Research On Medical Image Lesion Detection Algorithm Based On Deep Learnin

With the rapid development of medical image analysis technology and the sharp increase in clinical imaging data,how to diagnose and classify of different diseases based in medical image is a major challenge faced by modern medical workers quickly and accurately.Using artificial intelligence technology to detect lesion with medical image can improve the work efficiency of doctors and relieve the pressure on hospitals.With the great success of deep learning in the field of computer vision,it provides a new idea for the task of medical imaging lesion detection.The deep convolutional neural network can efficiently extract the features of various lesions in medical images to achieve tasks such as locating and classifying lesions.Based on this,this paper proposes a medical image focus detection model based on deep learning.Applying this model to the detection and recognition of pneumonia and cardiac hypertrophy in chest X-ray images has significantly improved performance and has achieved relatively excellent results in focus detection and interpretability.The main work of this paper:(1)To highlight important intrinsic features related to the lesion classification task,in this paper,Spatial Attention Superposition module,which takes advantage of the channel and spatial attention mechanisms,was designed to identify lesions and their locations,and an Multilayer Feature Fusion module was designed to harmonize disparate features from different channels and emphasize important information.These two modules were concatenated to extract critical image features serving as the basis for disease diagnosis.(2)To improve the performance of the lesion detector,we further embedded the proposed modules into a baseline neural network and developed a model called SASMFF-YOLO to diagnose lesions.To validate the effectiveness of our model,extensive experiments were conducted on two medical images datasets provided by the Radiological Society of North America(RSNA)and the AI Research Institute.SASMFF-YOLO achieved a precision of 88.1%,a recall of 98.2% for classification and an AP50 of 99% for lesion detection on the AI Research Institute dataset.The visualization of intermediate feature maps showed that our method could facilitate uncovering detect lesion related lesions in medical image.(3)In addition,to prove the robustness of the module and improve the interpretability of the model,this paper designs SProto PNet to diagnose disease.This model simulates the human classification logic for image reasoning and has strong interpretability.Our results demonstrated that our approach could be used to enhance the performance of overall lesions detection in medical image and improve model interpretability.