Shape-based kidney detection and segmentation in three-dimensional abdominal ultrasound images

Abstract

Due to recent technical advancements of three-dimensional ultrasound imaging systems, applications of this imaging modality have been expanding from the fetal imaging to cardiac- and abdominal-diagnosis. Among all internal organs, diagnosing the kidney has a paramount importance for rapid bedside treatment of trauma and kidney stone patients using ultrasound images. Although three-dimensional ultrasound provides higher level of structural information of kidneys, manual kidney diagnosis using three-dimensional ultrasound images requires a highly trained medical staff, due to the extensive visual complexity which three-dimensional images contain. Therefore, computer aided automated kidney diagnosis becomes very essential. Due to the challenging problems of ultrasound images, such as speckle noise and inhomogeneous intensity profile, kidney segmentation in three-dimensional ultrasound images has not been sufficiently investigated by researchers. In this paper, we first propose a new automated kidney detection approach using three-dimensional Morison's pouch ultrasound images. Then, we proposed a shape-based method to segment the detected kidneys. A preprocessing step is utilized to overcome the ultrasound challenges. Based on a set of 14 ultrasound volumes, we have evaluated the detection rate of our proposed kidney detection approach which is 92.86%. For kidney segmentation, we compared our proposed method with an existing approach, and the performed statistical analysis strongly validates the superiority of our proposed method with p = 0.000032.

Computer-Assisted 3-D Ultrasound Probe Placement for Emergency Healthcare Applications

“…Specifically, ultrasound imaging is the most popular tool to scan abdominal organs for triaging hemodynamically unstable patients with abdominal trauma [2]. The use of ultrasound imagery is preferred for emergency diagnosis over other popular imaging modalities, such as Magnetic Resonance Imaging (MRI) and Computed Tomography (CT), because ultrasound imaging provides the following advantages: (1) ultrasound imaging does not impose any risk to patients; (2) since ultrasound devices are portable, unstable patients are not required to be moved from a resuscitation room to an imaging room; (3) ultrasound imaging provides immediate results, which accelerates triaging [3]. Two-dimensional (2D) ultrasound imaging is widely used by radiologists for imaging abdominal organs.…”

“…In contrast, 3D ultrasound is less popular, especially in abdominal scans, because its data structural complexity makes it difficult for radiologists to interpret. However, localizing internal organs in the 3D space is only possible using 3D ultrasound, which is essential for new computer-assisted methods [3].…”

“…Abdominal organ detection in 3D ultrasound has been addressed rarely by researchers [3], [8], because processing of 3D ultrasound images faces the following challenges [3]: • Ultrasound-specific challenges: Speckle degrades the quality of images. Low contrast and inconsistent intensity profile reduces detectability of organs regions; • Organ-specific challenge: Due to the adjacency of organs with high scattering tissues, shadows may partially occlude the organs, leading to incorrect decisions; • Operator-specific challenge: ultrasound probe misalignment due to operators' inexperience results in partial visualization and mis-characterization of organ's shape.…”

“…This method simplifies the kidney anatomical structure into an elliptical shape, resulting in false detections. In another attempt, Marsousi et al [3] reported the shape-based kidney detection method in 3D ultrasound images. In this method, a kidney shape model is generated from a training set, and it is used to find a location inside a 3D ultrasound image with the maximum cross-correlation metric.…”

“…In this method, a kidney shape model is generated from a training set, and it is used to find a location inside a 3D ultrasound image with the maximum cross-correlation metric. The method in [3] is prone to fail when the organ shape in an acquired image is deformed by orientation and scaling, with respect to a reference shape.…”