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/

Zendy Plus

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

Zendy Tools

Zendy Open

There has recently been great progress in automatic segmentation of medicalimages with deep learning algorithms. In most works observer variation isacknowledged to be a problem as it makes training data heterogeneous but so farno attempts have been made to explicitly capture this variation. Here, wepropose an approach capable of mimicking different styles of segmentation,which potentially can improve quality and clinical acceptance of automaticsegmentation methods. In this work, instead of training one neural network onall available data, we train several neural networks on subgroups of databelonging to different segmentation variations separately. Because a priori itmay be unclear what styles of segmentation exist in the data and becausedifferent styles do not necessarily map one-on-one to different observers, thesubgroups should be automatically determined. We achieve this by searching forthe best data partition with a genetic algorithm. Therefore, each network canlearn a specific style of segmentation from grouped training data. We provideproof of principle results for open-sourced prostate segmentation MRI data withsimulated observer variations. Our approach provides an improvement of up to23% (depending on simulated variations) in terms of Dice and surface Dicecoefficients compared to one network trained on all data.

Observer variation-aware medical image segmentation by combining deep learning and surrogate-assisted genetic algorithms