RadHop-Net: A Lightweight Radiomics-to-Error Regression for False Positive Reduction In MRI Prostate Cancer Detection

TL;DR

The paper tackles the high false-positive rate in bpMRI-based clinically significant prostate cancer detection. It proposes RadHop features for Stage 1 and RadHop-Net, a lightweight CNN trained in a radiomics-to-error fashion to predict a residue $\varepsilon$ that corrects Stage 1 probabilities. A novel wrMSE loss with $\gamma=0.95$ balances the learning between false positives and true positives. On the pi-cai dataset, the two-stage pipeline improves lesion-level AP from $0.407$ to $0.468$ and achieves AUROC $0.839$, with approximately $54{,}585$ parameters, illustrating a practical, low-complexity approach for FP reduction in MRI prostate cancer detection.

Abstract

Clinically significant prostate cancer (csPCa) is a leading cause of cancer death in men, yet it has a high survival rate if diagnosed early. Bi-parametric MRI (bpMRI) reading has become a prominent screening test for csPCa. However, this process has a high false positive (FP) rate, incurring higher diagnostic costs and patient discomfort. This paper introduces RadHop-Net, a novel and lightweight CNN for FP reduction. The pipeline consists of two stages: Stage 1 employs data driven radiomics to extract candidate ROIs. In contrast, Stage 2 expands the receptive field about each ROI using RadHop-Net to compensate for the predicted error from Stage 1. Moreover, a novel loss function for regression problems is introduced to balance the influence between FPs and true positives (TPs). RadHop-Net is trained in a radiomics-to-error manner, thus decoupling from the common voxel-to-label approach. The proposed Stage 2 improves the average precision (AP) in lesion detection from 0.407 to 0.468 in the publicly available pi-cai dataset, also maintaining a significantly smaller model size than the state-of-the-art.

Figures (4)

• Figure 1: The proposed pipeline has two stages of processing. Stage 1 extracts RadHop features from each sequence to predict a heatmap and extract candidate ROIs. In Stage 2, RadHop-Net expands the receptive field about each ROI and compensates for the probability error from Stage 1 predictions.
• Figure 2: The architecture of the proposed RadHop-Net for Stage 1 probability error prediction.
• Figure 3: Quantitative comparison of the detection performance before and after adding the RadHop-Net.
• Figure 4: Three examples from the pi-cai dataset, demonstrating the effectiveness of the RadHop-Net. False positives are shown in red, and true positives in green.