Impact of uncertainty quantification through conformal prediction on volume assessment from DL-based MRI prostate segmentation

This study aimed to evaluate the uncertainty of a deep learning (DL)-based prostate segmentation algorithm using conformal prediction (CP) and its impact on prostate volume (PV) calculation in patients at risk of prostate cancer (PC). The study involved 377 patients’ 3-Tesla T2-weighted scans. By applying CP at an 85% confidence level, unreliable pixel segmentations of the DL model were flagged, and PV was recalculated. The results showed that CP improved PV accuracy, reducing the relative volume difference (RVD) compared to the DL algorithm alone. The study concluded that uncertainty quantification via CP increases the accuracy and reliability of DL-based PV assessments for PC risk patients.

Key points:

• Conformal prediction can flag uncertain pixel predictions of prostate segmentations at a user-defined confidence level.
• Deep learning with conformal prediction shows high accuracy in prostate volumetric assessment.
• Agreement between automatic and ellipsoid-derived volume was significantly larger with conformal prediction.

Article: Impact of uncertainty quantification through conformal prediction on volume assessment from deep learning-based MRI prostate segmentation

Authors: Marius Gade, Kevin Mekhaphan Nguyen, Sol Gedde & Alvaro Fernandez-Quilez