicting clinical findings associated with placental health (healthy, IUGR or PE). We also analyzed the correlations of SWAVE parameters with the pregnancy outcome in terms of admission of neonates into intensive care.

      Methods

      Phantoms and study population

      The proposed system was validated on a set of two homogeneous phantoms (Model 039) manufactured by CIRS (Norfolk, VA, USA). The elasticity measures obtained using a 3-D magnetic resonance elastography (MRE) system were considered as the gold standard. The MRE was performed at excitation frequencies of {50, 55, 60, 65} Hz using a 3-T Achieva scanner (Philips Inc., Amsterdam, Netherlands) at the University of British Columbia (UBC) MRI research center. A detailed description of the MRE data acquisition can be found in
      • Zeng Q
      • Honarvar M
      • Schneider C
      • Mohammad SK
      • Lobo J
      • Pang EHT
      • Lau KT
      • Hu C
      • Jago J
      • Erb SR
      • Rohling R
      • Salcudean SE.
      Three-dimensional multi-frequency shear wave absolute vibro-elastography (3D S-wave) with a matrix array transducer: Implementation and preliminary in vivo study of the liver.
      and

      Mohammed S, Honarvar M, Zeng Q, Hashemi H, Rohling R, Kozlowski P, Salcudean S. Model-based quantitative elasticity reconstruction using ADMM. IEEE Transactions on Medical Imaging (2022).

      .
      Forty-six placentas were collected from a group of women (n = 40) who delivered via cesarean delivery at BC Women's Hospital, Vancouver, Canada. We included 26 placentas from pregnancies affected by PE and/or IUGR. PE was defined as the presence of gestational hypertension with proteinuria, and a pregnancy was considered to be complicated by IUGR if the fetal abdominal circumference was below the 10th percentile measured from ultrasound examination (
      • Lessoway VA
      • Schulzer M
      • Wittmann BK
      • Gagnon FA
      • Wilson RD.
      Ultrasound fetal biometry charts for a North American Caucasian population.
      ). We included 10 controls from singleton pregnancies and 10 control twin placentas (from 5 dichorionic twin pregnancies) that were uncomplicated by PE and IUGR. The exclusion criteria included any gestational abnormality identified prior to delivery and pregnancies receiving treatment with investigational medication. The study (H17-00331) was performed under written informed consent of all participants after approval by the University of British Columbia/Children's and Women's Health Centre of British Columbia Research Ethics Board (UBC C&W REB). Study data were collected and managed using REDCap electronic data capture tools hosted at British Columbia Women's and Children's Hospital. Clinical data extracted from the participants’ health records (maternal and neonatal charts) included maternal age, parity, risk factors for placental disease (history of smoking, hypertension, diabetes), birth date and time, gestational age at diagnosis, gestational age at birth, birth weight, cord blood pH and admission to neonatal intensive care.

      Placenta preparation and region-of-interest identification

      Placentas were stored at 4°C until the examination. The study, including gross examination, region-of-interest identification, ultrasound, MRI and pathology data acquisition, was initiated between approximately 6 and 48 h after delivery and completed within 5 d of delivery. On an average, the time between placenta acquisition and completion of SWAVE data collection was 59 h. At the beginning of the examination, a perinatal pathologist (J.T.) removed the umbilical sac and identified externally appreciable placental disc lesions. Layers of acoustic absorbing pad (Aptflex F28, Precision Acoustics, UK) were placed beneath the maternal surface of the disc to reduce reverberation artifacts. The placenta was secured using two rubber bands at the edges to the absorbing pad layers and then placed in a rigid plastic container to minimize any deformation while transporting from one imaging modality to another. A sonographer (V.L.) scanned the placenta using a SonixTouch ultrasound machine (Analogic Corp., Richmond, BC, Canada) and an L14-5/38 linear transducer to locate two regions-of-interest: (i) a homogeneous region without macroscopic or sonographically visible abnormalities, and (ii) an inhomogeneous region with suspected lesions. The regions of interest were marked and photographed for future reference. With an objective of investigating the global elasticity measure, we considered the homogeneous regions of interest for all the experiments.

      Scanning system overview, data acquisition and data processing

      The elastography data of the placenta were acquired using a SWAVE (shear wave absolute vibro-elastography) imaging system (
      • Abeysekera JM
      • Ma M
      • Pesteie M
      • Terry J
      • Pugash D
      • Hutcheon JA
      • Mayer C
      • Lampe L
      • Salcudean S
      • Rohling R.
      SWAVE imaging of placental elasticity and viscosity: Proof of concept.
      ). The SWAVE system consists of an Ultrasonix SonixTouch ultrasound machine (Analogic, Richmond, BC, Canada) with a m4DC7-3/40 curved array transducer (Ultrasonix, Richmond, BC, Canada). The radiofrequency data sampling frequency was 20 MHz. The curved array transducer was set to operate at a transmit frequency of 3.33 MHz to perform a feasibility test for the future in vivo experiments. According to the American Institute of Ultrasound in Medicine (AIUM) (
      • Crino J
      • Finberg HJ
      • Frieden F
      • Kuller J
      • Odibo A
      • Robichaux A
      • Bohm-Velez M
      • Pretorius DH
      • Sheth S
      • Angtuaco TL
      • De Lange M
      • Gara B
      • Hertzberg B
      • Hoffenberg S
      • Jaffe R
      • Kurtz A
      • Mastrobattista J
      • McGahan J
      • Meilstrup J
      • Middleton W
      • Nelson T
      • Paushter D
      • Rapp C
      • Robbin M
      • H Kotlus Rosenberg
      • Toy E
      • Yeo L
      • Bohm-Velez M
      • Warren BH
      • Benacerraf BR
      • Benson C
      • Brown DL
      • Finberg HJ
      • Frates MC
      • Goldstein RB
      • Gooding GA
      • Hansen GC
      • Larson PA
      • Liebscher LA
      • Rumack CM
      • Timins JK
      • Jr Way WG
      AIUM practice guideline for the performance of obstetric ultrasound examinations.
      ), although a transmit frequency between 3 and 5 MHz is sufficient to achieve an optimum trade-off between penetration depth and resolution, a lower-frequency (∼ 3 MHz) transducer might be required for abdominal imaging in obese patients. Because of the hardware limitation and possible interference between different excitation frequencies, we acquired multi-frequency elasticity data in two sets of experiments at [80, 100, 120] (n = 46) and [40, 50, 60, 70] Hz (n = 41) and thereby limited the number of frequencies in each acquisition. The imaging depths were set to 5 cm with a focus at 2 cm to capture the full thickness of the full-term placentas. The placenta fixed within the container was placed on a custom-built plexiglass shaker platform. The shaker platform, along with the placenta container, was then submerged in a constant-temperature water bath (Cole-Parmer, Montreal, QC, Canada) with temperature set at 37°C, to ensure a speed of sound value of 1540 m/s. The shaker platform was connected to a voice coil motor through a rod that generates a multifrequency (between 40 and 120 Hz) excitation. The software modules of the SWAVE system include the excitation control module, which programs an Arduino microcontroller to generate the multi-frequency input signal from the voice coil motor. The amplitude ratios were set to 1:1.5:2.25:3 and 1:1:1.5 for the sinusoids of the two frequency sets of [40, 50, 60, 70] Hz and [80, 100, 120] Hz, respectively, to account for the increased attenuation effect at higher-frequency shear waves. The phase difference between the sinusoids was optimized to decrease the amplitude of the resultant multifrequency signal. A set of four springs were affixed to the platforms’ corner to allow vibration in the vertical direction. During data acquisition, the vibrating shaker platform generates shear waves within the placenta. The transducer was submerged in the water bath with a few millimeters between the transducer face and the fetal surface to avoid deformation and potential increase in stiffness measure. The transducer was positioned to image the placenta at the marked location while being parallel to the platform and the water bath. During the examination, the shaker platform generates shear waves in the placenta in a manner similar to that expected in vivo with a vibration board beneath a supine patient's back. Volumetric ultrasound radiofrequency data were captured and stored for offline processing to measure tissue displacements and reconstruct elasticity maps. We acquired a volume consisting of 15 planes, spaced at 0.95° angular positions. The volume acquisitions were completed within 6.3 s. The experimental setup of our SWAVE 2.0 system is shown in Fig. 1.
      Fig 1
      Fig. 1The experimental setup of shear wave absolute vibro-elastography (SWAVE) 2.0 for ex vivo placenta elasticity measurement. Top: The placenta is marked using India ink and oil pills. The container, holding the placenta as well as layers of acoustic absorbing pad, is placed on a plexiglass shaker platform and submerged in a water bath. The shaker platform is connected to a voice coil motor through a rod. The ultrasound transducer is submerged in the water bath and is imaging the fetal side of the placenta. Bottom: The transducer is submerged in the water bath with a few millimeters between the transducer face and the fetal surface (left). A graphical user interface displays the shear wave propagation along with the B-mode image to facilitate data acquisition (right).
      For tissue displacement tracking, a sector subdivision method (
      • Baghani A
      • Brant A
      • Salcudean S
      • Rohling R.
      A high-frame-rate ultrasound system for the study of tissue motions.
      ) was used, where each imaging plane was divided into eight sectors. Each sector is imaged 40 times before moving to the imaging of the next sector, enabling the motion estimation to be computed at the sector repetition frequency. The effective frame rate achieved in this method was 800 Hz. The relative tissue displacement along the axial direction between consecutive ultrasound frames was estimated using the time-domain cross-correlation with prior estimates (TDPE) method (
      • Zahiri-Azar R
      • Salcudean SE.
      Motion estimation in ultrasound images using time domain cross correlation with prior estimates.
      ). According to this method, the RF data were divided into blocks of 63 data samples with a 68% overlap, the normalized cross-correlation function was computed, and the position of the maximum peak of the normalized cross-correlation was measured using cosine interpolation, resulting in the steady-state tissue displacement at the respective block's location and time instance. From the steady-state displacement estimates, displacement phasors at each excitation frequency were computed using least-squares fitting. To obtain the synchronous displacement phasors required for subsequent elasticity measurement, the time delay between scanning points on a single scan line at different depth, between scan lines, and between sectors was compensated in the frequency domain (
      • Baghani A
      • Brant A
      • Salcudean S
      • Rohling R.
      A high-frame-rate ultrasound system for the study of tissue motions.
      ) based on the assumptions of periodic motion, constant ultrasound speed of sound within the region of interest and still transducer with respect to the tissue. The synchronized phasor volumes were then scan-converted based on the transducer specifications and spatially interpolated with a grid size of 0.6 × 0.6 × 0.6 mm3 in Cartesian coordinates.
      Local frequency estimation was used to estimate the local shear wave speed at each excitation frequency using filter banks operating on the displacement phasors (
      • Knutsson H
      • Westin CF
      • Granlund GH.
      Local multiscale frequency and bandwidth estimation.
      ). First, the phasor volume in 3-D LFE (while only the central plane of the phasor volume for 2-D LFE) at each excitation frequency is translated into k-space (spatial frequency domain) using fast Fourier transform and filtered using a fourth-order Butterworth bandpass filter with cutoff frequencies associated with an elasticity of E [1,100] kPa. The bandpass filter removes the compressional waves while effects of other waves, such as Lamb waves, have been neglected in this study. Six k-space directional filters along the axial, lateral and elevational directions (±x, ±y and ±z) were applied to the entire 3-D volume for 3-D LFE, while on the 2-D plane in 2-D LFE considering the elevational components to be zero. Now, the local shear wave speed can be measured using a ratio of the output of two filters operating on the displacements (
      • Oliphant T
      • Kinnick R
      • Manduca A
      • Ehman R
      • Greenleaf J.
      An error analysis of Helmholtz inversion for incompressible shear, vibration elastography with application to filter-design for tissue characterization.
      ). To attain this, a collection of log-normal filters at 11 different central frequencies spaced an octave apart from 1/210 to 1/20 were used. These filters were defined as a product between radial and directional components (
      • Knutsson H
      • Westin CF
      • Granlund GH.
      Local multiscale frequency and bandwidth estimation.
      ). The final LFE, expressed as wavenumber km, is a weighted summation of the estimates from the pairs of filters. For 3-D LFE, the local wavenumber km at each excitation frequency um is estimated for a 3-D voxel, while 2-D LFE provides the local wavenumber for a 2-D voxel. Finally, the elasticity estimate over all excitation frequencies is computed as a weighted average where the estimate at each frequency is weighted by the displacement phasor amplitude, Um,
      E=3ρcS2=3ρnm1nUmωm2km2
      (1)


