We have one paper accepted to MDPI Applied Science.  

[Title]
Automatic Detection and Segmentation of Thrombus in Abdominal Aortic Aneurysm Using Mask Region-based Convolutional Neural Network with Optimized Loss Functions

[Authors]

Byunghoon Hwang, Hyoseok Hwang (Corresponding Author)
– Department of Software Convergence, Kyung Hee University

Jihu Kim, Sungmin Lee, Younhyun Jung (Corresponding Author)
– Department of Software, Gachon University

Eunyoung Kim, Jeongho Kim
– Department of Radiology, Gachon University Gil Medical Center,

[abstract]

The detection and segmentation of thrombus regions are essential for monitoring disease progression on abdominal aortic aneurysm (AAA) and patient care and management. With the inherent capabilities to learn complex features, deep convolutional neural networks (CNNs) have been recently introduced to improve thrombus detection and segmentation. However, the investigation with CNN methods is in an early stage, and most of the existing methods are heavily concerned with the segmentation of the thrombus region, which only works after its detection process. In this work, we propose a fully automated method for the whole process of thrombus regions based on a well-established mask region-based convolutional neural network (Mask R-CNN) framework, where we improve the Mask R-CNN with optimized loss functions. Combined use of complete intersection over union (CIoU) and smooth L1 loss is designed for accurate thrombus detection, and the thrombus segmentation is improved with a modified focal loss. We evaluate our method using clinically approved 60 patient studies (i.e., computed tomography angiography (CTA) image volume data) by conducting 4-fold cross-validation. Comparison results with multiple state-of-the-art methods suggest the superior performance of our method by achieving the highest F1 score of 0.9197 for the thrombus detection and outperforming most metrics in the thrombus segmentation.