Prediction Of Rectal Cancer Lymph Node Metastasis Based On Deep Learning

The incidence and mortality of rectal cancer are high,and although surgical resection can improve survival,if lymph node metastasis occurs in rectal cancer,preoperative adjuvant treatment is needed to obtain surgical opportunities,so it is necessary to predict lymph node metastasis in rectal cancer before surgery.This paper studies the prediction method of lymph node metastasis,the specific contents are as follows:Firstly,VGG16 network is applied by migration learning to rectal site CT image classification,after fine-tuning the VGG16 network.Secondly,image segmentation is performed on the CT image with the tumor.This paper applies U-net network and spatial attention mechanism for image segmentation.Through the result comparative analysis,the U-net network algorithm that increases the spatial attention mechanism is more suitable for dividing the rectal tumor images,which yields a higher Jaccard coefficient and a faster convergence rate.Finally,the segmentation mask map based on the image segmentation,taking the tumor site out as training samples using Open CV,and lymph node transfer prediction is performed based on a deep learning algorithm.This paper studies lymph node transfer prediction based on two algorithms: custom 2D convolutional neural networks and custom 3D convolutional neural networks.Through comparative analysis,3D convolutional neural network algorithm can better predict lymph node metastasis in rectal cancer.The prediction results of 3D convolutional neural networks are better than those of 2D convolutional neural network algorithms in terms of convergence rate and accuracy.