Tumor-assisted Diagnosis based on U-Net Network
Juntong Chen, Aizeng Cao, Yiming Zhang, Tao Xu · 2020
With the increase of the number of patients in China, the diagnostic efficiency of doctors has decreased, and the rate of misdiagnosis has increased. With the help of medical imaging intelligent assisted diagnosis technology, doctors can obtain a preliminary diagnosis through an assisted diagnosis system. On this basis, doctors further inspect and review the diagnosis results, which greatly improve the efficiency and accuracy of cancer diagnosis, and reduce the rate of misdiagnosis. In this paper, the system focuses on rectal cancer to do research. Whether there is lymph node metastasis in patients with rectal cancer has an important impact on the decision-making of the treatment plan and the prognosis of patients, so the accurate judgment of lymph node metastasis is an important step in the treatment of rectal cancer, but there is no way to accurately judge the lymph node metastasis before the operation. Therefore, in this paper, the proposed U-net-based tumor-assisted diagnosis system can evaluate the status of lymph node metastasis by judging the characteristics of CT images of rectal cancer. The data set used is the seventh "Teddy Cup" data mining challenge-data set B. The system uses the U-Net network structure, using a cross-entropy loss function and Adam optimizer. After adjusting the parameters, the experiment used a piece of NVIDIA TESLA K80 to train for 50 rounds at a learning rate of 0.001 for about two days and finally kept the accuracy rate at about 85%. Obtained results have demonstrated that the proposed system is promising.