RESEARCHERS in Sweden are exploring the role of artificial intelligence (AI) in diagnosing prostate cancer. At present, diagnosing prostate cancer can be difficult as there is no single, definitive test used. Patients can be offered an examination of the prostate, a urine sample test to find infections, and a prostate-specific antigen test (PSA) to screen for prostate cancer. Biopsies are also sometimes necessary.
Investigators from the Karolinska Institutet, Stockholm, Sweden, examined >95,000 images taken from prostate cancer biopsies across Europe, and included many rare disease subtypes from Australian records, to establish whether AI is able to effectively diagnose prostate cancer.
Working with colleagues from the Radbound University Medical Center, Nijmegen, the Netherlands; the University of Turku, Finland; and Google Health, Mountain View, California, USA, researchers ran a competition for software developers to create algorithms that could grade prostate cancer tumours, training them by using 10,000 biopsy images from patients globally.
Researchers concluded that whilst AI illustrated the ability to diagnose prostate cancer early, algorithms differ due to sample preparation, differences in patient populations, and scan images. However, the top-performing algorithms in AI diagnosis outperformed general pathologists and matched the general performance of uropathologists.
Nita Mulliqi, a PhD student at the Karolinska Institutet, identified four important areas in improving the process of diagnosing and grading prostate cancer with the use of AI: consistent scanner calibration across all practices; improved algorithms that can be widely used; including larger datasets from hospitals across the world to widen the knowledge of the algorithms; and examining different subtypes of the disease. Mulliqi stressed: “we need an international effort to collect datasets that are representative of the variation in technical approaches and between patients. The combination of our vast database and our colleagues’ algorithms is beginning to show how we can really work together to make a big difference for clinicians and patients.”
As there is a global shortage of pathologists, it is sensible to explore other diagnostic technology options. AI could be a promising step forward into the future of prostate cancer screening and diagnosis.
