Research On The Key Technology Of Teaching Lung Cancer Medical Imaging Based On Deep Learning

Lung cancer has one of the highest incidence and mortality rates in the human body,with nearly 5 million people dying of lung cancer each year worldwide.Early diagnosis and treatment of lung cancer will greatly improve patients’ chances of survival.Numerous clinical data show that lung nodules are the initial manifestation of lung cancer.In the traditional medical imaging teaching,it mainly relies on the abstraction and labeling of lesions by teachers,which is not convenient for students to grasp the pathological features visually and graphically,much less to formulate effective treatment plans.With the wide application of computer-aided medical imaging technology in teaching,these problems have been better solved.However,pulmonary nodules have different morphology and often adhere to other tissues,resulting in too complex backgrounds and blurred boundaries,which makes the accuracy of extracting lesion features by computer imaging technology low,and the differences between benign and malignant pulmonary nodules are small and the features have high similarity,resulting in low classification accuracy.This study addresses the problems and key technologies that arise in the teaching of computer imaging technology,and takes CT image data commonly used for lung nodules as the object of study.First,this study proposes the AVnet network model for segmentation of lung nodule lesions.the AVnet model is based on the Vnet network with more lesion edge information preserved by adding residual blocks to the U-net model,and introduces an attention mechanism in feature fusion to improve the model’s focus on important features and increase the model generalization ability and robustness.Second,the proposed L-VGG model improves the classification accuracy of benign and malignant lung nodules.The proposed L-VGG model is based on the VGG16 model and introduces the idea of layer fusion with reference to the residual network to achieve the fusion of multiple features to better learn the feature information of lung nodules and improve the accuracy of benign and malignant lung nodules classification task.Finally,this study design a computer-aided lung cancer diagnosis teaching system with AVnet segmentation network and L-VGG classification network as the core algorithms,so that medical students can have a more intuitive and comprehensive understanding of lung cancer lesions,and thus can master the skills of diagnosing lung cancer more quickly and accurately.In this study,the performance of the proposed algorithm was verified by the LUNA16 dataset of lung CT images,and the AVnet network had higher accuracy,and the three evaluation indexes of MIou,PA,and Dice achieved 89.8%,91.5%,and 0.895 results,respectively;the three evaluation indexes of L-VGG’s classification ACC,SEN,and AUC achieved 0.855,0.903,0.915 results.By comparing the segmentation results of 3D U-net,Vnet and AVnet,the AVnet network was proved to have better robustness for the lung nodule segmentation task.the comparative classification results of VGG16,VGG19 and the L-VGG network proposed in this paper were also tested to have higher classification accuracy and combined the AVnet segmentation network and the L-VGG classification network The computer-aided lung cancer diagnosis system worked well.