A research team at the Korea Institute of Science and Technology (KIST) has developed an artificial intelligence (AI) system that uses urine testing to improve prostate cancer’s diagnosis accuracy.
Prostate cancer is one of the most common cancers in men, and hospitals conventionally use blood tests to determine the presence of prostate cancer. However, such tests’ accuracy only came around 30 percent, forcing most patients to undergo an invasive biopsy after a blood test and suffer side effects such as bleeding and pain.
To resolve the issue, the team, led by Professor Lee Kwan-hee at the university, developed a technology that diagnoses prostate cancer in the urine with near 100 percent accuracy in just 20 minutes. Professor Jung In-gap at Asan Medical Center also participated in the study.
The research team succeeded in developing the technology by using an AI analysis based on ultra-sensitive electrical signals.
“The diagnostic test using urine is excellent for patient convenience and does not require an invasive biopsy, allowing them to diagnose cancer without side effects or pain of the patient,” the team said. “Until now, the concentration of cancer factors in urine was too low, and urine-based biosensors have been used to classify risk groups rather than precise diagnosis. There was also a limit in raising the accuracy of diagnosis to more than 90 percent with a single cancer factor.”
The team overcame such problems by simultaneously utilizing different cancer factors instead of just one cancer factor.
In detail, the device can simultaneously measure four very small amounts of cancer factors in urine and improves the problems of the existing prostate-specific antigen (PSA)-based detection. Afterward, the team introduced artificial intelligence machine learning to correlate the four cancer factors obtained through this sensor and prostate cancer. It developed an algorithm that can diagnose cancer according to the complex pattern of the received detection signals.
As a result of diagnosing prostate cancer using the artificial intelligence analysis method, the team could diagnose 95.5 percent of prostate cancer patients in 76 urine samples.
The researchers said they also planned to increase the diagnosis algorithm’s accuracy by expanding clinical trials in the future to learn more patient information.
“The AI smart biosensor can also be used for precise diagnosis of other cancers using urine,” Professor Lee said.
ACS Nano published the results of the study.