Risk-based algorithm could improve cancer screenings
This time the intelligence is not artificial
An algorithm capable of estimating the risk that a particular patient will develop prostate cancer over the next five years should be used in a national screening program in the UK, one of the software's creators has said.
One in eight men in the United Kingdom will be diagnosed with prostate cancer in their lifetime, according to charity Prostate Cancer UK. Over 47,500 men are diagnosed with the disease, and more than 11,500 die from it, every year. Despite these grim statistics, prostate cancer is one of the most curable forms of cancer if it's caught early.
The most common form of screening is a regular blood test. Doctors look for prostate-specific antigen (PSA), a protein produced by cells in the prostate gland. Elevated levels could mean the cells are malignant. Patients in those cases are typically called in for a follow-up biopsy exam to confirm whether they have cancer or not.
PSA tests, however, aren't always reliable and tend to flag false positives, which results in unwanted and unnecessary biopsies.
Alternatively, one can screen for PSA as well as human kallikrein peptidase (hK2), another cancer biomarker. Though hK2 is a weak marker, it can help improve the accuracy of PSA testing. Scientists led by the University College London paired this duo-marker approach with an algorithm that factors in a patient's age to ultimately determine their risk of developing prostate cancer.
To come up with this software, the academics analyzed blood samples from more than 21,000 men, singling out data from 571 men who died of prostate cancer, and 2,169 men who did not contract the disease. The team then took the PSA and hK2 levels from those samples, and the ages of the men, and used this information to devise a bog-standard algorithm capable of automatically estimating a man's risk of developing prostate cancer from their age and PSA and hK2 levels.
The team said their screening approach could achieve, for regularly tested men aged 55 and over, a 90 percent detection rate with a false positive rate of 1.2 percent.
The PSA test alone had a detection rate of 86 percent, with a false positive rate of two percent. Scaling the algorithm approach down to 86 per cent detection rate yielded a 0.5 false positive rate, according to the scientists, allowing them to claim their software-based double-marker technique can reduce false positives by three quarters while maintaining the same level of detection as today's PSA testing.
These results were published in the Journal of Medical Screening.
- US drug watchdog green-lights first prostate-cancer-predicting AI software
- AI-enhanced frog stem cells start to replicate in entirely new ways
- Prisons transcribe private phone calls with inmates using speech-to-text AI
- Study: AI designed to detect diabetic eye disease blinks in the real world, makes more work for doctors
"Our study shows a different screening approach could reduce the number of false positives by three quarters," said Sir Nicholas Wald, lead author of the paper and a professor at UCL's Institute of Health Informatics. "This would make screening for prostate cancer safer and more accurate, reducing overdiagnosis and overtreatment."
Sir Nicholas wants to test the algorithm on more patients to see whether it can be used in real clinical settings.
"The next step is to test the feasibility of this approach in practice with a pilot project inviting healthy men for screening. If the project is successful, we believe this approach ought to be considered as part of a national screening program for all men," he concluded. ®