Katsuki, Takayuki (IBM Research AI, IBM Research - Tokyo) | Ono, Masaki (IBM Research AI, IBM Research - Tokyo) | Koseki, Akira (IBM Research AI, IBM Research - Tokyo) | Kudo, Michiharu (IBM Research AI, IBM Research - Tokyo) | Haida, Kyoichi (The Dai-ichi Life Insurance Company, Ltd.) | Kuroda, Jun (The Dai-ichi Life Insurance Company, Ltd.) | Makino, Masaki (Fujita Health University) | Yanagiya, Ryosuke (Fujita Health University) | Suzuki, Atsushi (Fujita Health University)
This paper describes a feature extraction technology from event sequence of lab tests in electronic health record (EHR) for modeling diabetic nephropathy. We used a stacked convolutional autoencoder which can extract both local and global temporal information from the event sequence. The extracted features can be interpreted as similarities to a small number of typical sequences of lab tests. The extracted features in our prototyping experiment were promising for understanding of the long-term course of the disease.
Apple's Heath Records feature, which allows iPhone users to download their medical records to their smartphone, is now available in the UK and is being used by two hospital trusts. The feature, part of Apple's Health app, gives iPhone users the ability to request and download their electronic health records as stored by hospitals using a direct, encrypted connection between their iPhone and the system used by the clinic. The Health app then periodically connects to pull across any new health records and notify the user when new records are available. Currently in the US over 500 institutions support the Health Records feature on the iPhone across 11,000 locations. In many cases, patients' medical records are held in multiple locations, requiring patients to log in to each healthcare provider's website to piece together their health information manually.