Can MRI-Based AI Bolster Biopsy Decision-Making in PI-RADS 3 Cases?

New research suggests that artificial intelligence (AI) may have a significant impact in reducing prostate biopsies for PI-RADS 3 lesions on magnetic resonance imaging (MRI).

For the retrospective study, recently presented at the Radiological Society of North America (RSNA) conference, researchers evaluated an AI model for 302 PI-RADS lesions in a total of 248 patients. All patients in the cohort had MRI/ultrasound-guided fusion biopsy preceded by multiparametric MRI, according to the study. The study authors noted that 44 of the 302 biopsies were positive for clinically significant prostate cancer (csPCa).¹

The researchers noted a prostate-specific antigen density (PSAD) > .15 ng/mL² was associated with a greater than sixfold likelihood of csPCa in PI-RAD 3 index lesions but only a 61 percent sensitivity. However, when the AL model was combined with the PSAD threshold, the study authors noted a 17 percent increase in sensitivity (78 percent) and a 93 percent negative predictive value (NPV).¹

“Combining a bpMRI-based AI model with PSAD can notably enhance the diagnostic process for patients with PI-RADS 3 lesions,” wrote lead study author Omer Esengur, M.D., a postdoctoral researcher associated with the Molecular Imaging Branch of the National Institutes of Health (NIH).

(Editor’s note: For additional coverage of RSNA, click here.)

While the AI model itself had an overall sensitivity of 57 percent and specificity of 62 percent for diagnosing csPCa, the researchers noted it still offered promise in ruling out csPCa with an 89 percent NPV.¹

“The AI model alone demonstrated a high NPV, particularly in detecting csPCa, which is crucial for minimizing unnecessary biopsies,” added Esengur and colleagues.

Reference

1. Esengur OT, Yilmaz EC, Ozyoruk KB, et al. Multimodal approach to optimize biopsy-decision-making for PI-RADS 3 lesions at multiparametric MRI. Poster presented at the Radiological Society of North America (RSNA) 2024 110^th Scientific Assembly and Annual Meeting Dec. 1-5, 2024. Available at: https://www.rsna.org/annual-meeting .