      where ρ is tissue density and is assumed to be equal to 1000 kg/m3. cs is the local shear wave speed evaluated over all excitation frequencies.

      Rheological modeling

      In our proposed 3-D multi-frequency SWAVE method, data were acquired at multiple frequencies: 40, 50, 60, 70, 80, 100 and 120 Hz. For rheological modeling, shear wave speed at each excitation frequency cs(ω), can be computed as
      cs(ω)=ω/k
      (2)


      resulting in the single-frequency elasticity measure, E(ω) = 3ρcs(ω)2. Analyzing the shear wave speed dispersion, that is, the change of speed cs as a function of frequency u, both tissue elasticity and viscosity can be quantified (
      • Chen S
      • Fatemi M
      • Greenleaf JF.
      Quantifying elasticity and viscosity from measurement of shear wave speed dispersion.
      ). Voigt's model was used previously for the characterization of viscoelastic properties of normal placentas (
      • Abeysekera JM
      • Ma M
      • Pesteie M
      • Terry J
      • Pugash D
      • Hutcheon JA
      • Mayer C
      • Lampe L
      • Salcudean S
      • Rohling R.
      SWAVE imaging of placental elasticity and viscosity: Proof of concept.
      ), which models the tissue with a complex shear modulus G* as a parallel arrangement of a spring of shear modulus μ and a dashpot of viscosity η(G*(ω) = μ + jωη). According to this model, the dispersion of shear wave speed is related to μ and η as follows:
      cs(ω)=2(μ2+ω2η2)ρ(μ+(μ2+ω2η2))
      (3)


      The parameters μ and η were computed using a least-squares fit between the shear wave speed measures obtained from the model and that obtained using eqn (3).
      A previous study reported that a fractional rheological model can more accurately describe the power law behavior of the complex shear modulus G*(ω) for placental tissue compared with the Voigt model (
      • Simon E
      • Calle S.
      Safety of elastography applied to the placenta: Be careful with ultrasound radiation force.
      ). According to the fractional rheological model,
      G*(ω)=Ge+K·ωn
      (4)


      where G* is the complex shear modulus, Ge is the equilibrium shear modulus, K is the coefficient of consistence and n is the constitutive parameter. The dispersion of shear wave speed is related to Ge, K and n as follows:
      cs(ω)=2(Ge2+ω2nK2)ρ(Ge+(Ge2+ω2nK2))
      (5)


      The parameters Ge, K and n were computed using a least-squares fit between the shear wave speed measures obtained from the model and that obtained using eqn (5). We measured the root mean square error as a metric of goodness of the fit for both rheological models. We apply the rheological model fitting on the 41 placentas for which data at all 7 frequencies [40, 50, 60, 70, 80,100,120] Hz were available.

      Classifier training and evaluation

      We train and evaluate a linear discriminant analysis (LDA) classifier to assess the potential of elasticity parameters to discriminate between tissue from control and diseased placentas. Classifier performance was assessed using AUROC values obtained from ROC analyses and sensitivity, specificity and accuracy obtained with fivefold cross validation. We also reported the cutoff values of corresponding parameters.

      Histopathology and ultrasound registration

      After imaging, each placenta underwent pathological examination including measurement of the placental weight and dimensions. The placenta was sampled in such a way that the acquired full-thickness sections were aligned with the regions of interest marked during placental preparation and imaged with ultrasound. The registration of the center plane in ultrasound volume to the histopathology plane was fine-tuned using anatomic landmarks that were clearly identifiable in both modalities. These landmarks included the specimen contour along the maternal and fetal surface and lesion boundaries. The rigid transformation matrix was generated using the MatchPoint Registration plug-in in MITK (
      • Wolf I
      • Vetter M
      • Wegner I
      • Büottger T
      • Nolden M
      • Schüobinger M
      • Hasten-teufel M
      • Kunert T
      • Meinzer HP.
      The medical imaging interaction toolkit.
      ).

      Statistical analysis

      Statistical analyses were performed using MATLAB R2020b (The MathWorks, Natick, MA, USA) and IBM SPSS statistics for Windows, version 27.0 (IBM Corp, Armonk, NY, USA). We summarized continuous variables using the mean and standard deviation or median and interquartile range. Nominal or ordinal variables were described as frequencies with percentages. Because the sample size was relatively small and the distributions of the estimated quantitative ultrasound measures were not strictly Gaussian, the Wilcoxon rank-sum test was used to compare the statistical significance of the difference between placentas from different groups. A p value < 0.05 was considered to indicate statistical significance.

