On-site strip-type drug test kit enhanced by AI-based image analysis

Area of Science:

• Forensic Science
• Analytical Chemistry
• Artificial Intelligence

Background:

• Increasing demand for rapid, precise field drug detection due to beverage tampering incidents.
• Existing strip-type drug tests rely on subjective visual interpretation, leading to variability.
• Need for objective, quantitative methods to improve drug test kit reliability.

Purpose of the Study:

• To develop and validate a user-friendly, portable strip-type drug test kit.
• To implement an AI-based image analysis method for objective result interpretation.
• To enhance the automation and reliability of field drug detection.

Main Methods:

• Development of a novel strip-type drug test kit minimizing hazardous reagents and waste.
• Utilized artificial intelligence (AI) image analysis techniques for quantitative evaluation.
• Trained and evaluated deep-learning models (CNN, ViT, CNN-GRU) on 7,295 kit images.

Main Results:

• The CNN-GRU model demonstrated superior performance in accuracy, precision, recall, F1-score, and Cohen's kappa (1.0000).
• AI algorithm achieved perfect accuracy, exceeding the performance of 32 police officers.
• Validated AI-based precision reading algorithms for strip-type drug test kits.

Conclusions:

• AI-based precision reading algorithms significantly enhance automation and reliability of strip-type drug test kits.
• The developed method offers a reliable solution for field drug detection in forensic and investigative settings.
• This technology has strong potential for improving accuracy and reducing errors in drug testing.