An artificial intelligence (AI) solution developed by DeepHealth, a subsidiary of RadNet, could help nearly one in five patients avoid prostate biopsies while maintaining high sensitivity for detecting clinically significant prostate cancer, according to new research. The study found that the AI tool could reduce biopsies by approximately 19% while maintaining a sensitivity of 98.4% for identifying clinically significant prostate cancer. The findings highlight the potential of AI-assisted prostate MRI to help clinicians make more informed biopsy decisions and reduce unnecessary invasive procedures.
Prostate MRI Plays a Key Role in Early Cancer Detection
Early detection of clinically significant prostate cancer is essential to reducing mortality and avoiding unnecessary treatment. In recent years, multiparametric magnetic resonance imaging (mpMRI) has become a key part of the diagnostic pathway. This advanced imaging technique combines anatomical images with functional measurements to help doctors identify suspicious areas within the prostate.
Radiologists commonly use the Prostate Imaging Reporting and Data System (PI-RADS) to standardise the interpretation of prostate MRI scans and assess the likelihood of clinically significant cancer. The research, published in European Radiology, compared DeepHealth’s AI-based risk classification tool with radiologist-assigned PI-RADS scores. Researchers also assessed whether AI-assisted biopsy triage could reduce unnecessary procedures without compromising cancer detection.
Study Evaluates AI Across Multiple Countries and MRI Systems
The retrospective study analysed prostate MRI examinations conducted between 2014 and 2025 across the United States, Germany and the Netherlands. The researchers evaluated nearly 800 patients undergoing clinical assessment for suspected prostate cancer. Of these, 380 patients, representing approximately 48% of the study population, had clinically significant disease. The scans came from equipment manufactured by five different imaging vendors. According to the researchers, this supports the potential for the AI solution to perform across different MRI systems rather than relying on a single manufacturer’s equipment.
Three-Tier AI Classification Simplifies Biopsy Triage
DeepHealth’s Prostate MR Solution uses a simplified, three-tier risk classification system to help clinicians assess whether a patient may benefit from a biopsy. The system categorises findings as low, medium or high risk. In comparison, PI-RADS uses five categories, ranging from category 1, indicating that clinically significant cancer is highly unlikely, to category 5, indicating that it is highly likely.
In routine clinical practice, healthcare providers often group these five categories into three broader classifications:
Low likelihood: PI-RADS 1 and 2.
Equivocal: PI-RADS 3.
High likelihood: PI-RADS 4 and 5.
The AI tool aims to support clinical decision-making by providing an additional risk assessment, particularly when MRI findings are uncertain.
AI Maintains High Sensitivity While Reducing Biopsies
The researchers found that the AI solution achieved non-inferior patient-level sensitivity and higher patient-level discrimination than PI-RADS. However, it demonstrated lower lesion-level sensitivity at the predefined threshold, highlighting an important limitation in detecting individual cancer lesions. Overall, the findings suggest that AI-assisted biopsy triage could help clinicians avoid approximately 19% of biopsies while maintaining 98.4% sensitivity for clinically significant prostate cancer. The researchers believe the technology could be particularly useful for patients with equivocal PI-RADS 3 findings, where deciding whether to recommend a biopsy can be challenging. Nevertheless, radiologists would continue to be responsible for identifying and mapping suspicious lesions.
AI Could Help Reduce Unnecessary Invasive Procedures
Although prostate biopsies remain essential for confirming a cancer diagnosis, the procedures are invasive and can sometimes identify no cancer or only low-grade disease that may not require immediate treatment. Reducing unnecessary biopsies could therefore spare patients the associated risks and discomfort while helping clinicians focus diagnostic procedures on those most likely to benefit.
Commenting on the findings, Francesco Giganti, MD, senior author of the study and professor of radiology at University College London, said, “When an MRI finding is uncertain, deciding whether to recommend a biopsy can be difficult. Our findings suggest that AI could provide another piece of evidence to help clinicians make that decision. The opportunity is to use DeepHealth’s Prostate MR Solution to better identify which men may benefit from a biopsy, with the final medical decision remaining with the clinical team.”
DeepHealth Highlights AI’s Role in Clinical Decision-Making
Niccolò Stefani, MD, business and product leader for clinical AI at DeepHealth, said the findings demonstrated how AI could support clinicians at a critical decision point in prostate cancer care. “Our aim is to help care teams make more informed decisions about the next step for each patient,” he said. The study highlights the potential of AI-assisted prostate MRI to improve biopsy selection and support more personalised diagnostic pathways. However, the technology is intended to complement radiologists’ assessments rather than replace clinical judgement. As reported by radiologybusiness.com, by combining AI-based risk classification with expert radiological interpretation, healthcare teams may be able to reduce avoidable biopsies while maintaining a strong focus on detecting clinically significant prostate cancer.