      Results

      Participants

      The present study included 46 placentas. As mentioned, these placentas were divided into two groups: control and PE and/or IUGR. Of the 26 placentas complicated by PE and/or IUGR, there were 16 IUGR placentas, 1 PE placenta and 9 placentas affected by both IUGR and PE. Among all cases affected by IUGR, 15 were categorized as early onset (diagnosed at <32 wk), 2 were late onset (diagnosed at >32 wk) and the remainder were not recorded. Among the PE cases for which the gestational age at diagnosis was recorded, 4 were early onset (diagnosed at <32 wk) and one was late onset. Among our control placentas, 1 singleton and 4 twin placentas were from pregnancies affected by gestational diabetes mellitus. Table 1 summarizes the clinical characteristics of the study cohort by group. The mean gestational age at delivery for the pregnant women in the IUGR/PE group was 33.1 wk (standard deviation = 14.2 wk), which was significantly lower (p < 0.05) compared with that of the control group, 38.6 wk (standard deviation =1.2 weeks). Also, the pregnant women in the IUGR/PE group had significantly higher occurrences of hypertension compared with the control group (p < 0.05). The birth weight (1560 ± 717 g) and the placenta weight (234 ± 110 g) associated with the IUGR/PE group were also significantly lower compared with the birth weight (3379 ± 610 gm) and the placenta weight (464 ± 102 gm) in the control group (p < 0.05). However, placenta thickness measures did not significantly differ between the control group (1.89 ± 0.49 cm) and the IUGR/PE group (1.73 ± 0.51 cm) (p > 0.05). Twenty of 26 IUGR/PE-affected neonates were admitted into the neonatal intensive care unit (NICU).
      Table 1Clinical characteristics of pregnancies for the placentas studied.
      Maternal parameterControls (n = 15)IUGR or PE (n = 24)p Value
      GravidityMedian = 2Range = 1–8Median = 2Range = 1–60.33
      ParityMedian = 1Range = 0–3Median = 0Range = 0-30.18
      Gestational age (wk)Mean = 38.6SD = 1.2Mean = 33.1SD = 4.23.5e–6
      Maternal age (y)Mean = 34.5SD = 5.7Mean = 32.5SD = 6.40.43
      Smoking statusEx-smoker n = 0Non-smoker n = 12Ex-smoker n = 3Non-smoker n=180.64
      HypertensionYesNoYesNo
      n =0n = 15n = 12n = 80.01
      DiabetesYesNoYesNo
      n = 3n = 12n = 4n = 160.41
      Neonatal parameterControls (n = 20)IUGR or PE (n = 26)p Value
      Birth weight (g)Mean = 3379SD = 610Mean = 1560SD = 7172.1e–8
      Placenta weight (g)Mean = 464SD = 102Mean = 234SD = 1101.6e–6
      Placenta thickness (cm)Mean = 1.89SD = 0.49Mean =1.73SD = 0.510.23
      GenderFemaleMaleFemaleMale
      n = 9n = 10n = 15n = 90.62
      NICUYesNoYesNo
      Admission0202061.6e–7
      IUGR = intra-uterine growth retardation; LFE = local frequency estimation; NICU = neonatal intensive care unit; PE = pre-eclampsia.

      Phantom validation

      Rectangular regions of interest were selected from the center plane of the acquired image volume for each of the two phantoms. The means and standard deviations obtained from the regions of interest obtained using the single-frequency and multi-frequency approaches of 2-D LFE and 3-D LFE are listed in Table 2. For phantom 1, the SWAVE 2.0 system using multi-frequency 3-D LFE results in elasticity measures with a bias of 15% compared with the gold standard values measured using the MRE, while the multi-frequency 2-D LFE method results in a bias of 48%. For phantom 2, the multi-frequency 2-D LFE and multi-frequency 3-D LFE result in biases of 59% and 24%, respectively. This result validates the findings obtained in several previous studies, which reported that 3-D LFE provides more accurate estimation of tissue stiffness (
      • Yin M
      • Manduca A
      • Romano AJ
      • Glaser KJ
      • Drapaca C
      • Lake DS
      • Ehman RL.
      3-D local frequency estimation inversion for abdominal MR elastography.
      ;
      • Hemily J.
      Towards liver shear wave vibro-elastography: Method repeatability and image registration technique.
      ). Two-dimensional LFE tends to overestimate the tissue stiffness as it computes the wavelength from the projection of the 3-D wave onto the 2-D axial-lateral plane.
      Table 2Elasticity measures for the tissue-mimicking phantoms
      MethodFrequencyPhantom 1 (kPa)Phantom 2 (kPa)
      SWAVE (2-D LFE)8010.04 ± 2.6523.98 ± 7.25
      10010.10 ± 3.9223.46 ± 5.31
      1209.53 ± 3.0823.91 ± 4.83
      [80, 100, 120]9.69 ± 2.5324.89 ± 3.90
      SWAVE 2.0 (3-D LFE)805.14 ± 0.979.07 ± 1.48
      1005.31 ± 0.8410.90 ± 1.69
      1206.27 ± 0.6713.20 ± 1.45
      [80, 100, 120]5.55 ± 0.8511.97 ± 1.03
      3-D MRE[50, 55, 60, 65]6.56 ± 0.3215.67 ± 0.33
      IUGR = intra-uterine growth retardation; MRE = magnetic resonance elastography; PE = pre-eclampsia; SWAVE = shear wave absolute vibro-elastography.
      Elasticity measurements are expressed as the mean ± standard deviation of Young's modulus (E) measures.

      Elasticity of placenta ex vivo

      Shear wave absolute vibro-elastography imaging examples of ex vivo placental elasticity along with the histology overlays are provided in Fig. 2a. Representative examples for both the control group and the PE/IUGR group do not contain any focal lesions, as defined by our perinatal pathologist (J.T.), who reviewed the histopathology slides. These were validated by the histology images. From the B-mode images and the histology images, there was not any discernible difference between normal placenta and PE/IUGR placenta. The elasticity results obtained from both 2-D LFE and 3-D LFE illustrate that PE/IUGR placentas are associated with elevated elasticity. However, the elasticity range for 2-D LFE is much higher than that for 3-D LFE. Fig. 2b indicates that the shear waves for the PE/IUGR case have longer wavelengths compared with those for the control case. Also, the projected waves on the axial-lateral plane are longer compared with those in the other two planes. Therefore, elasticity measures obtained from 2-D LFE using the projected waves on the axial-lateral plane yield high stiffness artifacts and consequently high mean elasticity and high variance. The elasticity measures from 3-D LFE were comparatively homogeneous, as was expected from the histology image in this particular case. Also, Fig. 3a illustrates that the elasticity measures obtained using 2-D LFE were higher than those obtained using 3-D LFE, which were statistically significant for both control (p = 1.25e–6) and PE/IUGR (p = 2.48e–5) groups. Fig. 3b compares the elasticity measures for the control, IUGR and PE groups. The elasticity measures obtained using 3-D LFE for both the IUGR (15.30 ± 2.96 kPa, p = 3.35e–5) and PE (12.33 ± 4.88 kPa, p = 0.017) groups were significantly higher compared with those for the control group (8.32 ± 3.67 kPa). On the other hand, the elasticity values obtained using 2-D LFE for the IUGR (23.40 ± 9.56 kPa, p = 0.75) and PE (23.80 ± 8.73, p = 0.46) groups were slightly higher compared with that for the control group (22.46 ± 10.02 kPa), which did not reach statistical significance. For the subsequent analysis, we present only the 3-D LFE multi-frequency measurements.
      Fig 2
      Fig. 2Shear wave absolute vibro-elastography (SWAVE) imaging of ex vivo placental elasticity for representative examples from the control and the pre-eclampsia (PE)/intra-uterine growth retardation (IUGR) group. (a) Elasticity results obtained using 2-D local frequency estimation (LFE) and 3-D LFE are shown as overlays. The regions of interest are selected from the registered histological sections. (b) Phasor planes obtained at 80 Hz corresponding to the slices shown on the B-mode images.
      Fig 3
      Fig. 3(a) Elasticity results obtained using 2-D local frequency estimation (LFE) versus elasticity results obtained using 3-D LFE. Two-dimensional LFE overestimates the stiffness for most placenta examples from both the control and pre-eclampsia (PE)/intra-uterine growth retardation (IUGR) groups (indicated by the points on the right side of the identity line). The histograms of the differences between E measures obtained using 2-D LFE and 3-D LFE for both classes are also shown. (b) Violin plots of placentas from the control, IUGR and PE groups obtained using 2-D LFE (left) and 3-D LFE (right). All violin and boxplots indicate the first, second and third quartiles; the whiskers indicate the minimum and maximum values. * denotes p < 0.05 and NS denotes non-significant. .
      We also evaluate the elasticity for placentas grouped according to clinical outcome (i.e., NICU admission). The elasticity measures for the placentas corresponding to the neonates required to be admitted to the NICU (13.69 ± 3.67 kPa) were significantly higher compared with those for the neonates not admitted to the NICU (9.37 ± 4.80 kPa, p = 0.0013). Finally, we measure the elasticity in placentas in late versus early onset of IUGR and PE. The stiffness values of placentas in early IUGR (12.82 ± 5.22 kPa) and late IUGR (13.57 ± 3.50 kPa) were similar (p = 0.72). The stiffness of placentas in late-onset PE (17.13 kPa) was higher compared with those in early onset (10.65 ± 2.73 kPa); however, the sample size was too small to compute a meaningful statistical significance.

