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

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The interpretation of medical images is a challenging task, often complicatedby the presence of artifacts, occlusions, limited contrast and more. Mostnotable is the case of chest radiography, where there is a high inter-ratervariability in the detection and classification of abnormalities. This islargely due to inconclusive evidence in the data or subjective definitions ofdisease appearance. An additional example is the classification of anatomicalviews based on 2D Ultrasound images. Often, the anatomical context captured ina frame is not sufficient to recognize the underlying anatomy. Current machinelearning solutions for these problems are typically limited to providingprobabilistic predictions, relying on the capacity of underlying models toadapt to limited information and the high degree of label noise. In practice,however, this leads to overconfident systems with poor generalization on unseendata. To account for this, we propose a system that learns not only theprobabilistic estimate for classification, but also an explicit uncertaintymeasure which captures the confidence of the system in the predicted output. Weargue that this approach is essential to account for the inherent ambiguitycharacteristic of medical images from different radiologic exams includingcomputed radiography, ultrasonography and magnetic resonance imaging. In ourexperiments we demonstrate that sample rejection based on the predicteduncertainty can significantly improve the ROC-AUC for various tasks, e.g., by8% to 0.91 with an expected rejection rate of under 25% for the classificationof different abnormalities in chest radiographs. In addition, we show thatusing uncertainty-driven bootstrapping to filter the training data, one canachieve a significant increase in robustness and accuracy.

Quantifying and Leveraging Predictive Uncertainty for Medical Image Assessment