Doc in a Box
Abstract
The goal of the Doc-in-a-box project is to develop systems that can help Navy medics and corpsmen in providing needed care to the sailors in a ship which may be deployed in a remote location. The developed systems need to address the challenge that a ship in such a location may have limited connectivity and limited computational specifications (computational power, memory, etc.).The developed systems need to help the Navy medic in avoiding misdiagnosis and incorrect treatment, in following current guidelines to the extent possible, and overcome - to the extent possible - lack of advanced machines and instruments, with the available resources in the ship. We aim to develop the project in three phases. In phase 1, our focus will be on developing an information retrieval and initial question answering system that will find relevant paragraphs in the document repository related to her queries and highlight answers to her specific questions in those paragraphs. The document repository will contain general medical and health information (such as fundamental textbook and Physicians Desk Reference) as well as health records of the sailors, if possible. In phase 2, we will incorporate more advanced question answering that can do differential diagnosis and pick the right condition when multiple conditions share many similar symptoms. We will also take into account genomic information. In phase 3 we will proceed to build a personal assistant for medics that can not only respond to medics requests but also proactively alert the medics, can have longer conversations with the medics and in general work together with the medics towards achieving a predefined set of goals. Our focus here is on phase 1. Working together with our partners, we will identify a set of documents that need to be included in the document repository. During the Information Retrieval (IR) and Question Answering (QA) system development phase we may not have accesses to the final document repository and hence we plan to use TREC datasets to develop, train and evaluate our systems. For demo we will use those datasets and freely available sources. We propose to use SolarSPELL Health - the Solar-Powered Educational Learning Library as an alternative hardware box for our digital Doc. Our team consists of PI Baral a professor in Computer Science with significant experience in Natural Language Processing, Question Answering, and Biomedical Informatics, and Co-PI Ross a practicing, doctorally-prepared nurse practitioner in cardiac electrophysiology.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 29, 2020
- Source ID
- N000142012332
Entities
People
- Chitta Baral
Organizations
- Arizona State University
- Office of Naval Research
- United States Navy