Transfer Learning of Pre- Trained Inception-V3 Model for Colorectal Cancer Lymph Node Metastasis Classification

Abstract

Colorectal cancer (CRC) lymph node (LN) metastasis is a critical index for CRC pathological staging. Surgeons determine the treatment plan based on magnetic resonance imaging (MRI) of CRC LN. However, current methods need to manually extract features, the accuracy depends on the feature extraction. In the paper, we present a method which use pre-train Inception-v3 model to classify CRC LN, the model could automatically extract feature from MRI of CRC LN. Experiments verified the performance of the model. A high classification accuracy is obtained, which is better than the results in the previous studies. The results suggest that the method can improve the accuracy of classification of CRC LN.