GPT-Powered Medical Assistant for Summarizing IoT Patient Health Records

Abstract

The rapid expansion of IoT-enabled health monitoring devices has resulted in substantial amounts of unstructured patient data, creating considerable hurdles in clinical decision-making. This research introduces a GPT-driven medical assistant that autonomously condenses IoT-generated patient health records into succinct, clinically pertinent narratives. The system consolidates data from wearable sensors that monitor heart rate, oxygen saturation, temperature, and activity levels, utilizing a cloud-based architecture for processing. Optimized GPT-3.5 using 10,000 annotated IoT health records and assessed the system with ROUGE and BLEU criteria. The proposed model attained a ROUGE-1 score of 82.4, a ROUGE-L score of 79.6, and a BLEU-4 score of 67.2, improving baseline summarization models by 18–22%. In a clinical usability study including 25 clinicians, 88% deemed the summaries as "highly useful" for expediting patient evaluations. Furthermore, the assistant decreased the average review duration from 7.3 minutes to 2.1 minutes per patient record. These findings illustrate the capability of GPT-based summarization to improve healthcare efficiency and facilitate urgent medical decision-making.

Ähnliche Arbeiten