Integrative feature-enhanced network model predicting prostate biopsy results in patients with negative MRI

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Abstract Background According to the Prostate Imaging Reporting and Data System, version 2.1 (PI-RADS v2.1), lesions with an intermediate or high level of suspicion (PI-RADS $$\ge$$ 3) typically undergo MRI-targeted biopsy, with or without systematic biopsies. However, among patients with negative multi-parametric MRI (mpMRI) (PI-RADS < 3), there exists a current lack of consensus regarding the circumstances under which systematic biopsies should be performed, which leads to unnecessary biopsies and patient morbidity. To discern patients with negative prostate MRI who could potentially forgo unnecessary biopsies, we employed an integrative feature-enhanced deep learning approach that leverages both imaging and clinical information to predict biopsy results. Methods An Integrative Feature-enhanced Network (IFN) was proposed to predict the prostate biopsy results, confirmed by the histopathologic examination. The IFN was built based on a 3D ResNet with extra proposed feature-enhanced (FE) blocks, and a fully-connected layer combining the imaging and clinical information. The study cohort consisted of 508 patients with negative prostate 3 T mpMRI between 2016 and 2020. The proposed IFN was trained and validated through fivefold cross-validation with bootstrapping, and the model’s performance was measured by area under the curve (AUC), sensitivity, specificity, and negative predictive value (NPV). AUCs were compared via the DeLong test with a 95% confidence interval (CI), and the rest of the measurements were compared via the Chi-squared test. Results Overall, of 508 men included in the study cohort, 48 (9.4%) harbored positive biopsies and 460 (90.6%) harbored negative biopsies. The proposed IFN achieved an AUC of 0.753, a NPV of 0.967, a sensitivity of 0.787, and a specificity of 0.642. In comparative evaluations, the IFN model outperformed implementations utilizing only clinical data (AUC: 0.721, NPV: 0.942) or image features alone (AUC: 0.652, NPV: 0.948). Compared with pre-existing risk metrics, the proposed network achieved a higher NPV than conventional PSAD-based thresholds of 0.10 ng/ml/ml (NPV: 0.923) and 0.15 ng/ml/ml (NPV: 0.922), an existing radiomics-based approach (NPV: 0.953), and established clinical risk calculators, including SWOP #3&4 (NPV: 0.911) and PCPTRC (NPV: 0.914). Conclusion The proposed deep-learning-based IFN may be potentially feasible to predict the prostate biopsy results for patients with negative mpMRI. With the improved predictability, the IFN may be able to help stratify which patients might avoid biopsies and thus potentially reduce the number of unnecessary biopsies.