Information scientists develop method to detect doping cases using artificial intelligence

Thousands of athletes are currently competing for medals at the Paris Olympics. And in some cases, questions will be asked about whether the medals were won fairly or whether doping was involved. Software developed by a team led by Wolfgang Maaß, a professor of business informatics at Saarland University, could help answer these questions in future competitions.

The software, currently being demonstrated at the International Joint Conference on Artificial Intelligence (August 3-9 in South Korea), needs just a few pieces of data to predict with unprecedented accuracy which athletes were definitely not doping — and in doing so can identify cases that warrant closer scrutiny.

Looking for a needle in a haystack or tilting at windmills are good metaphors for the challenge of detecting athletes who are doping. With thousands of athletes competing in major sporting events such as the Olympics, World Cups or professional leagues such as soccer, it can take a lab weeks to analyze urine samples to determine whether any of them have taken performance-enhancing drugs.

“Currently, all samples are analyzed manually,” said Maaß, professor of information systems for the service industry at Saarland University and scientific director of the research department for intelligent service engineering at the German Research Center for Artificial Intelligence (DFKI).

Given the huge number of athletes competing in major events like the Olympics — there are around 10,500 in Paris — and how time-consuming current testing methods are, it’s easy to see that many cheaters are simply getting away with it.

Only a fraction of urine samples can be analyzed in a lab. As we know from the doping scandal at the 2014 Winter Olympics in Sochi, some cheating athletes try to swap their own urine samples for “clean” samples provided by someone else.

Until now, DNA analysis has been the only reliable method for identifying whether samples have been switched. “But that is both expensive and time-consuming,” Maaß explained.

It is simply not possible to analyze the DNA of every single sample. Maaß and other colleagues from the DFKI (German Research Center for Artificial Intelligence), the German Sports University Cologne and the World Anti-Doping Agency (WADA) decided to combine their experience to find a simpler, more feasible solution. “This problem simply cries out for machine analysis,” Maaß said.

To solve this problem, they developed software that uses artificial intelligence to analyze data from urine samples both quickly and cost-effectively. “Doping tests measure the concentrations and ratios of different steroids, which are then checked for plausibility,” Maaß explained. This provides a biochemical fingerprint that Saarbrücken AI’s software can use to reliably flag any anomalies.

The machine learning program only needs data from three urine samples provided by each athlete during their career. Because one athlete’s natural steroid profile can be very different from another’s, the program learns what concentrations of specific substances are typical for a given athlete.

For each sample in the biochemistry lab, seven characteristics are determined, such as steroid concentrations and ratios. Like a child solving a spot-the-difference picture puzzle, the software looks for deviations from the usual pattern.

“If you compare three or more ‘images’ with the measurement data from the individual urine samples, the software finds the ones where everything matches,” Maaß said, explaining in simple terms how the computer program works. There remains a residual number of samples in which the ‘images’ do not match, i.e. in which inconsistencies were detected.

“The small number of remaining cases can then be investigated in more detail by biochemists in the laboratory using DNA analysis. If an athlete has taken a performance-enhancing substance and this substance can be detected in urine, our software can help identify this athlete with a high degree of certainty,” Maaß said.

Rather than directly detecting doping offenders, the software is designed to identify clean athletes with 99% certainty to ensure that innocent people are not wrongly accused. While this may mean that a small number of doping offenders go undetected, positive doping cases, where athletes have taken banned substances to achieve higher, further or faster results, are identified with a very high degree of certainty.

“There will certainly be a doping case among the remaining cases, and they can then be investigated in more detail using DNA tests,” Maaß explained.

With this innovative and incredibly accurate method, finding a needle in a haystack can become a lot easier.