Intelligent Doping Diagnosis (iD2): Development of an AI-Based Diagnostic System for Accurate and Rapid Doping Detection in Athletes

Abstract

The aim of this research is to develop an automated diagnostic system based on artificial intelligence for the accurate and rapid diagnosis of doping in athletes. The current process of human evaluation of vast amounts of biological sample data to identify atypical results for banned substances is time-consuming, prone to human error, and requires expert opinion for accurate interpretation. New strategies are needed to overcome these challenges and improve the accuracy and efficiency of the assessment process. We developed an evaluation diagnostic system called Intelligent Doping Diagnosis (iD2), which uses a machine learning-based approach to analyze mass spectrometer data generated by mass spectrometers to identify atypical findings of banned substances. The study used data collected over four years at the Doping Control Center in South Korea, consisting of four data sets of retention times and MRM pairs for hundreds of doping substances, and a total of 145,544 data points was used. The accuracy of doping diagnosis was determined using ten classification algorithm models and the results showed an accuracy of 93% for the total group and 99% for the alpha, beta, and gamma groups. These results demonstrate the potential of using AI technology in anti-doping efforts to improve the accuracy, efficiency, and speed of the testing process. We suggest that iD2 be further developed to include more drugs, devices, metabolites, MRLs, and DLs, making it an effective next-generation anti-doping strategy.