Development of a Wearable Camera and AI Algorithm for Medication Behavior Recognition.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Digital Health

Background:

• Patient non-adherence to long-term medications for chronic conditions affects 40-50% of individuals.
• Limited medication monitoring hinders timely intervention and research modification.
• Accurate medication behavior recognition is crucial for patient care and clinical research.

Purpose of the Study:

• To develop and validate an AI-driven system for monitoring patient medication adherence using wearable camera sensors.
• To improve healthcare provider intervention capabilities through prompt medication error reporting.
• To enhance clinical research by enabling easier experiment modification and health prediction based on compliance.

Main Methods:

• Utilized a wearable device with a camera image sensor to record videos of patients taking medication.
• Applied convolutional neural network (CNN) techniques, including a Faster R-CNN object detection model (Model 1) and an action recognition model (Model 2).
• Trained the AI algorithm using a dataset of 50,000 labeled images from 89 participants.

Main Results:

• Achieved a high accuracy of 92.7% for medication behavior recognition through the combined AI models.
• Demonstrated the effectiveness of the integrated object detection and action recognition approach.
• Established a robust method for analyzing medication-taking behavior from video data.

Conclusions:

• The developed AI system offers a novel solution for monitoring medication adherence in real-time.
• High accuracy in recognizing medication behavior facilitates rapid intervention for healthcare providers.
• This technology has the potential to significantly improve patient outcomes and streamline clinical research.