Leveraging Auxiliary Classification for Rib Fracture Segmentation

TL;DR

A sophisticated deep-learning model with an auxiliary classification task designed to enhance the accuracy of rib fracture segmentation is introduced, which aims to improve feature representation at the bottleneck layer by highlighting the regions of interest.

Abstract

Thoracic trauma often results in rib fractures, which demand swift and accurate diagnosis for effective treatment. However, detecting these fractures on rib CT scans poses considerable challenges, involving the analysis of many image slices in sequence. Despite notable advancements in algorithms for automated fracture segmentation, the persisting challenges stem from the diverse shapes and sizes of these fractures. To address these issues, this study introduces a sophisticated deep-learning model with an auxiliary classification task designed to enhance the accuracy of rib fracture segmentation. The auxiliary classification task is crucial in distinguishing between fractured ribs and negative regions, encompassing non-fractured ribs and surrounding tissues, from the patches obtained from CT scans. By leveraging this auxiliary task, the model aims to improve feature representation at the bottleneck layer by highlighting the regions of interest. Experimental results on the RibFrac dataset demonstrate significant improvement in segmentation performance.

Figures (6)

• Figure 1: The first row of images displays fractured regions of different sizes and shapes, appearing as irregularities and disruptions in the bone structure. In the second row, the ground truth annotations are overlaid on the fractured regions in the corresponding CT images from the first row.
• Figure 2: Schematic view of our proposed segmentation network with an auxiliary classifier at the bottleneck.
• Figure 3: Schematic view of CAM module.
• Figure 4: Baseline Models Comparison with volume and slice DSC: First row: Segmentation of Sample 431, Slice Number 268, Second row: Segmentation of Sample 440, Slice Number 159, Third row: Segmentation of Sample 464, Slice Number 278.
• Figure 5: Visualization of results: The first row displays the ground truth masks, the second row shows the class activation maps, and the third row presents the predicted heat maps.