Prostate MRI AI Study Highlights the Trade-Off Between Biopsy Avoidance and Missed Cancers
Published in Artificial Intelligence, Journals, Scanners and Imaging, Tools.
A study published in BMC Medical Imaging on September 28, 2026 evaluated QP-Prostate, an AI tool for classifying suspicious prostate MRI lesions already identified by radiologists. The retrospective, single-centre analysis included 113 men who underwent MRI-targeted biopsy.
For clinically significant cancer, treating moderate and high AI scores as positive produced 92% sensitivity and 32% specificity. Using only high scores reduced sensitivity to 79% and increased specificity to 43%, illustrating how the threshold changes the balance between missed cancers and false positives.
A theoretical biopsy-sparing analysis estimated that the broader threshold could withhold 20 of 113 biopsies while missing 5 of 66 clinically significant cancers. The stricter threshold could withhold 34 biopsies but miss 14 such cancers. These were simulated decisions, not demonstrated improvements in patient care.
The authors call for prospective validation before AI-guided biopsy avoidance. Several coauthors work for the developer, Quibim, including its CEO, who holds company stock.
Source: Vilanova and colleagues, BMC Medical Imaging (September 28, 2026).