Automatic analysis system of calcaneus radiograph: Rotation-invariant landmark detection for calcaneal angle measurement, fracture identification and fracture region segmentation

Calcaneus is the largest tarsal bone to withstand daily stresses of weight-bearing. The calcaneal fracture most common type in fractures. After a suspected, plain radiographs should be taken first. Bohler's Angle (BA) and Critical Gissane (CAG), measured by four anatomic landmarks lateral foot radiograph, can guide diagnosis facilitate operative recovery fractured calcaneus. This study aims develop an analysis system that automatically locate landmarks, measure BA CAG for assessment, identify calcaneus, segment regions. For landmark detection, we proposed coarse-to-fine Rotation-Invariant Regression-Voting (RIRV) detection method based on regressive Multi-Layer Perceptron (MLP) Scale Invariant Feature Transform (SIFT) patch descriptor, which solves problem fickle rotation By implementing novel normalization approach, RIRV explicitly rotation-invariance comparing with traditional methods. identification segmentation, convolution neural network (CNN) U-Net auxiliary classification head (U-Net-CH) designed. input ROIs CNN are normalized detected uniform view, orientation, scale. advantage this approach multi-task learning combines segmentation. Our accurately mean angle error 3.8? 6.2? respectively. region our presents good performance F1-score 96.55%, recall 94.99%, segmentation IoU-score 0.586. A powerful radiograph including anatomical angles measurement, identification, built. aid orthopedists improve efficiency accuracy calcaneus diagnosis.