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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

In medical image classification tasks, it is common to find that the numberof normal samples far exceeds the number of abnormal samples. In suchclass-imbalanced situations, reliable training of deep neural networkscontinues to be a major challenge. Under these circumstances, the predictedclass probabilities may be biased toward the majority class. Calibration hasbeen suggested to alleviate some of these effects. However, there isinsufficient analysis explaining when and whether calibrating a model would bebeneficial in improving performance. In this study, we perform a systematicanalysis of the effect of model calibration on its performance on two medicalimage modalities, namely, chest X-rays and fundus images, using various deeplearning classifier backbones. For this, we study the following variations: (i)the degree of imbalances in the dataset used for training; (ii) calibrationmethods; and (iii) two classification thresholds, namely, default decisionthreshold of 0.5, and optimal threshold from precision-recall curves. Ourresults indicate that at the default operating threshold of 0.5, theperformance achieved through calibration is significantly superior (p < 0.05)to using uncalibrated probabilities. However, at the PR-guided threshold, thesegains are not significantly different (p > 0.05). This finding holds for bothimage modalities and at varying degrees of imbalance.

Does deep learning model calibration improve performance in class-imbalanced medical image classification?