      Viscoelastic characterization

      Fig. 4a indicates that the shear wave speed, and hence the resulting elasticity E, increases monotonically with excitation frequency u. The slopes of the elasticity dispersion for the placentas in the IUGR and PE groups are steeper than those for the control group, indicating higher viscosity values are associated with pregnancy complications. Fig. 4b illustrates that the fractional rheological model (right) more accurately represents the frequency dependence of the dispersion curve compared with the Voigt model (left). We obtained mean root mean square (RMS) error values of 0.03 (2.11% of average shear wave speed) for the control group, 0.05 (3.19% of average shear wave speed) for the IUGR group and 0.04 (2.61% of average shear wave speed) for the PE group using the fractional model fit. On the other hand, the mean RMS errors obtained for the Voigt fit for the control, IUGR and PE groups were 0.07 (4.94% of average shear wave speed), 0.17 (10.85 % of average shear wave speed) and 0.22 (14.33% of average shear wave speed), respectively. Therefore, we chose the fractional rheological model for the viscoelastic characterization of the placenta and subsequently present the results obtained by fitting the fractional model. We found similar shear modulus values in placentas with IUGR (n = 13) and PE (n = 10) group and the control group (n = 17) (Fig. 4c, left). The placentas in the control group have a shear modulus value of 1192.10 ± 454.88 Pa (mean ± standard deviation), while the placentas in the IUGR and PE groups have shear modulus values of 1264.80 ± 483.73 Pa (p = 0.71) and 1053.00 ± 580.11 Pa (p = 0.44). On the other hand, we obtained a significantly increased constitutive parameter, n, in the placentas in IUGR group (2.3 ± 0.21, p = 0.0039) and PE group (2.11 ± 0.52, p = 0.0042), compared with the n value of placentas in the control group (1.45 ± 0.65).
      Fig 4
      Fig. 4Viscoelastic characterization of the placentas from control, intra-uterine growth retardation (IUGR) and pre-eclampsia (PE) groups. (a) Elasticity measures as a function of frequency (n = 41). The line and shaded regions indicate the mean elasticity and standard deviations at corresponding frequency. The boxplots are spread apart by a small distance about each excitation frequency for improved visualization. (b) Shear wave speed dispersion and the fitted Voigt model (left) and fitted fractional rheological model (right) for the mean values of placentas in the control, IUGR and PE group. (c) Violin plots and boxplots of shear modulus (left) and viscosity parameter n (right) of placentas from the control (n = 17), IUGR (n = 13) and PE (n = 10) groups. All violin plots and boxplots indicate the first, second and third quartiles; the whiskers indicate the minimum and maximum values. * denotes p < 0.05; ** denotes p < 0.01; NS denotes non-significant.

      Diagnostic performance evaluation of SWAVE

      In Fig. 5 are the ROC curves illustrating the classification performance of different SWAVE parameters in discriminating between control and PE/IUGR placentas using a LDA classifier. The corresponding AUROC, sensitivity, specificity, accuracy and cutoff values obtained for the minimum misclassification cost are reported in Table 3, except for the parameters with an AUROC <0.5 (indicating an incorrect inverse relationship between SWAVE parameters and placenta disease state). Among the single frequency elasticity measures obtained using 3-D LFE (Fig. 5, left), low-frequency measures corresponding to 40 to 60 Hz were unable to detect the increase in stiffness corresponding to placental abnormality. Higher-frequency elasticity measures tend to attain increasingly improved classification performance, where elasticity at 120 Hz achieved 91% sensitivity, 85% accuracy and an AUROC of 0.84 (confidence interval [CI]: 0.7–0.98). The difference among the AUROCs for the elasticity measures for 70, 80, 100 and 120 Hz did not reach statistical significance.
      Fig 5
      Fig. 5Receiver operating characteristic (ROC) curves for elasticity estimates between control and pre-eclampsia (PE)/intra-uterine growth retardation (IUGR)-affected placentas using single frequency and 3-D local frequency estimation (LFE) technique (left), elasticity estimates using multifrequency (middle), and for shear modulus and viscosity parameter, n (right). The dashed line indicates an area under the ROC curve (AUROC) of 0.5.
      Table 3Thresholds, AUROC, sensitivity and specificity for ROC analyses of elasticity for control and PE/IUGR-affected placentas
      MeasureAUROC (95% CI)Cutoff (kPa)SensitivitySpecificityAccuracy
      70 Hz0.65 (0.47–0.83)5.10.740.440.61
      80 Hz0.70 (0.52–0.89)5.90.830.560.73
      100 Hz0.78 (0.62–0.93)8.80.780.610.73
      120 Hz0.84 (0.7–0.98)13.60.910.780.85
      Multifrequency

      ([80, 100, 120] Hz)


      0.90 (0.8–0.99)


      10.52


      0.87


      0.78


      0.83
      n0.79 (0.61–0.96)1.730.710.960.83
      AUROC = area under the receiver operating characteristic curve; CI = confidence interval; ROC = receiver operating characteristic IUGR = intra-uterine growth retardation; PE = pre-eclampsia.
      Multi-frequency elasticity measure had improved classification performance, whereas 2D LFE demonstrated a performance of a random classifier, with an AUROC of 0.41 (Fig. 5, middle). Finally, classification analyses on fractional rheological parameters reveal that shear wave modulus had poor predictive value (AUROC = 0.28) in distinguishing between control and diseased placentas. However, the viscosity parameter n had improved classification performance (AUROC = 0.79, CI: 0.61–0.96) in identifying placental tissue from complicated pregnancies (Fig. 5, right). Among all the parameters, multifrequency elasticity measures obtained using 3-D LFE were found to yield the best classification performance, with AUROC significantly different from most other parameters except single frequency measures for 100 and 120 Hz.
      Fig. 6 illustrates the classification performance of the SWAVE parameters that were selected based on superior discriminating abilities between outcomes in terms of NICU admission. Multi-frequency elasticity measures obtained using 3-D LFE produced the best result with an AUROC of 0.78 (CI: 0.62–0.93) in determining pregnancy outcome in terms of NICU admission. The accuracy, sensitivity and specificity obtained with this parameter were 0.71, 0.70 and 0.71, respectively. The single-frequency elasticity measure obtained with the curved-array transducer had a sensitivity and specificity of 0.65 and 0.76, with an accuracy of 0.71. Viscosity n was found to be correlated with NICU outcome, with a sensitivity of 0.70 and specificity of 0.55, resulting in an accuracy of 0.63.
      Fig 6
      Fig. 6Receiver operating characteristic (ROC) curves and corresponding area under the ROC curve (AUROC) values with 95% confidence intervals for elasticity estimates between cases with and without neonatal intensive care unit admission. All elasticity measures were obtained using 3-D local frequency estimation. The dashed line indicates an AUROC of 0.5.

