Here's a medical paradox that keeps urologists up at night: by age 60, roughly half of all men have prostate cancer. By 80, it's 70 percent. Yet most of these men will die with prostate cancer, not from it.
The challenge isn't finding prostate cancer—it's figuring out which cases actually need treatment. Enter PROVIZ, an artificial intelligence system developed at the Norwegian University of Science and Technology (NTNU) that might finally thread this needle.
The clinical dilemma is real. As PSA blood testing has become more widespread, prostate cancer detection has surged—Norway alone now diagnoses about 5,000 new cases annually. Each suspected case requires an MRI scan, and each scan needs an experienced radiologist to interpret it and decide: biopsy or watch?
That's where the healthcare system starts to buckle. "The surge creates significant diagnostic bottlenecks," explains Professor Tone Frost Bathen, who manages the project at NTNU. Radiologists are drowning in scans, and the wait times keep growing.
PROVIZ analyzes MRI images of the prostate gland, helping radiologists work more efficiently and determine whether patients need a biopsy—and if so, where exactly to take it. The tool doesn't replace human judgment; it augments it.
What's particularly interesting about this project is what happened when researchers interviewed patients about AI in cancer diagnostics. The finding: trust in doctors remains crucial. Patients were comfortable with AI assistance, but only if their doctor understood how the AI reached its conclusions and could explain the results.
Dr. Simon A. Berger, a PhD research fellow on the team, emphasizes this point. The doctor must serve as "communicator and guarantor of safety." This isn't about replacing clinical expertise—it's about giving experts better tools.
Now, let's be clear about what this solves and what it doesn't. PROVIZ addresses the diagnostic bottleneck—who needs a biopsy. It doesn't yet tackle the deeper question of which cancers, once found, actually need aggressive treatment versus active surveillance. That's a separate research frontier involving genomic markers and long-term outcome data.
But solving the first problem matters. Faster, more accurate initial screening means patients get appropriate care sooner, and it frees up specialist time for the complex cases that truly need human expertise.


