A Novel Lung Nodule Detection and Recognition Model Based on Deep Learning

To solve the problems of missing and false detection of pulmonary nodules in complex lung environments, as well as trivial and inefficient detection procedures, an end-to-end pulmonary nodules detection and recognition model based on deep learning was proposed. Innovation and improvement are made on the basis of YOLOv5. In the feature extraction stage of the model, a convolutional structure integrating self-attention mechanism is proposed to capture the global feature and the dependence relationship of long-distance information, and screen the key pathological information. Then, a convolution structure integrating internal convolution operators is proposed to reduce the computational redundancy in the feature channel and improve the inference speed of the model. In the feature fusion stage of the model, the structure of cross-scale coordinate attention feature fusion is proposed, and the different features enhanced with attention are weighted by jumping links to promote the fusion of multi-scale feature information. The proposed model obtained 97.8% mAP@0.5 indexes in the self-built diagnosis and treatment data set of pulmonary nodules in Huaihai area. The pulmonary nodule detection model proposed in this paper can significantly reduce the false positive rate and obtain the location and classification results of diseased nodules with higher detection accuracy and faster detection speed, which has important practical value in clinical application.