Automatic Diagnosis of Rectal Cancer Based on CT Images by Deep Learning Method

Rectal cancer is one of the most common malignant tumors in the digestive tract. Rectal cancer often invades the extra intestinal space and metastases to lymph nodes and other organs. Computed tomography (CT) scans are performed to determine the location and status of the tumor. In this paper, the CT image of rectal cancer was analyzed using deep neural network (DNN) and back propagation (BP) neural network. The mask image was extracted from the CT image of rectal with DNN. A Dice coefficient was used to measure the similarity between the actual cancer region and the region segmented by DNN. Based on the segmented mask image, the features of tumors were extracted and a metastasis evaluation model established to determine lymph node metastasis of the rectal cancer. The experimental results have shown that the mean value of Dice coefficient was 96.7%, the diagnosis rate of rectal cancer was 100%, and the identification rate of lymph node metastasis was 91.6%.