TinyML-Based Lightweight AI Healthcare Mobile Chatbot Deployment

Area of Science:

• Artificial Intelligence in Healthcare
• Edge Computing
• Machine Learning for Medical Applications

Background:

• Artificial Intelligence (AI) is transforming healthcare, improving patient interactions and satisfaction.
• Online AI solutions pose risks to sensitive patient data security.
• Advancements in AI impact nursing, diagnostics, and critical medical procedures.

Purpose of the Study:

• To develop a secure, AI-driven chatbot for hospital environments.
• To address patient data security concerns using local processing.
• To improve the efficiency of patient data management and diagnostic support.

Main Methods:

• Utilized Edge Computing for secure, on-site data processing.
• Implemented Histogram of Gradient (HOG)-based classification for patient identification.
• Integrated TinyML for efficient processing of patient data, including temperature measurement and demographic recording via a chatbot.

Main Results:

• Achieved 95.8% accuracy in patient detection and 95.3% accuracy for the TinyML model.
• Demonstrated rapid on-device processing time (217 ms) and low latency (4 ms).
• The TinyML model requires minimal resources (8.8Kb RAM, 50.3Kb Flash memory).

Conclusions:

• The developed system offers enhanced security and faster patient data recording compared to existing AI solutions.
• All data processing occurs locally, ensuring patient data privacy and integrity.
• The AI chatbot facilitates secure data storage for hospital management and diagnostic purposes.