RESUMEN
Although fully automated volumetric approaches for monitoring brain tumor response have many advantages, most available deep learning models are optimized for highly curated, multi-contrast MRI from newly diagnosed gliomas, which are not representative of post-treatment cases in the clinic. Improving segmentation for treated patients is critical to accurately tracking changes in response to therapy. We investigated mixing data from newly diagnosed (n = 208) and treated (n = 221) gliomas in training, applying transfer learning (TL) from pre- to post-treatment imaging domains, and incorporating spatial regularization for T2-lesion segmentation using only T2 FLAIR images as input to improve generalization post-treatment. These approaches were evaluated on 24 patients suspected of progression who had received prior treatment. Including 26% of treated patients in training improved performance by 13.9%, and including more treated and untreated patients resulted in minimal changes. Fine-tuning with treated glioma improved sensitivity compared to data mixing by 2.5% (p < 0.05), and spatial regularization further improved performance when used with TL by 95th HD, Dice, and sensitivity (6.8%, 0.8%, 2.2%; p < 0.05). While training with ≥60 treated patients yielded the majority of performance gain, TL and spatial regularization further improved T2-lesion segmentation to treated gliomas using a single MR contrast and minimal processing, demonstrating clinical utility in response assessment.
RESUMEN
RATIONALE AND OBJECTIVES: Early prostate cancer detection and staging from MRI is extremely challenging for both radiologists and deep learning algorithms, but the potential to learn from large and diverse datasets remains a promising avenue to increase their performance within and across institutions. To enable this for prototype-stage algorithms, where the majority of existing research remains, we introduce a flexible federated learning framework for cross-site training, validation, and evaluation of custom deep learning prostate cancer detection algorithms. MATERIALS AND METHODS: We introduce an abstraction of prostate cancer groundtruth that represents diverse annotation and histopathology data. We maximize use of this groundtruth if and when they are available using UCNet, a custom 3D UNet that enables simultaneous supervision of pixel-wise, region-wise, and gland-wise classification. We leverage these modules to perform cross-site federated training using 1400+ heterogeneous multi-parameteric prostate MRI exams from two University hospitals. RESULTS: We observe a positive result, with significant improvements in cross-site generalization performance with negligible intra-site performance degradation for both lesion segmentation and per-lesion binary classification of clinically-significant prostate cancer. Cross-site lesion segmentation performance intersection-over-union (IoU) improved by 100%, while cross-site lesion classification performance overall accuracy improved by 9.5-14.8%, depending on the optimal checkpoint selected by each site. CONCLUSION: Federated learning can improve the generalization performance of prostate cancer detection models across institutions while protecting patient health information and institution-specific code and data. However, even more data and participating institutions are likely required to improve the absolute performance of prostate cancer classification models. To enable adoption of federated learning with limited re-engineering of federated components, we open-source our FLtools system at https://federated.ucsf.edu, including examples that can be easily adapted to other medical imaging deep learning projects.