      Discussion

      Among the different system settings and methods we investigated in the current work, the SWAVE elasticity measures obtained using a multi-frequency 3-D LFE approach performed best in distinguishing placentas ex vivo from normal and diseased pregnancies. Our findings agree with the previous reports in the literature indicating that placentas from pregnancies affected by PE (12.33 ± 4.88 kPa) and IUGR (15.30 ± 2.96 kPa) are significantly stiffer, corresponding to increases of 48% and 84%, respectively, compared with those from normal pregnancies (8.32 ± 3.67 kPa). However, because of differences in elastography techniques used in the previous works, the different quantitative measures of placenta stiffness in any disease group will be different. Table 4 outlines the range of elasticity measures, sample sizes, elasticity techniques and study types in previous studies and our studies. We observe that the point shear wave elastography (pSWE) techniques for in vivo studies yield a mean/median Young's modulus value within 2.28 and 3.85 kPa for those in the control group. Two-dimensional shear wave elastography techniques, on the other hand, are associated with higher Young's modulus values ranging from 5.8–11.06 kPa for the similar placenta group. The elasticity measures obtained using the proposed SWAVE 2.0 method fall within the ranges obtained with those obtained using SWEI (2-D), TE and the previous SWAVE method.
      Table 4Comparison between the elasticity (E) values obtained in this study and those obtained in previous studies
      StudyMethodControlElasticity (kPa)
      PEIUGR
      • Spiliopoulos M
      • Kuo CY
      • Eranki A
      • Jacobs M
      • Rossi CT
      • Iqbal SN
      • Fisher JP
      • Fries MH
      • Kim PC.
      Characterizing placental stiffness using ultrasound shear-wave elastography in healthy and preeclamptic pregnancies.
      SWE (2-D)

      (in vivo)
      10.45 ± 7.6

      (n = 18)
      26.36 ± 14.1

      (n = 18)
      NA
      • Akbas M
      • Koyuncu FM
      • Artunç-Ulkumen B.
      Placental elasticity assessment by point shear wave elastography in pregnancies with intrauterine growth restriction.
      pSWE

      (in vivo)
      3.85 ± 1.2

      (n = 81)
      NA5.51 ± 2.09

      (n = 66)
      • Fujita Y
      • Nakanishi TO
      • Sugitani M
      • Kato K.
      Placental elasticity as a new non-invasive predictive marker of pre-eclampsia.
      pSWE

      (in vivo )
      3.18 (1.22–6.48)

      (n = 181)
      5.96 (3.56–17.14)

      (n = 13)
      NA
      • Calle S
      • Simon E
      • Dumoux MC
      • Perrotin F
      • Remenieras JP.
      Shear wave velocity dispersion analysis in placenta using 2-D transient elastography.
      TE

      (ex vivo)
      9.57 ± 0.75

      (n = 20)
      NANA
      • Habibi HA
      • Davutoglu EA
      • Kandemirli SG
      • Aslan M
      • Ozel A
      • Ucar AK
      • Zeytun P
      • Madazli R
      • Adaletli I.
      In vivo assessment of placental elasticity in intrauterine growth restriction by shear-wave elastography.
      SWE (2-D)

      (in vivo)
      6 (4.38–7.45)

      (n = 42)
      NA28 (16.8–35)

      (n = 42)
      • Abeysekera JM
      • Ma M
      • Pesteie M
      • Terry J
      • Pugash D
      • Hutcheon JA
      • Mayer C
      • Lampe L
      • Salcudean S
      • Rohling R.
      SWAVE imaging of placental elasticity and viscosity: Proof of concept.
      SWAVE

      (ex vivo)
      11.18 ± 1.76
      Average of single frequency measures at 80, 100 and 100 Hz.


      (n = 61)
      NANA
      • Karaman E
      • Arslan H
      • Cetin O
      • Şahin HG
      • Bora A
      • Yavuz A
      • Elasan S
      Akbu-dak I. Comparison of placental elasticity in normal and pre-eclamptic pregnant women by acoustic radiation force impulse elastosonography.
      pSWE

      (in vivo)
      2.48 ± 0.2

      (n = 38)
      6.40 ± 0.69

      (n = 35)
      NA
      • McAleavey SA
      • Parker KJ
      • Ormachea J
      • Wood RW
      • Stodgell CJ
      • Katzman PJ
      • Pressman EK
      • Miller RK.
      Shear wave elastography in the living, perfused, post-delivery placenta.
      SWE (2-D)

      (ex vivo)
      11.06 ± 0.01

      (n = 11)
      NANA
      • Alan B
      • Tunc S
      • Agacayak E
      • Bilici A.
      Diagnosis of pre-eclampsia and assessment of severity through examination of the placenta with acoustic radiation force impulse elastography.
      pSWE

      (in vivo )
      3.43 (3.0–3.90) (n = 42)5.80 (5.23–7.02)

      (n = 44)
      NA
      • Cimsit C
      • Yoldemir T
      • Akpinar IN.
      Shear wave elastography in placental dysfunction: Comparison of elasticity values in normal and preeclamptic pregnancies in the second trimester.
      pSWE

      (in vivo )
      2.28 (2.02–2.58) (n = 101)6.67 (3.60–12.34)

      (n = 28)
      NA
      • Kiliç F
      • Kayadibi Y
      • Yüksel MA
      • Adaletli I
      • Ustabaşioğlu FE
      • Öncül M
      • Madazli R
      • Yilmaz MH
      • Mihmanli İ
      • Kantarci F.
      Shear wave elastography of placenta: In vivo quantitation of placental elasticity in preeclampsia.
      SWE (2D)

      (in vivo)
      5.8 (2.5–13)
      Expressed as median (maximum – minimum).


      (n = 27)
      25 (2.4–66.5)
      Expressed as median (maximum – minimum).


      (n = 23)
      NA
      • Ohmaru T
      • Fujita Y
      • Sugitani M
      • Shimokawa M
      • Fukushima K
      • Kato K.
      Placental elasticity evaluation using Virtual Touch tissue quantification during pregnancy.
      pSWE

      (in vivo)
      2.88 ± 0.13

      (n = 143)
      7.68 ± 0.61
      E measures for pregnancy-induced hypertension group have been classified under PE group.


      (n = 15)
      4.92 ± 0.46

      (n = 21)
      • Sugitani M
      • Fujita Y
      • Yumoto Y
      • Fukushima K
      • Takeuchi T
      • Shimokawa M
      • Kato K.
      A new method for measurement of placental elasticity: Acoustic radiation force impulse imaging.
      pSWE (ex vivo )5.15 ± 0.37

      (n = 74)
      6.66 ± 0.81
      E measures for pregnancy-induced hypertension group have been classified under PE group.


      (n=17)
      11.29 ± 1.64

      (n = 24)
      • Li W
      • Wei Z
      • Yan R
      • Zhang Y.
      Detection of placenta elasticity modulus by quantitative real-time shear wave imaging.
      SWE (2-D)

      (in vivo)
      7.84 ± 1.68

      (n = 30)
      NANA
      Our studySWAVE 2.0

      (ex vivo)
      8.32 ± 3.67

      (n = 20)
      12.33 ± 4.88

      (n = 10)
      15.30 ± 2.96

      (n = 16)
      AUROC = area under the receiver operating characteristic curve; CI = confidence interval; ROC = receiver operating characteristic IUGR = intra-uterine growth retardation; PE = pre-eclampsia; pSWE = point SWE; SWAVE = shear wave absolute vibro-elastography; SWE = shear wave elastography.
      Elasticity measurements are expressed as the mean ± standard deviation or median (25%–75% quartile) of Young's modulus (E) measures. Young's modulus (E) was converted from shear wave speed (cs) using the equation E = 3ρcs2 when necessary.
      low asterisk Average of single frequency measures at 80, 100 and 100 Hz.
      Expressed as median (maximum – minimum).
      E measures for pregnancy-induced hypertension group have been classified under PE group.
      Additionally, a fraction of the previous studies performed ROC analyses to assess the efficacy of elasticity measures as a predictor of placental abnormality (
      • Alan B
      • Tunc S
      • Agacayak E
      • Bilici A.
      Diagnosis of pre-eclampsia and assessment of severity through examination of the placenta with acoustic radiation force impulse elastography.
      ;
      • Fujita Y
      • Nakanishi TO
      • Sugitani M
      • Kato K.
      Placental elasticity as a new non-invasive predictive marker of pre-eclampsia.
      ).
      • Fujita Y
      • Nakanishi TO
      • Sugitani M
      • Kato K.
      Placental elasticity as a new non-invasive predictive marker of pre-eclampsia.
      reported a cutoff elasticity of 4.2 kPa to predict PE, resulting in an AUROC of 0.91.
      • Alan B
      • Tunc S
      • Agacayak E
      • Bilici A.
      Diagnosis of pre-eclampsia and assessment of severity through examination of the placenta with acoustic radiation force impulse elastography.
      identified a cutoff maximum elasticity measure of 9.08 kPa with an AUROC for PE diagnostic performance of 0.883. Similarly,
      • Spiliopoulos M
      • Kuo CY
      • Eranki A
      • Jacobs M
      • Rossi CT
      • Iqbal SN
      • Fisher JP
      • Fries MH
      • Kim PC.
      Characterizing placental stiffness using ultrasound shear-wave elastography in healthy and preeclamptic pregnancies.
      found a cutoff of 16.3 kPa to identify the presence of PE with a classification performance of AUROC = 0.82. On the other hand, our study found an optimum cutoff elasticity measure of 10.52 kPa resulting in an AUROC of 0.9 (CI: 0.8–0.99) in predicting PE and IUGR cases, which was consistent with the previous findings.
      To ensure that the reported elasticity data are acquired from regions where the shear wave components have been adequately transmitted, a mask was applied to select regions with a displacement magnitude >3 μm. The normalized correlation coefficient was used as a measure of speckle tracking accuracy, which was found to be >0.95 over the majority of regions of interest, with an average of 0.996 ± 0.016 for the [40, 50, 60, 70] Hz frequency set and 0.975 ± 0.027 for the [80, 100, 120] Hz frequency set.
      We also report the efficacy of the proposed method in predicting the pregnancy outcome in terms of NICU admission. The multi-frequency elasticity measure using 3-D LFE obtained for the neonatal group requiring NICU admission was significantly higher than the measure for the group not requiring NICU admission (13.69 ± 3.67 kPa vs. 9.37 ± 4.80 kPa, p = 0.0013). Also, the elasticity measure attained an AUROC of 0.78 (CI: 0.63–0.93) in predicting NICU admission.
      We also investigated the effect of the 3-D LFE reconstruction method compared with 2-D LFE. We found that 2-D LFE results in higher elasticity measures than those from 3-D LFE in both phantoms and the placentas. In the phantom study, 3-D LFE results in improved accuracy and precision in elasticity measures compared with 2-D LFE. However, from our single frequency LFE measurements, we found that there is an increase in Young's modulus with the increase in function for 3-D LFE, particularly for the stiffer phantom (Phantom 2). The 2-D LFE measures, on the contrary, did not exhibit a similar trend. The CIRS 039 phantom has a negligible storage modulus, with almost constant shear wave speed values at different frequencies (

