Stethonotes

• 

  Home page

• 

  Record page

• 

  Results page

Inspiration

"EHRs were invented to improve the efficiency and quality of patient care, but they have yet to achieve that promise" - Susan Hingle, MD

One of the greatest challenges in medicine today is how to retain the classical human-to-human bond of a patient and a physician while continuing to implement new technologies that can help improve care and save lives. Current EHRs give this promise, but often end up harming the efficiency they are meant to improve. One of the biggest culprits is that in office visits, doctors must either spend most of their time typing away at a computer (hampering their ability to interact with the patient) or spend the money to hire and train scribes and charting professionals. We propose a new solution: an EHR system that automatically acts as scribe and charter, gleaning information directly from listening to the conversation, without noticeably interjecting itself like current computer screens.

What it does

Stethonotes records and transcribes the audio of a clinic or ER visit, then uses natural language processing to extract important information directly into the patient's chart, eliminating the need for physicians to spend valuable time and energy typing away at their computers and enabling them to spend more of their day working directly with patients.

Challenges we ran into

We initially attempted to use Microsoft's Azure hosting and LUIS API, but ran into roadblocks and eventually abandoned Microsoft in favor of Google.

What we learned

We learned about using Google APIs and also about how to use machine learning in natural language processing

What's next for Stethonotes

The next step for Stethonotes would be to more fully flesh out and test all specified features. The next significant feature we would like to add is showing color-coded confidence values on each extracted item, so the physician can easily see at a glance which lines should be manually checked. We would also like to store the recording so the doctor can manually listen to the lines in question.

Our end goal with Stethonotes is not an entirely self-sufficient EHR system. To be truly transformational, this technology would ideally integrate directly with Epic's or Cerner's systems (the dominant EHR providers in the current market). So, we would like to develop Stethonotes into a marketable proof-of-concept so that we could sell the technology to a company like Epic or Cerner.