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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Meningiomas are the most common type of primary brain tumor, accounting forapproximately 30% of all brain tumors. A substantial number of these tumors arenever surgically removed but rather monitored over time. Automatic and precisemeningioma segmentation is therefore beneficial to enable reliable growthestimation and patient-specific treatment planning. In this study, we proposethe inclusion of attention mechanisms over a U-Net architecture: (i)Attention-gated U-Net (AGUNet) and (ii) Dual Attention U-Net (DAUNet), using a3D MRI volume as input. Attention has the potential to leverage the globalcontext and identify features' relationships across the entire volume. To limitspatial resolution degradation and loss of detail inherent to encoder-decoderarchitectures, we studied the impact of multi-scale input and deep supervisioncomponents. The proposed architectures are trainable end-to-end and eachconcept can be seamlessly disabled for ablation studies. The validation studieswere performed using a 5-fold cross validation over 600 T1-weighted MRI volumesfrom St. Olavs University Hospital, Trondheim, Norway. For the best performingarchitecture, an average Dice score of 81.6% was reached for an F1-score of95.6%. With an almost perfect precision of 98%, meningiomas smaller than 3mlwere occasionally missed hence reaching an overall recall of 93%. Leveragingglobal context from a 3D MRI volume provided the best performances, even if thenative volume resolution could not be processed directly. Overall, near-perfectdetection was achieved for meningiomas larger than 3ml which is relevant forclinical use. In the future, the use of multi-scale designs and refinementnetworks should be further investigated to improve the performance. A largernumber of cases with meningiomas below 3ml might also be needed to improve theperformance for the smallest tumors.

Meningioma segmentation in T1-weighted MRI leveraging global context and attention mechanisms