The research group of Professor Shunpei Ishikawa of Tokyo Medical and Dental University, in collaboration with Professor Tatsuya Harada of the University of Tokyo Graduate School, uses artificial intelligence technology "deep learning" to accurately determine cancer cells from histopathological images. Developed an algorithm to do.In the international competition Cameryon 17 using this, we won the 4th place in the world in the accuracy of determining lymph node metastasis in breast cancer patients.The only prize among the participating teams in Japan.
The presence or absence of lymph node metastasis in breast cancer is an important factor in determining a patient's treatment strategy and prognosis.Usually, it is judged by observing the pathological tissue using a microscope.However, small lesions are easily overlooked, and there are differences in the diagnosis results by pathologists, so improvement by using IT technology has been sought.
In this study, in the histopathological images of lymph node tissue of breast cancer patients, about 30 images as training data from the cancer cell region and other regions were read into the "deep neural network (Note)" and networked. Was optimized.At that time, we succeeded in efficiently learning by extracting the characteristic information of the histopathological image from the intermediate layer of the neural network, and it became possible to make a highly accurate judgment.From this result, we created a map of the existence probability of cancer cells superimposed on the pathological image, and determined the presence or absence of metastasis for each lymph node and the stage of breast cancer patients. Achieved.
The results of this research showed that artificial intelligence technology is effective in detecting lymph node metastasis in breast cancer, and that essential morphological information on cancer cells and cancer tissues can be obtained.In the future, it is expected that the use of artificial intelligence will reduce the difference between facilities and achieve high accuracy for the pathological diagnosis of various cancers other than breast cancer.
(Note) A neural network is an information processing model that imitates the neural system of the brain, and a deep hierarchical structure is called a deep neural network.Deep learning is one of the artificial intelligence technologies using this.