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

Although deep learning prediction models have been successful in thediscrimination of different classes, they can often suffer from poorcalibration across challenging domains including healthcare. Moreover, thelong-tail distribution poses great challenges in deep learning classificationproblems including clinical disease prediction. There are approaches proposedrecently to calibrate deep prediction in computer vision, but there are nostudies found to demonstrate how the representative models work in differentchallenging contexts. In this paper, we bridge the confidence calibration fromcomputer vision to medical imaging with a comparative study of four high-impactcalibration models. Our studies are conducted in different contexts (naturalimage classification and lung cancer risk estimation) including in balanced vs.imbalanced training sets and in computer vision vs. medical imaging. Ourresults support key findings: (1) We achieve new conclusions which are notstudied under different learning contexts, e.g., combining two calibrationmodels that both mitigate the overconfident prediction can lead tounder-confident prediction, and simpler calibration models from the computervision domain tend to be more generalizable to medical imaging. (2) Wehighlight the gap between general computer vision tasks and medical imagingprediction, e.g., calibration methods ideal for general computer vision tasksmay in fact damage the calibration of medical imaging prediction. (3) We alsoreinforce previous conclusions in natural image classification settings. Webelieve that this study has merits to guide readers to choose calibrationmodels and understand gaps between general computer vision and medical imagingdomains.

A Comparative Study of Confidence Calibration in Deep Learning: From Computer Vision to Medical Imaging