Anatomically guided self-adapting deep neural network for clinically significant prostate cancer detection on bi-parametric MRI: a multi-center study

Abstract

Objective: To evaluate the effectiveness of a self-adapting deep network, trained on large-scale bi-parametric MRI data, in detecting clinically significant prostate cancer (csPCa) in external multi-center data from men of diverse demographics; to investigate the advantages of transfer learning.

Methods: We used two samples: (i) Publicly available multi-center and multi-vendor Prostate Imaging: Cancer AI (PI-CAI) training data, consisting of 1500 bi-parametric MRI scans, along with its unseen validation and testing samples; (ii) In-house multi-center testing and transfer learning data, comprising 1036 and 200 bi-parametric MRI scans. We trained a self-adapting 3D nnU-Net model using probabilistic prostate masks on the PI-CAI data and evaluated its performance on the hidden validation and testing samples and the in-house data with and without transfer learning. We used the area under the receiver operating characteristic (AUROC) curve to evaluate patient-level performance in detecting csPCa.

Results: The PI-CAI training data had 425 scans with csPCa, while the in-house testing and fine-tuning data had 288 and 50 scans with csPCa, respectively. The nnU-Net model achieved an AUROC of 0.888 and 0.889 on the hidden validation and testing data. The model performed with an AUROC of 0.886 on the in-house testing data, with a slight decrease in performance to 0.870 using transfer learning.

Conclusions: The state-of-the-art deep learning method using prostate masks trained on large-scale bi-parametric MRI data provides high performance in detecting csPCa in internal and external testing data with different characteristics, demonstrating the robustness and generalizability of deep learning within and across datasets.

Clinical relevance statement: A self-adapting deep network, utilizing prostate masks and trained on large-scale bi-parametric MRI data, is effective in accurately detecting clinically significant prostate cancer across diverse datasets, highlighting the potential of deep learning methods for improving prostate cancer detection in clinical practice.

Keywords: Deep learning; Magnetic resonance imaging; Prostate cancer.

Fig. 1

The datasets used in the study. The ProstateX data and Prostate Imaging: Cancer AI [PI-CAI] training data were used for the model training in the study. The PI-CAI data was used to train an ensemble of 3D nnU-Net models to detect clinically significant PCa

Fig. 2

Creating the probabilistic zone masks. The ProstateX data was used to train a 3D nnU-Net for creating probabilistic prostate zone masks on the PI-CAI training data. Afterward, the probabilistic masks were used to augment the ensemble model's clinically significant prostate cancer detection performance

Fig. 3

The 3D nnU-Net model for detecting clinically significant prostate cancer. a The 3D nnU-Net was fed with T2W imaging, diffusion-weighted imaging, and apparent diffusion coefficient maps along with probabilistic prostate masks via five different channels. The model was trained on the publicly available Prostate Imaging: Cancer AI training data using the significant cancer masks provided by the organizers as the ground truth. b The 3D nnU-Net model was trained using a fivefold cross-validation approach. Then, the ensemble of five nnU-Net models was used to make the final predictions

Fig. 4

The AUROC, FROC, and PR curves of the nnU-Net in detecting clinically significant prostate cancer with and without transfer learning. The area under the receiver operating characteristic (AUROC), Free-Response Receiver Operating Characteristic (FROC), and Precision–Recall (PR) curves of the ensemble of five nnU-Net models in detecting clinically significant prostate cancer in the in-house dataset with and without transfer learning. The AUROC and FROC slightly decreased, and average precision slightly increased using transfer learning, not reaching a statistical significance

Fig. 5

A patient with clinically significant prostate cancer in the right peripheral zone from the in-house data. The T2W (a), diffusion-weighted image with a b-value of 1400 s/mm² (b), apparent diffusion coefficient map (c), and the predictions of the deep learning model overlaid on the T2W image (d). The model correctly predicted the lesion and drew its borders