Project Description:
"The AI Scribe" project addresses the prevalent challenge of documentation overload among healthcare professionals, a leading cause of dissatisfaction and burnout. Conventional manual approaches to clinical note-taking are both resource-intensive and prone to errors. To alleviate this documentation burden, our product introduces an innovative solution integrating Whisper AI's speech-to-text capabilities and the BART (Bidirectional Auto-Regressive Transformer) model to automate the generation of History of Present Illness (HPI) from audio recordings of clinician-patient interactions. The process begins with real-time conversation capture, followed by automatic speech recognition models (ASR). Subsequently, employing advanced Natural Language Processing (NLP) techniques, we extract vital details such as symptoms and chief complaints, resulting in concise and clinically relevant HPI summaries. Using a diverse open-source dataset (aci-bench), our algorithm continuously refines model accuracy through iterative training. By combining these techniques, we were able to generate HPIs demonstrating high accuracy indicators through our performance metric “ROUGE,” which compares our generated summary to a gold standard written by a physician or medical scribe. Overall, "The AI Scribe" represents a transformative advancement that enhances documentation processes, reduces clinician burden, and fosters improvements in patient care quality.
