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

Global: Zendy Plus monthly

As machine learning becomes increasingly prevalent in critical fields such ashealthcare, ensuring the safety and reliability of machine learning systemsbecomes paramount. A key component of reliability is the ability to estimateuncertainty, which enables the identification of areas of high and lowconfidence and helps to minimize the risk of error. In this study, we propose amachine learning pipeline called U-PASS tailored for clinical applications thatincorporates uncertainty estimation at every stage of the process, includingdata acquisition, training, and model deployment. The training process isdivided into a supervised pre-training step and a semi-supervised finetuningstep. We apply our uncertainty-guided deep learning pipeline to the challengingproblem of sleep staging and demonstrate that it systematically improvesperformance at every stage. By optimizing the training dataset, activelyseeking informative samples, and deferring the most uncertain samples to anexpert, we achieve an expert-level accuracy of 85% on a challenging clinicaldataset of elderly sleep apnea patients, representing a significant improvementover the baseline accuracy of 75%. U-PASS represents a promising approach toincorporating uncertainty estimation into machine learning pipelines, therebyimproving their reliability and unlocking their potential in clinical settings.

U-PASS: an Uncertainty-guided deep learning Pipeline for Automated Sleep Staging