      Mohammed S, Honarvar M, Zeng Q, Hashemi H, Rohling R, Kozlowski P, Salcudean S. Model-based quantitative elasticity reconstruction using ADMM. IEEE Transactions on Medical Imaging (2022).

      ), a finding that has also been confirmed by CIRS.
      Therefore, the presence of shear wave speed dispersion as a function of frequency is not expected. The reason for the apparent dispersion in elasticity in 3-D LFE could be owing to the small elevational dimension of the acquired ultrasound volume, which ranges from 0.5 to 1.44 cm across the depth. An elevational dimension with 1 cm could capture a half-wavelength of a shear wave corresponding to maxima of 7.7, 12.0 and 17.3 kPa at 80, 100 and 120 Hz, respectively. This limits the maximum elasticity that can be measured at a particular frequency, with a pronounced effect on stiffer material at low frequencies. The limited elevational dimension could also lead to the high bias obtained for phantom 2 in our study, with more accurate results obtained with the high frequency (15% bias at 120 Hz) compared with the lower frequency (42% bias at 80 Hz).
      In the ex vivo placenta experiments, we selected the homogeneous regions of interest to obtain relatively uniform elasticity measures suitable for stiffness comparison among different disease groups. However, the average standard deviation obtained using 2-D LFE was 8.80 kPa for intra-placenta elasticity measurement (38.2% of the mean value) and 2.56 kPa (22.76% of the mean value) for the 3-D LFE method. This agrees with previous findings that indicated that 2-D LFE is prone to overestimating tissue stiffness and results in high-stiffness artifacts (
      • Yin M
      • Manduca A
      • Romano AJ
      • Glaser KJ
      • Drapaca C
      • Lake DS
      • Ehman RL.
      3-D local frequency estimation inversion for abdominal MR elastography.
      ;
      • Hemily J.
      Towards liver shear wave vibro-elastography: Method repeatability and image registration technique.
      ). Simple planar wave imaging using 2-D LFE fails to capture the complex shear wave propagation, specifically at the boundary and low-SNR regions (
      • Yin M
      • Manduca A
      • Romano AJ
      • Glaser KJ
      • Drapaca C
      • Lake DS
      • Ehman RL.
      3-D local frequency estimation inversion for abdominal MR elastography.
      ). Furthermore, unlike its 3-D LFE counterpart, 2-D LFE could not statistically differentiate between different disease groups.
      We estimate elasticity measures using single-frequency and multi-frequency techniques. We also analyze the frequency dependence of placenta elasticity and perform a viscoelastic characterization of placentas belonging to different groups using a fractional rheological model fitting. Previous studies revealed the presence of a significant viscous component (
      • Abeysekera JM
      • Ma M
      • Pesteie M
      • Terry J
      • Pugash D
      • Hutcheon JA
      • Mayer C
      • Lampe L
      • Salcudean S
      • Rohling R.
      SWAVE imaging of placental elasticity and viscosity: Proof of concept.
      ;
      • Calle S
      • Simon E
      • Dumoux MC
      • Perrotin F
      • Remenieras JP.
      Shear wave velocity dispersion analysis in placenta using 2-D transient elastography.
      ) by fitting a rheological model to fit the shear wave speed dispersion with frequency variation. These studies were limited to the study of clinically normal placentas. Our current study confirms the previous findings and also identifies increased values of the exponent parameter of the power law, n, indicating increased viscosity in the diseased placenta group. This characteristic is evident through a steeper slope of shear wave speed dispersion (Fig. 4a). The viscosity parameter n was found to be a discriminating feature to classify normal placentas from PE- and IUGR-affected placentas with significantly different mean values (1.45 ± 0.65 vs. 2.11 ± 0.52 and 2.3 ± 0.21). The viscosity parameter n also exhibited superior diagnostic performance with an AUROC of 0.79 (CI: 0.61–0.96) in differentiating between normal and PE/IUGR cases and an AUROC of 0.78 (CI: 0.63–0.93). We further note that the RMS error in the Voigt fits for the PE (14.33% of average shear wave speed) and IUGR (10.85% of average shear wave speed) groups were higher compared with that for the control group (4.95% of average shear wave speed). This could be an indication that the diseased group increasingly becomes non-linear and therefore deviates from the linear Voigt model. The fractional rheological model could successfully describe the non-linear behavior, resulting in reduced RMS error in the fit for all three groups: 2.11% of average shear wave speed for the control group, 3.19% of average shear wave speed for the IUGR group and 2.61% of average shear wave speed for the PE group. Additionally, we found an inverse relationship between stiffness and the placental disease group (low elasticity for PE/IUGR group compared with the control group) in a lower frequency range (40–70 Hz). This could have occurred because the placental dimension captured by the ultrasound was not sufficiently large to capture a half-wavelength of the larger shear wave at low frequencies for PE/IUGR-affected stiffer placentas, which is a requirement for elasticity estimation using LFE (
      • Manduca A
      • Oliphant TE
      • Dresner MA
      • Mahowald J
      • Kruse SA
      • Amromin E
      • Felmlee JP
      • Greenleaf JF
      • Ehman RL.
      Magnetic resonance elastography: Non-invasive mapping of tissue elasticity.
      ). Therefore, the single frequency elasticity estimates at lower frequencies have greater errors. This is further evidenced by the poor ROC performance of the associated elasticity features (Fig. 6). On the contrary, elasticity measures at higher frequencies (70–120 Hz) have increasingly improved AUROC, that is, better diagnostic performance (Table 3 and Fig. 6). On the other hand, multi-frequency acquisition further improves the elasticity estimate by reducing artifacts arising from noisy displacement at one of the frequencies. In our study, the multi-frequency elasticity obtained by averaging the single frequency measures weighted by the displacement amplitude was found to outperform the single-frequency counterpart.
      Our placenta study has several limitations. First, the ex vivo nature of the current study limits its ability to capture the variation in stiffness owing to the altered blood perfusion owing to onset of PE or IUGR. Also, this post-delivery study does not identify the time of disease onset, and therefore it is not clear whether SWAVE 2.0 is able to detect the early symptoms of the placental abnormality required for effective intervention. Although this study aims to detect the global variation caused by placental abnormality, we do not incorporate the histopathological findings, and therefore, some of the elasticity measures can arise from the presence of local lesions. To address these limitations, we will design and perform a longitudinal in vivo study. We would also incorporate histopathological information from the registered pathology slides to identify the source of elasticity variation. A second limitation of our study was the relatively small population. Nevertheless, our sample size achieved an overall 99% power to reject the null hypothesis that the population mean is the same for all three groups when the group means are 8.32, 12.33 and 15.30 kPa and the pooled standard deviation is 3.74 kPa, with a significance level of 0.050 using a one-way analysis of variance. However, a pairwise power analysis found that the power attained for rejecting the equal mean hypothesis between the control and PE groups was 61%, whereas a power of 100% was achieved for the control and IUGR groups. The analyses underpin the importance of a larger sample size for the PE group. A third limitation of the study was the variable wait time to perform SWAVE imaging of the placentas. We compared the elasticity results and the quality of the measurements obtained on the first day, on the second day and on the fifth day after delivery on a small set of placentas (n = 4). There were no significant differences between the elasticity measures. However, the quality indices (including radiofrequency data, signal-to-noise ratio, correlation, displacement amplitude and phasor signal-to-noise ratio) were relatively low in the 5-d-old placentas compared with the 1-d-old placentas. Finally, the current study lacks a formal safety analysis of the SWAVE technology to be applied on pregnant women and fetus. We argue that the mechanical vibration applied in SWAVE is similar to the vibration of a mechanical massager, which has not been found to have deleterious effects on pregnant women and produces a relaxation response (
      • Diego MA
      • Field T
      • Sanders C
      • Hernandez-Reif M.
      Massage therapy of moderate and light pressure and vibrator effects on EEG and heart rate.
      ). However, a formal safety analysis will be conducted in future studies to establish the safety of the proposed method.

