ABOUT ISB
ISB Experience Tailored to you
Help us tailor the website experience to your taste. Make some selections to let the website work better for you.
I am Interested in
Suggestions for You
Kumar S, Arora P, Chagarlamudi T.R, Shriram K, Sur S, Kumar V
In the evolving field of medical data management, unstructured data sources like clinical notes, electronic health records, and medical literature present significant opportunities for enhancing healthcare. This project focuses on developing a technological framework that aids healthcare professionals, particularly doctors and hospitals, in managing patient records and making informed clinical decisions. The team used framework like a chatbot-like UI for querying patient histories and a portal for accessing archival medical data, all integrated with large language models (LLMs) to provide contextually relevant, citation-supported responses.
The project initiative aims to revolutionize clinical decision-making and patient record management, enhancing efficiency and accuracy to empower healthcare providers. By leveraging advanced technologies, the application ensures patient histories are comprehensive, accessible, and insightful.
The team employed a diverse technology stack including Amazon Textract, Comprehend, MySQL RDS, GPT-4, NEO4J, Falcon, MedLlama, Biomedical-NER[35], Streamlit, and Langsmith. These tools synergistically process medical data from diverse sources like electronic health records and medical articles, ensuring robust data extraction, summarization, and question answering capabilities.
They used key methodologies including Amazon Textract and Comprehend for extracting and understanding medical information, Biomedical-NER[35] for named entity recognition in medical texts, advanced language models like GPT-4 and MedLlama to facilitate medical question answering and text summarization, and the Falcon-7b model for enhanced summarization tasks. They used Langchain framework to integrate these tools for accurate and efficient outcomes, monitored by Langsmith for performance evaluation.
The application developed by the team enhances information retrieval by quickly summarizing medical topics, allowing healthcare professionals to access essential knowledge without extensive research. It offers domain-specific question answering, providing context-specific responses to aid decision-making. The interactive design promotes continuous learning, with tailored experiences based on user personas and preferred information sources, making it highly relevant for individual professionals. The user-friendly interface ensures accessibility for a wide range of medical practitioners, while the quick access to expert-generated summaries and answers significantly supports critical decision-making and saves time, allowing professionals to focus on patient care.
The application represents a significant advancement in medical informatics, combining cutting-edge technologies to streamline clinical workflows and improve patient care. With ongoing advancements, fine-tuning the machine learning model, robust monitoring, and adherence to high governance standards, it promises to continually meet and exceed the healthcare industry's needs.
