AI tool outperforms MR in predicting prostate cancer spread, study finds
May 02, 2025
by Gus Iversen, Editor in Chief
A study presented at the American Urological Association’s 2025 Annual Meeting suggests that Avenda Health’s cancer mapping tool, Unfold AI, may offer significantly greater accuracy than MR alone in predicting seminal vesicle invasion (SVI) in prostate cancer patients.
Researchers from Stanford University School of Medicine and UCLA’s David Geffen School of Medicine compared Unfold AI’s performance with standard MR interpretations in two cohorts of men undergoing prostate cancer surgery. The AI tool, which integrates MR and clinical data to generate a 3D cancer map, achieved a 92% accuracy rate in detecting SVI in the first cohort, compared to 52% accuracy by radiologists interpreting MR alone. In the second, smaller cohort, Unfold AI missed two of 10 confirmed SVI cases, while MR missed four.
The study, titled “Prediction of Seminal Vesicle Invasion Using Artificial Intelligence Prostate Cancer Risk Mapping,” also found that Unfold AI produced fewer false positives. The tool reported an area under the curve (AUC) of 0.95 versus 0.80 for MR in the first group (p = 0.003), and 0.85 versus 0.74 in the second group.
Determining SVI is a key factor in staging prostate cancer and guiding treatment decisions, but MR has been criticized for inconsistent accuracy in this context.
“These results demonstrate how Unfold AI continues to improve the diagnosis and staging of prostate cancer, enabling the physician to recommend and deliver the best therapy to the patient,” said Shyam Natarajan, Ph.D., co-founder and CEO of Avenda Health.
The Culver City, California-based company previously won the BJUI Compass Prize 2025 for earlier work on Unfold AI’s ability to predict extracapsular extension risk. The current project received partial funding from the National Cancer Institute.
Researchers from Stanford University School of Medicine and UCLA’s David Geffen School of Medicine compared Unfold AI’s performance with standard MR interpretations in two cohorts of men undergoing prostate cancer surgery. The AI tool, which integrates MR and clinical data to generate a 3D cancer map, achieved a 92% accuracy rate in detecting SVI in the first cohort, compared to 52% accuracy by radiologists interpreting MR alone. In the second, smaller cohort, Unfold AI missed two of 10 confirmed SVI cases, while MR missed four.
The study, titled “Prediction of Seminal Vesicle Invasion Using Artificial Intelligence Prostate Cancer Risk Mapping,” also found that Unfold AI produced fewer false positives. The tool reported an area under the curve (AUC) of 0.95 versus 0.80 for MR in the first group (p = 0.003), and 0.85 versus 0.74 in the second group.
Determining SVI is a key factor in staging prostate cancer and guiding treatment decisions, but MR has been criticized for inconsistent accuracy in this context.
“These results demonstrate how Unfold AI continues to improve the diagnosis and staging of prostate cancer, enabling the physician to recommend and deliver the best therapy to the patient,” said Shyam Natarajan, Ph.D., co-founder and CEO of Avenda Health.
The Culver City, California-based company previously won the BJUI Compass Prize 2025 for earlier work on Unfold AI’s ability to predict extracapsular extension risk. The current project received partial funding from the National Cancer Institute.