2nd Seminar of Circulation and Unitilization of Medical and Genome Data

Koiti Hashida PI

Title:PLR for Autonomous, Decentralized, and Collaborative Healthcare

Abstract: If your health record can be disclosed to medical organizations near
your new address or travel destination, for instance, then the safety
and efficacy of medical services will be improved. However, the
traditional method for data sharing centralized by a middleman other
than the data owner and user cuases a huge risk of data leakage while
spending a large cost for data management. A decentralized
data-sharing without such a middleman is far safer, much less
expensive, and versatile. PLR (personal life repository) realizes this
decentralized data sharing by storing encrypted data in basically free
online storages such as Google Drive. It has been already utilized for
elderly care and is about to be applied to healthcare in cooperation
with medical information systems such as EMR and EHR.

Sipp Doug PI

Title: Information quality issues in machine learning strategies for biomedical research

Abstract: Machine learning applications have achieved notable early successes in biomedical fields from rational drug design to radiological image analysis to predictive pathology. Large data sets encompassing omics, epidemiology, and individual medical histories have led to hopes that new deep learning techniques will facilitate a better understanding of the causes of disease, and the development of new therapies. However, the incompleteness, noisiness, and complexity of human biological systems, and the unavoidable difficulties in experimental validation of predictions from human data, present fundamental challenges to machine learning approaches. In this talk, I will introduce some of the problematic characteristics of human physiological and pathological systems, and briefly discuss implications of the use of machine learning in biomedical research.