Automated Detection of a Blood Pool in Ultrasound Images of Abdominal Trauma

Abstract 

Ultrasound imaging is commonly used for emergency diagnosis of blunt trauma. Portable scanners are able to provide adequate imaging in remote and dangerous areas; however, medical expertise may not be available in the immediate local area to interpret the acquired images. The presence of pooled blood in the abdomen is a critical clinical symptom after trauma. This article describes an automated algorithm to detect blood pools in ultrasound images of abdominal trauma. The algorithm creates and uses a feature space consisting of local intensities, averaged local gradient magnitudes and second-order central rotation invariant moments. Successful tests were performed with a set of clinical images of a liver-kidney interface covering the Morrison’s pouch, which is the most likely space for blood from an abdominal injury to gather. When implemented in a portable scanner, the reported algorithm will provide rapid, on-the-spot detection of trauma-induced blood pooling and advance notice of a significant blunt traumatic injury. (E-mail: rshekhar@umm.edu)

Key Words: Trauma ultrasound, Feature detection, Abdominal blood pool, Blunt trauma