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

Purpose Automated segmentation of anatomical structures in medical imageanalysis is a prerequisite for autonomous diagnosis as well as various computerand robot aided interventions. Recent methods based on deep convolutionalneural networks (CNN) have outperformed former heuristic methods. However,those methods were primarily evaluated on rigid, real-world environments. Inthis study, existing segmentation methods were evaluated for their use on a newdataset of transoral endoscopic exploration. Methods Four machine learningbased methods SegNet, UNet, ENet and ErfNet were trained with supervision on anovel 7-class dataset of the human larynx. The dataset contains 536 manuallysegmented images from two patients during laser incisions. TheIntersection-over-Union (IoU) evaluation metric was used to measure theaccuracy of each method. Data augmentation and network ensembling were employedto increase segmentation accuracy. Stochastic inference was used to showuncertainties of the individual models. Patient-to-patient transfer wasinvestigated using patient-specific fine-tuning. Results In this study, aweighted average ensemble network of UNet and ErfNet was best suited for thesegmentation of laryngeal soft tissue with a mean IoU of 84.7 %. The highestefficiency was achieved by ENet with a mean inference time of 9.22 ms perimage. It is shown that 10 additional images from a new patient are sufficientfor patient-specific fine-tuning. Conclusion CNN-based methods for semanticsegmentation are applicable to endoscopic images of laryngeal soft tissue. Thesegmentation can be used for active constraints or to monitor morphologicalchanges and autonomously detect pathologies. Further improvements could beachieved by using a larger dataset or training the models in a self-supervisedmanner on additional unlabeled data.

A Dataset of Laryngeal Endoscopic Images with Comparative Study on Convolution Neural Network Based Semantic Segmentation