A01-KB108 Construction of Classyfier of Tumore Cell Types of Pancreas Cancer Based on Pathological Images Using Deep Learning
- Primary Investigator
- Naoaki Ono (Nara Institute of Science and Technology, Graduate School of Information Science, Assistant Professor)
- Kenoki Ohuchida (Kyushu University, Department of Medicine and Surgery, Graduate School of Medical Sciences, Assistant Professor)
- Chika Iwamoto, Kyushu University, Department of Advanced Medical Initiatives, Faculty of Medical Sciences, Post-doctroal Fellow
We are addressing to construct a computational model of tumor cell type classification of pancreas cancer. This system is based on feature extraction from tumor pathological images using Deep Convolutional Neural Networks (DCNN). Pancreas cancers are known as one of the most deadly cancer because their difficulty of early detection, rapid progression and easy metastasis. Therefore assisting their classification of their cell types or malignancy from their phenotypic features will provide useful clues to decide better treatment. In preceding study, we have developed an image analysis model of lung adenocarcinoma from pathological images using DCNN and showed that the model can distinguish their subtypes very accurately. In this study we will apply a similar DCNN model to extract cellular features from several different dyeing methods and analyze the correlation between their patterns. And we will develop a model for the clustering and prediction of their tumor types.