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

Purpose: Manual annotations for training deep learning (DL) models inauto-segmentation are time-intensive. This study introduces a hybridrepresentation-enhanced sampling strategy that integrates both density anddiversity criteria within an uncertainty-based Bayesian active learning (BAL)framework to reduce annotation efforts by selecting the most informativetraining samples. Methods: The experiments are performed on two lower extremity(LE) datasets of MRI and CT images, focusing on the segmentation of the femur,pelvis, sacrum, quadriceps femoris, hamstrings, adductors, sartorius, andiliopsoas, utilizing a U-net-based BAL framework. Our method selects uncertainsamples with high density and diversity for manual revision, optimizing formaximal similarity to unlabeled instances and minimal similarity to existingtraining data. We assess the accuracy and efficiency using Dice and a proposedmetric called reduced annotation cost (RAC), respectively. We further evaluatethe impact of various acquisition rules on BAL performance and design anablation study for effectiveness estimation. Results: In MRI and CT datasets,our method was superior or comparable to existing ones, achieving a 0.8\% Diceand 1.0\% RAC increase in CT (statistically significant), and a 0.8\% Dice and1.1\% RAC increase in MRI (not statistically significant) in volume-wiseacquisition. Our ablation study indicates that combining density and diversitycriteria enhances the efficiency of BAL in musculoskeletal segmentationcompared to using either criterion alone. Conclusion: Our sampling method isproven efficient in reducing annotation costs in image segmentation tasks. Thecombination of the proposed method and our BAL framework provides asemi-automatic way for efficient annotation of medical image datasets.

Hybrid Representation-Enhanced Sampling for Bayesian Active Learning in Musculoskeletal Segmentation of Lower Extremities