The future application of artificial intelligence to pathology diagnosis
Junya Fukuoka, Kishio Kuroda, Tomoi Furukawa, Tomoya Oguri · Annals of Oncology · 2018
Artificial intelligence, especially deep learning (DL) using convolutional neural network has been recognized as a useful and powerful tool for image analysis, and applications to radiology, endoscopy, and pathology is now considered a promising movement. More than 100 publications are found by the key-word search of “Pathology” and “Deep learning”, recently. Among them, one published with a strongest impact to the medical society was a manuscript from JAMA. In the paper, 129 slides of whole slide imaging from lymph nodes with and without breast cancer metastasis were compared between deep learning platforms and pathologists, which showed specificity and sensitivity of diagnosis by DL was almost identical to pathologists' diagnosis. In the lecture, recent progress of pathological AI will be introduced by literature review and introduction of our in house data. Then, I will touch upon to the topics of future aspect of AI diagnostics especially in the pathology field and will include big issue of workflow change in pathology at Post-AI era. As a matter of course, the future requirement for pathologists, and probably for clinicians, will change rapidly. This indicates us a need of vision of how to develop new generation doctors.