Research On Lung Tumor Detection Algorithm Based On Convolutional Neural Network

With the development of computer vision,auxiliary treatment based on image detection technology in the medical field has gradually entered people’s field of vision.The image detection technology assists doctors in the investigation of lesions,which effectively improves the efficiency of doctors.This paper focuses on the lung tumor detection algorithm based on convolutional neural network.The accuracy of the detection algorithm is improved by designing a convolutional neural network that facilitates image feature extraction and feature learning.This paper studies the lung tumor detection algorithm from the following two aspects.1.A dual-path detection network based on multi-scale feature extraction.Aiming at the problem that the current convolutional neural network is insufficient for the extraction of feature information;this paper designs a feature extraction network that can extract multi-scale feature information based on the residual network.By dividing the input feature information into blocks,each block performs convolution feature extraction separately,and the extraction results of each step are input into the next step to fuse with the original input features.Finally,the feature extraction results of each step are spliced to obtain the output feature map.Feature extraction in this way can not only extract deep feature information,but also realize the multiplexing of multi-layer feature information.These all improve the detection network’s ability to extract features and utilize the input feature information.According to the ten-fold cross-validation results trained on the lung tumor dataset,the dual-path detection network model with multi-scale feature extraction is better than the original network.2.Image detection network based on CBAM attention mechanism.The attention mechanism is a feedback mechanism that can feed back the feature learning in the deep learning network to speed up the convergence of the deep learning network.In this paper,the CBAM attention mechanism is added to the lung tumor detection network to generate the weight matrix of the channel relationship of the feature information and the weight matrix of the spatial region relationship.The convolutional neural network will preferentially learn high-weight features,which improves the feature learning ability of the network model.And the attention mechanism can speed up the convergence of the network model and reduce the training time.The experimental results show that the performance of the multi-scale dual-path detection network with CBAM attention mechanism is better than that of the original network,which proves the feasibility of the attention mechanism.