Bone Fracture Detection Based on Faster R-CNN with Bi-Directional Feature Pyramid Module
Wenlong Wu, Qian Yang, Zhihao Su
Abstract
With the widespread application of deep learning technology in the field of medical image processing, research on using convolutional neural networks for fracture detection has gradually increased. This paper proposes a Faster R-CNN network based on a bidirectional feature transfer mechanism (Bi-FPM), aiming to improve the accuracy of fracture detection. Through experimental comparison on the Kaggle Bone fracture dataset, this study explores the performance of different feature pyramid structures in the Faster R-CNN framework. Using ResNext-101 as the backbone, the proposed Bi-FPM method improves the accuracy by 5.8% in the fracture detection task compared to the baseline Faster R-CNN, and is higher than the traditional Feature Pyramid Network (FPN) Out 4.1%. Experimental results show that Bi-FPM’ s bidirectional feature transfer mechanism can effectively integrate deep and shallow features, improve the model's ability to identify fracture features and achieve better performance in fracture detection tasks. Through testing on the Kaggle Bone fracture dataset containing 350 test samples and 150 training samples, this paper confirms the effectiveness and superiority of the proposed method.