Research on Abnormal Target Detection Method in Chest Radiograph Based on YOLO v5 Algorithm

Now, chest X-rays are a common test to detect lung diseases. With the development of medical imaging technology and the establishment of chest radiograph classification database, the rational use of image processing technology to study lung disease diagnosis has very important practical significance. For the detection of 14 lung diseases, this paper presents a method of abnormal target detection in chest X-rays based on YOLO v5 algorithm. The Vindr-CXR dataset provided by the Kaggle Competition is used to verify the accuracy of the YOLO v5 anomaly detection algorithm. Experimental result shows that compared with other anomaly detection algorithms, the YOLO v5 algorithm used in this paper has better performance. The metric is 7.28% and 5.89% higher than the Faster RCNN algorithm and the EfficientDet algorithm, respectively. This verifies the effectiveness of the algorithm in detecting abnormal targets in chest X-rays.