Enhancing Clinically Significant Prostate Cancer Prediction in T2-weighted Images through Transfer Learning from Breast Cancer
Chi-en Amy Tai, Alexander Wong
TL;DR
The paper addresses the challenge of predicting clinically significant prostate cancer from T2-weighted MRI under limited data. It applies cross-domain transfer learning by pretraining a volumetric deep radiomic feature extractor on breast cancer data (BCa weights) and comparing to MONAI pretraining. The breast-to-prostate transfer yields a large LOOCV accuracy gain, reaching $97.50\%$ versus $63.00\%$ with MONAI and improving sensitivity, specificity, and F1 to above $90\%$, demonstrating cross-domain knowledge transfer as a promising approach to mitigate data limitations in medical imaging. The work suggests extending the method to other cancer types to enhance generalization when data are scarce.
Abstract
In 2020, prostate cancer saw a staggering 1.4 million new cases, resulting in over 375,000 deaths. The accurate identification of clinically significant prostate cancer is crucial for delivering effective treatment to patients. Consequently, there has been a surge in research exploring the application of deep neural networks to predict clinical significance based on magnetic resonance images. However, these networks demand extensive datasets to attain optimal performance. Recently, transfer learning emerged as a technique that leverages acquired features from a domain with richer data to enhance the performance of a domain with limited data. In this paper, we investigate the improvement of clinically significant prostate cancer prediction in T2-weighted images through transfer learning from breast cancer. The results demonstrate a remarkable improvement of over 30% in leave-one-out cross-validation accuracy.