English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Multiparametric magnetic resonance imaging (mpMRI) has demonstrated promisingresults in prostate cancer (PCa) detection using deep convolutional neuralnetworks (CNNs). Recently, transformers have achieved competitive performancecompared to CNNs in computer vision. Large-scale transformers need abundantannotated data for training, which are difficult to obtain in medical imaging.Self-supervised learning can effectively leverage unlabeled data to extractuseful semantic representations without annotation and its associated costs.This can improve model performance on downstream tasks with limited labelleddata and increase generalizability. We introduce a novel end-to-endCross-Shaped windows (CSwin) transformer UNet model, CSwin UNet, to detectclinically significant prostate cancer (csPCa) in prostate bi-parametric MRimaging (bpMRI) and demonstrate the effectiveness of our proposedself-supervised pre-training framework. Using a large prostate bpMRI datasetwith 1500 patients, we first pre-train CSwin transformer using multi-taskself-supervised learning to improve data-efficiency and networkgeneralizability. We then finetuned using lesion annotations to perform csPCadetection. Five-fold cross validation shows that self-supervised CSwin UNetachieves 0.888 AUC and 0.545 Average Precision (AP), significantlyoutperforming four state-of-the-art models (Swin UNETR, DynUNet, AttentionUNet, UNet). Using a separate bpMRI dataset with 158 patients, we evaluated ourmodel robustness to external hold-out data. Self-supervised CSwin UNet achieves0.79 AUC and 0.45 AP, still outperforming all other comparable methods anddemonstrating generalization to a dataset shift.

Cross-Shaped Windows Transformer with Self-supervised Pretraining for Clinically Significant Prostate Cancer Detection in Bi-parametric MRI