      Conclusions

      We have described the application of SWAVE 2.0, a 3-D multifrequency shear wave vibro-elastography method, for ex vivo placental imaging. Using the method, we computed the Youngs modulus for 46 placentas and determined that it can differentiate between placentas from normal pregnancies and placentas from pregnancies affected by pre-eclampsia and intra-uterine growth restriction. We identified a threshold of 10.52 kPa for the optimum system setting to classify placentas into control and diseased groups with an accuracy of 83%, sensitivity of 87%, specificity of 78% and AUROC of 0.9. Moreover, the proposed elasticity measure was found to be correlated with pregnancy outcomes in terms of NICU admission with an AUROC of 0.78, and accuracy, sensitivity and specificity of 0.71, 0.70 and 0.71, respectively. A viscoelastic characterization established viscosity as a discriminating feature that was able to differentiate between placentas from control and PE/IUGR groups and was correlated with NICU admission outcome. This work illustrates the feasibility of using SWAVE 2.0 in an in vivo pregnancy screening system to identify placental abnormality and to predict adverse outcome.

      Conflict of interest disclosure

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.

      Acknowledgments

      We sincerely thank Dr. Mohammad Honarvar, Julio Lobo, and Shahed Khan Mohammed for their suggestions and guidance for the system development and reconstruction algorithm. We acknowledge support from Microsoft Corporation, Schlumberger Foundation, Canadian Institutes of Health Research (CIHR) and the Natural Sciences and Engineering Research Council. Dr. Robert Rohling and Dr. Septimiu Salcudean hold director, executive and equity positions in Sonic Incytes, the commercial licensee of the S-WAVE technology, and are co-inventors on S-WAVE technology and patents. The authors also thank BC Children's Hospital BioBank for support in placenta sample collection.

      References

        • Abeysekera JM
        • Ma M
        • Pesteie M
        • Terry J
        • Pugash D
        • Hutcheon JA
        • Mayer C
        • Lampe L
        • Salcudean S
        • Rohling R.
        SWAVE imaging of placental elasticity and viscosity: Proof of concept.
        Ultrasound Med Biol. 2017; 43: 1112-1124
        • Akbas M
        • Koyuncu FM
        • Artunç-Ulkumen B.
        Placental elasticity assessment by point shear wave elastography in pregnancies with intrauterine growth restriction.
        J Perinat Med. 2019; 47: 841-846
        • Alan B
        • Tunc S
        • Agacayak E
        • Bilici A.
        Diagnosis of pre-eclampsia and assessment of severity through examination of the placenta with acoustic radiation force impulse elastography.
        Int J Gynecol Obstet. 2016; 135: 43-46
        • Albayrak E
        • Dogru HY
        • Ozmen Z
        • Altunkas A
        • Kalayci TO
        • Inci MF
        • Server S
        • Sonmezgoz F
        • Aktas F
        • Demir O.
        Is evaluation of placenta with real-time sonoelastography during the second trimester of pregnancy an effective method for the assessment of spontaneous preterm birth risk?.
        Clin Imaging. 2016; 40: 926-930
        • Baghani A
        • Brant A
        • Salcudean S
        • Rohling R.
        A high-frame-rate ultrasound system for the study of tissue motions.
        IEEE Trans Ultrason Ferroelectr Freq Control. 2010; 57: 1535-1547
        • Bildaci TB
        • Cevik H
        • Desteli GA
        • Tavasli B
        • Ozdogan S.
        Placental elasticity on patients with gestational diabetes: Single institution experience.
        J Chin Med Assoc. 2017; 80: 717-720
        • Calle S
        • Simon E
        • Dumoux MC
        • Perrotin F
        • Remenieras JP.
        Shear wave velocity dispersion analysis in placenta using 2-D transient elastography.
        J Appl Phys. 2018; 123234902
        • Chen S
        • Fatemi M
        • Greenleaf JF.
        Quantifying elasticity and viscosity from measurement of shear wave speed dispersion.
        J Acoust Soc Am. 2004; 115: 2781-2785
        • Cimsit C
        • Yoldemir T
        • Akpinar IN.
        Shear wave elastography in placental dysfunction: Comparison of elasticity values in normal and preeclamptic pregnancies in the second trimester.
        J Ultrasound Med. 2015; 34: 151-159
        • Crino J
        • Finberg HJ
        • Frieden F
        • Kuller J
        • Odibo A
        • Robichaux A
        • Bohm-Velez M
        • Pretorius DH
        • Sheth S
        • Angtuaco TL
        • De Lange M
        • Gara B
        • Hertzberg B
        • Hoffenberg S
        • Jaffe R
        • Kurtz A
        • Mastrobattista J
        • McGahan J
        • Meilstrup J
        • Middleton W
        • Nelson T
        • Paushter D
        • Rapp C
        • Robbin M
        • H Kotlus Rosenberg
        • Toy E
        • Yeo L
        • Bohm-Velez M
        • Warren BH
        • Benacerraf BR
        • Benson C
        • Brown DL
        • Finberg HJ
        • Frates MC
        • Goldstein RB
        • Gooding GA
        • Hansen GC
        • Larson PA
        • Liebscher LA
        • Rumack CM
        • Timins JK
        • Jr Way WG
        AIUM practice guideline for the performance of obstetric ultrasound examinations.
        J Ultrasound Med. 2013; 32: 1083-1101
        • Diego MA
        • Field T
        • Sanders C
        • Hernandez-Reif M.
        Massage therapy of moderate and light pressure and vibrator effects on EEG and heart rate.
        Int J Neurosci. 2004; 114: 31-44
        • Dietrich CF
        • Barr RG
        • Farrokh A
        • Dighe M
        • Hocke M
        • Jenssen C
        • Dong Y
        • Saftoiu A
        • Havre RF.
        Strain elastography—How to do it?.
        Ultrasound Int Open. 2017; 3: E137-E149
        • Durhan G
        • Unverdi H
        • Deveci C
        • Büyüksireci M
        • Karakaya J
        • Degirmenci T
        • Bayrak A
        • Kosar P
        • Hucumenoglu S
        • Ergün Y.
        Placental elasticity and histopathological findings in normal and intra-uterine growth restriction pregnancies assessed with strain elastography in ex vivo placenta.
        Ultrasound Med Biol. 2017; 43: 111-118
        • Eroglu H
        • Tolunay HE
        • Tonyali NV
        • Orgul G
        • Şahin D
        • Yücel A.
        Comparison of placental elasticity in normal and intrauterine growth retardation pregnancies by ex vivo strain elastography.
        Arch Gynecol Obstet. 2020; 302: 109-115
        • Eskandari H
        • Goksel O
        • Alcudean E
        • Rohling R.
        Bandpass sampling of high-frequency tissue motion.
        IEEE Trans Ultrason Ferroelectr Freq Control. 2011; 58: 1332-1343
        • Feltovich H.
        Elastography applications in pregnancy.
        in: Alam SK Feltovich H Tissue elasticity imaging. Elsevier, San Diego2020: 181-196
        • Fujita Y
        • Nakanishi TO
        • Sugitani M
        • Kato K.
        Placental elasticity as a new non-invasive predictive marker of pre-eclampsia.
        Ultrasound Med Biol. 2019; 45: 93-97
        • Gabbay-Benziv R
        • Baschat AA.
        Gestational diabetes as one of the great obstetrical syndromes—The maternal, placental, and fetal dialog.
        Best Pract Res Clin Obstet Gynaecol. 2015; 29: 150-155
        • Habibi HA
        • Davutoglu EA
        • Kandemirli SG
        • Aslan M
        • Ozel A
        • Ucar AK
        • Zeytun P
        • Madazli R
        • Adaletli I.
        In vivo assessment of placental elasticity in intrauterine growth restriction by shear-wave elastography.
        Eur J Radiol. 2017; 97: 16-20
        • Hemily J.
        Towards liver shear wave vibro-elastography: Method repeatability and image registration technique.
        University of British Columbia, 2017 (Ph.D. thesis)
        • Hill CR.
        Telehistology.
        in: Hill CR Bamber JC Ter Haar GR Physical principles of medical ultrasonics. Wiley, New York1986: 321-337
        • Issaoui M
        • Debost-Legrand A
        • Skerl K
        • Chauveau B
        • Magnin B
        • Delabaere A
        • Boyer L
        • Sauvant-Rochat MP
        • Lemery D.
        Shear wave elastography safety in fetus: A quantitative health risk assessment.
        Diagn Interv Imaging. 2018; 99: 519-524
        • Karaman E
        • Arslan H
        • Cetin O
        • Şahin HG
        • Bora A
        • Yavuz A
        • Elasan S
        Akbu-dak I. Comparison of placental elasticity in normal and pre-eclamptic pregnant women by acoustic radiation force impulse elastosonography.
        J Obstet Gynaecol Res. 2016; 42: 1464-1470
        • Kiliç F
        • Kayadibi Y
        • Yüksel MA
        • Adaletli I
        • Ustabaşioğlu FE
        • Öncül M
        • Madazli R
        • Yilmaz MH
        • Mihmanli İ
        • Kantarci F.
        Shear wave elastography of placenta: In vivo quantitation of placental elasticity in preeclampsia.
        Diagn Interv Radiol. 2015; 21: 202-207
        • Knutsson H
        • Westin CF
        • Granlund GH.
        Local multiscale frequency and bandwidth estimation.
        in: Proceedings of the 1st International Conference on Image Processing (ICIP), New York1. IEEE, 1994: 36-40
        • Lai HW
        • Lyv GR
        • Wei YT
        • Zhou T.
        The diagnostic value of two-dimensional shear wave elastography in gestational diabetes mellitus.
        Placenta. 2020; 101: 147-153
        • Lessoway VA
        • Schulzer M
        • Wittmann BK
        • Gagnon FA
        • Wilson RD.
        Ultrasound fetal biometry charts for a North American Caucasian population.
        J Clin Ultrasound. 1998; 26: 433-453
        • Li W
        • Wei Z
        • Yan R
        • Zhang Y.
        Detection of placenta elasticity modulus by quantitative real-time shear wave imaging.
        Clin Exp Obstet Gynecol. 2012; 39: 470-473
        • Magann EF
        • Evans SF
        • Newnham JP.
        Placental implantation at 18 weeks and migration throughout pregnancy.
        South Med J. 1998; 91: 1025-1027
        • Manduca A
        • Oliphant TE
        • Dresner MA
        • Mahowald J
        • Kruse SA
        • Amromin E
        • Felmlee JP
        • Greenleaf JF
        • Ehman RL.
        Magnetic resonance elastography: Non-invasive mapping of tissue elasticity.
        Med Image Anal. 2001; 5: 237-254
        • McAleavey SA
        • Parker KJ
        • Ormachea J
        • Wood RW
        • Stodgell CJ
        • Katzman PJ
        • Pressman EK
        • Miller RK.
        Shear wave elastography in the living, perfused, post-delivery placenta.
        Ultrasound Med Biol. 2016; 42: 1282-1288
      1. Mohammed S, Honarvar M, Zeng Q, Hashemi H, Rohling R, Kozlowski P, Salcudean S. Model-based quantitative elasticity reconstruction using ADMM. IEEE Transactions on Medical Imaging (2022).

