Professor Wen-Yu Hu's team at National Taiwan University Hospital has developed a "Smart Medication Verification System" that leverages AI image recognition technology to accurately verify whether the medication selected by nurses matches the electronic medical orders. This system aims to significantly reduce medication errors and enhance patient safety.

Overcoming Manual Verification Limitations: Medication Check Completed in 3 Seconds!

Medication errors are a significant challenge for global healthcare systems. In Taiwan, nearly 20,000 medication error incidents are reported annually, with approximately 20% occurring during the nursing administration stage. Traditional manual medication verification is prone to errors due to personal and environmental factors, leading to mistakes in drug names, dosages, frequencies, and timing. The smart medication verification system uses AI image recognition technology to complete medication checks within just 3 seconds, effectively addressing the limitations of traditional methods and reducing the risk of medication errors.

The core of this system is an AI image recognition model that combines deep learning and convolutional neural networks (CNN). By collecting images of over 400 types of medications, totaling tens of thousands of images, and meticulously labeling detailed information (including drug names, dosage forms, and doses), the system enhances training data diversity through techniques like rotation, scaling, and cropping. This has resulted in a comprehensive training database covering the common appearances of hospital medications.

Key features of the Smart Medication Verification System include:

1. Visualized Medication Tray: Clearly displays medications placed in the tray, accommodating various forms such as oral drugs, injections, and topical ointments, facilitating nurses’ review.
2. AI Image Recognition Verifier: Quickly compares medication images with electronic medical orders using AI image recognition, ensuring accurate medication administration and issuing immediate alerts when errors are detected.
3. Medication Guidance Function: Automatically generates multilingual medication instructions and sends medication tray photos to patients and their families, enhancing patient engagement in their healthcare.
4. Quality Control Monitoring and Teaching Materials: Records each verification result and automatically exports error cases to create teaching materials for medical staff, continually improving medication safety.
5. API Integration, High Compatibility: Easily integrates with existing electronic medication systems at medical institutions through API without altering original systems or operational habits, facilitating adoption and use.

Successful Initial Clinical Verification with Over 90% Accuracy!

Professor Hu stated that the team’s goal is to develop a practical and easily promotable medication assistance tool to enhance patient medication safety. This system not only reduces medication errors but also improves nursing quality and lowers healthcare costs, offering high application value. The system has been successfully verified in initial clinical trials at National Taiwan University Hospital, achieving a recognition accuracy rate of over 90%. The team plans to continue optimizing system functions and collaborate with other medical institutions to extend this technology’s application to broader fields.

Resource (mandarin) :
防範藥命錯誤！台大智慧藥物檢核系統 AI圖像3秒辨識給藥正確性！