The Learning Curve in Prostate MRI Interpretation: Self-Directed Learning Versus Continual Reader Feedback

Abstract

The learning curve in prostate tumor detection largely reflected self-directed learning. Continual feedback had a lesser effect. Clinical prostate MRI interpretation by novice radiologists warrants caution.

Prostate Magnetic Resonance Imaging Interpretation Varies Substantially Across Radiologists

“…In addition, sensitivity analyses did not detect differences in cancer yield performance over time, either collectively or for individual radiologists. This suggests that quality improvement initiatives are necessary to improve the real-world effectiveness of the PIRADS scoring system, as increased experience over time may be insufficient [13].…”

National implementation of multi‐parametric magnetic resonance imaging for prostate cancer detection – recommendations from a UK consensus meeting

“…Third, in order to maintain quality mpMRI reporting and guarantee accurate and safe prostate mpMRI reports, minimum standards for reporting radiologists were tackled. Although the effect of dedicated training on the accuracy of prostate cancer localisation on mpMRI, the effect of continual feedback on reporting confidence and a ‘learning curve’ effect have been documented, the establishment of a threshold number of prostate mpMRI required during training, to reach independent reporting and to maintain expertise are lacking . While some may not agree with the concept of quantitative metrics to gauge quality or experience, the majority of the panel agreed that an independent radiologist should report >100 prostate mpMRI scans per year with regular attendance to MDT meetings of at least twice a month.…”

Proposed Adjustments to PI-RADS Version 2 Decision Rules: Impact on Prostate Cancer Detection

“…The indications for MR imaging in these patients were suspicion of prostate cancer without prior prostate biopsy (n = 161), prior negative biopsy findings (n = 107), and prior positive biopsy findings (n = 75). Between two and 195 of the patients were included in earlier unrelated studies from our institution that evaluated the optimization of the acquisition and interpretation of prostate MR imaging (10)(11)(12)(13), results from MR imaging-US fusion-targeted prostate biopsy (14)(15)(16), or the interobserver reproducibility of PI-RADS V2 (4,17,18); none of these explored the impact of proposed adjustments to the PI-RADS V2 decision rules, as is the subject of the present investigations. Given the possibility of misregistration error at the time of fusion targeting (8,9), standardized options within the database were selected at the time of the initial search to filter patients from the search results on the basis of the results of concurrent systematic biopsy; the relative position of the fusion and systematic cores was not explicitly captured in the database as a searchable field and thus not taken into consideration in this process.…”