        • Nightingale K.
        Acoustic radiation force impulse (ARFI) imaging: A review.
        Curr Med Imaging. 2011; 7: 328-339
        • Ohmaru T
        • Fujita Y
        • Sugitani M
        • Shimokawa M
        • Fukushima K
        • Kato K.
        Placental elasticity evaluation using Virtual Touch tissue quantification during pregnancy.
        Placenta. 2015; 36: 915-920
        • Oliphant T
        • Kinnick R
        • Manduca A
        • Ehman R
        • Greenleaf J.
        An error analysis of Helmholtz inversion for incompressible shear, vibration elastography with application to filter-design for tissue characterization.
        Proc IEEE Ultrason Symp. 2000; 2: 1795-1798
        • Ophir J
        • Céspedes E
        • Ponnekanti H
        • Yazdi Y
        • Li X.
        Elastography: A quantitative method for imaging the elasticity of biological tissues.
        Ultrason Imaging. 1991; 13: 111-134
        • Romero R
        • Romero R.
        Prenatal medicine: The child is the father of the man.
        J Matern Fetal Neonatal Med. 2009; 22: 636-639
        • Sandrin L
        • Fourquet B
        • Hasquenoph J
        • Yon S
        • Fournier C
        • Mal F
        • Christidis C
        • Ziol M
        • Poulet B
        • Kazemi F
        • Beaugrand M
        • Palau R
        Transient elastography: A new noninvasive method for assessment of hepatic fibrosis.
        Ultrasound Med Biol. 2003; 29: 1705-1713
        • Sarvazyan A
        • Hall TJ
        • Urban MW
        • Fatemi M
        • Aglyamov SR
        • Garra BS.
        An overview of elastography—An emerging branch of medical imaging.
        Cur Med Imaging. 2011; 7: 255-282
        • Simon E
        • Calle S.
        Safety of elastography applied to the placenta: Be careful with ultrasound radiation force.
        J Obstet Gynaecol Res. 2017; 43: 1509
        • Spiliopoulos M
        • Kuo CY
        • Eranki A
        • Jacobs M
        • Rossi CT
        • Iqbal SN
        • Fisher JP
        • Fries MH
        • Kim PC.
        Characterizing placental stiffness using ultrasound shear-wave elastography in healthy and preeclamptic pregnancies.
        Arch Gynecol Obstet. 2020; 302: 1103-1112
        • Staff A
        • Benton S
        • von DP
        • Roberts J
        • Taylor R
        • Powers R
        • Charnock-Jones D
        • Redman C
        Redefining preeclampsia using placenta-derived biomarkers.
        Hypertension. 2013; 61: 932-942
        • Sugitani M
        • Fujita Y
        • Yumoto Y
        • Fukushima K
        • Takeuchi T
        • Shimokawa M
        • Kato K.
        A new method for measurement of placental elasticity: Acoustic radiation force impulse imaging.
        Placenta. 2013; 34: 1009-1013
        • Wolf I
        • Vetter M
        • Wegner I
        • Büottger T
        • Nolden M
        • Schüobinger M
        • Hasten-teufel M
        • Kunert T
        • Meinzer HP.
        The medical imaging interaction toolkit.
        Med Image Anal. 2005; 9: 594-604
        • Yin M
        • Manduca A
        • Romano AJ
        • Glaser KJ
        • Drapaca C
        • Lake DS
        • Ehman RL.
        3-D local frequency estimation inversion for abdominal MR elastography.
        Proceedings of the International Society for Magnetic Resonance in Medicine. 2007; 15: 960
        • Yuksel MA
        • Kilic F
        • Kayadibi Y
        • Alici Davutoglu E
        • Imamoglu M
        • Bakan S
        • Mihmanli I
        • Kantarci F
        • Madazli R
        Shear wave elastography of the placenta in patients with gestational diabetes mellitus.
        J Obstet Gynaecol. 2016; 36: 585-588
        • Zahiri-Azar R
        • Salcudean SE.
        Motion estimation in ultrasound images using time domain cross correlation with prior estimates.
        IEEE Trans Biomed Eng. 2006; 53: 1990-2000
        • Zeng Q
        • Honarvar M
        • Schneider C
        • Mohammad SK
        • Lobo J
        • Pang EHT
        • Lau KT
        • Hu C
        • Jago J
        • Erb SR
        • Rohling R
        • Salcudean SE.
        Three-dimensional multi-frequency shear wave absolute vibro-elastography (3D S-wave) with a matrix array transducer: Implementation and preliminary in vivo study of the liver.
        IEEE Trans Med Imaging. 2020; 40: 648-660