Quantitative imaging metrics forming multiparametric prostate MRI have been shown to correlate with Gleason grade. We therefore aimed to develop logistic regression models which predict aggressive prostate cancer in focal lesions using quantitative MRI parameters. Models were constructed separately for the transition and peripheral zones, using data from 176 examinations.
In the peripheral zone, a combination of 3 simple parameters were found to predict a Gleason 4/5 component with a similar sensitivity and specificity to experienced radiologists. However, performance was relatively poor in the transition zone. Logistic regression models may therefore prove useful when training radiologists to characterise prostate cancer.