Research on Automatic Detection Algorithm for Lumbar Disc Herniation Based on FCOS

Lumbar disc herniation is a common spinal disease, and its diagnosis and treatment usually rely on medical imaging technology. Traditional medical image diagnosis methods require high labor and time costs. This article proposes an object detection network based on FCOS to achieve automated detection of lumbar disc herniation. This network introduces the idea of group convolution and incorporates attention mechanisms. In addition, the Soft NMS algorithm was used in the inference stage to avoid the problem of target boxes being violently deleted. The method proposed in this article achieved a mAP of 0.661 on the MRI test set, which has higher accuracy compared to other networks. This method provides a feasible and effective solution for detecting lumbar disc herniation, which is expected to assist doctors in clinical diagnosis.