Deep Learning-based Image Recognition Technology in Medical Diagnosis

Abstract In the context of the computer era, the system of deep learning and other intelligent algorithms to assist medical diagnosis is gradually applied in the diagnosis and treatment of medical clinical diseases. To further study the application value of intelligent algorithms in medical diagnosis, this paper analyzes the depth of deep learning algorithms. Then, for the needs of medical diagnosis, the network architecture of convolutional neural is improved on the basis of deep learning algorithms, and then the MobileNet V2 network model is constructed by using residual neural, and a distraction (SA) mechanism module is introduced for image recognition and classification. This paper presents a CAD medical diagnosis system that uses deep learning image recognition to facilitate its use in medical clinical diagnosis. The optimized network model in this paper achieves smooth accuracy and loss rate of approximately 0.975 left and -0.969 in training. Compared with other network models, the four indexes of frequency weight intersection and merger ratio (0.904), mean pixel accuracy (0.881), background intersection and merger ratio (0.941) and diagnostic region intersection and merger ratio (0.807) of this paper’s model are all optimal. The F1-score evaluation indexes for image recognition in all three lung diseases in this paper reached more than 97%, and the AUC was as high as 99%. At the same time, the sensitivity of the CAD system in this paper is much higher than that of other systems, and the image recognition algorithm, as well as the CAD system designed in this paper, can improve the diagnostic efficiency of the primary health care units, and it can provide a reference for the detection of diseases.