Clinical decision support system for comprehensive analysis and long-term surveillance of post-endovascular repair in abdominal aortic aneurysms

초록

An abdominal aortic aneurysm is a critical vascular condition involving the continuous dilatation of thrombus, even after surgical intervention such as endovascular aneurysm repair. Clinical decision support systems (CDSSs) can enhance the comprehensive analysis and long-term surveillance of post-operative complications and outcomes. However, research on their development and implementation remains largely untapped. In this work, we design, develop, and evaluate a novel CDSS for post-operative care in collaboration with two senior domain experts. Our image processing module enables automated, precise, and rapid detection and segmentation of regions of interest (ROIs) containing thrombus in post-operative computed tomography angiography (CTA) image studies. It utilizes both mask region-based convolutional neural network (Mask R-CNN) and U-Net++ models and optimizes their sub-networks and loss functions. In addition, the longitudinal knowledge exploration module goes beyond analyzing individual time points (i.e., individual CTA studies) to identify changes in thrombus ROIs over a sequence of studies. It also highlights crucial diagnostic indicators, such as endoleaks and ruptures, and provides synchronized visualization and tracking. Our experiments demonstrate that the image processing module achieves high accuracy when compared to five CNN models across 48 patient studies. The longitudinal knowledge exploration module is validated through informal expert reviews of two representative clinical cases, with positive feedback regarding its clinical feasibility and usability. The CDSS is shown to be computationally efficient, thereby allowing accessibility on standard laptops for routine use. To promote widespread adoption, we are releasing it as a plug-in extension for the well-established 3D Slicer platform.

키워드