Improving Doctor-Patient Interaction with ML-Enabled Clinical Note Taking

Опубликовано 14 октября 2019, 16:33

While the majority of hospital appointments in the ambulatory setting as well as General Practitioners use Electronic Health Record (EHR) systems, these are poorly designed for use during patient encounters. The introduction of EHRs in hospitals is often met with resistance from clinicians, due to the increase in administrative burden: with EHRs many specialists spend more time documenting in EHR systems than on patient interaction. We believe that the solution is smarter bedside technology that can free up doctors to communicate naturally with patients -- while notes are taken through a combination of ML-based technologies that analyze the conversation as well as brief hand-written notes for information the clinician may not wish to verbalize.

As a proof of concept we developed PhenoPad -- an ML-based clinical tool enables the digitization of highly structured patient data and facilitates patient interaction, while providing physicians enough freedom to perform their jobs efficiently. Information is captured via a variety of modalities (video, speech, stylus, or typing) and quickly and unobtrusively presented to a physician for validation on a mobile device (tablet). To improve the quality of the captured data we have designed Deep Learning based methods for speaker diarization, extraction of key clinical terms from clinical text, and abbreviation disambiguation. All information extracted from notes is presented in real-time for clinical verification, and can be exported to EHR systems, enabling clinicians to pay more attention to patients and their concerns, and provide better quality care.

Talk slides: microsoft.com/en-us/research/u...

Learn more about this and other talks at Microsoft Research: microsoft.com/en-us/research/v...