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

T2-weighted magnetic resonance imaging (MRI) and diffusion-weighted imaging(DWI) are essential components for cervical cancer diagnosis. However,combining these channels for training deep learning models are challenging dueto misalignment of images. Here, we propose a novel multi-head framework thatuses dilated convolutions and shared residual connections for separate encodingof multiparametric MRI images. We employ a residual U-Net model as a baseline,and perform a series of architectural experiments to evaluate the tumorsegmentation performance based on multiparametric input channels and featureencoding configurations. All experiments were performed using a cohortincluding 207 patients with locally advanced cervical cancer. Our proposedmulti-head model using separate dilated encoding for T2W MRI, and combinedb1000 DWI and apparent diffusion coefficient (ADC) images achieved the bestmedian Dice coefficient similarity (DSC) score, 0.823 (95% confidence interval(CI), 0.595-0.797), outperforming the conventional multi-channel model, DSC0.788 (95% CI, 0.568-0.776), although the difference was not statisticallysignificant (p>0.05). We investigated channel sensitivity using 3D GRAD-CAM andchannel dropout, and highlighted the critical importance of T2W and ADCchannels for accurate tumor segmentations. However, our results showed thatb1000 DWI had a minor impact on overall segmentation performance. Wedemonstrated that the use of separate dilated feature extractors andindependent contextual learning improved the model's ability to reduce theboundary effects and distortion of DWI, leading to improved segmentationperformance. Our findings can have significant implications for the developmentof robust and generalizable models that can extend to other multi-modalsegmentation applications.

Deep Learning Framework with Multi-Head Dilated Encoders for Enhanced Segmentation of Cervical Cancer on Multiparametric Magnetic Resonance Imaging