Calibrating the Dice Loss to Handle Neural Network Overconfidence for Biomedical Image Segmentation.
The Dice similarity coefficient (DSC) is both a widely used metric and loss function for biomedical image segmentation due to its robustness to class imbalance. However, it is well known that the DSC loss is poorly calibrated, resulting in overconfident predictions that cannot be usefully interprete...
| Publicado en: | Journal of Digital Imaging Vol. 36; no. 2; pp. 739 - 753 |
|---|---|
| Autores principales: | , , , , , |
| Formato: | diagnostic images equations & formulas pictorial research tables/charts Journal Article |
| Publicado: |
Springer Nature
Apr2023
|
| Acceso en línea: | Ver este registro en EBSCOhost |
| fields | @attributes: recordID: 1 pdfLink: plink: https://search.ebscohost.com/login.aspx?direct=true&db=ccm&AN=162679411&site=ehost-live header: @attributes: shortDbName: ccm uiTerm: 162679411 longDbName: CINAHL Complete uiTag: AN controlInfo: bkinfo: dissinfo: jinfo: jid: 08971889 DOQ jtl: Journal of Digital Imaging issn: 08971889 maglogo: N pubinfo: dt: Apr2023 vid: 36 iid: 2 pid: 237 pub: Springer Nature place: New York, New York artinfo: ui: 162679411 160624064 162679411 162679411 10.1007/s10278-022-00735-3 162679411 ppf: 739 ppct: 14 formats: fmt: – @attributes: type: T – @attributes: type: P tig: atl: Calibrating the Dice Loss to Handle Neural Network Overconfidence for Biomedical Image Segmentation. aug: au: Yeung, Michael Rundo, Leonardo Nan, Yang Sala, Evis Schönlieb, Carola-Bibiane Yang, Guang affil: Department of Radiology, University of Cambridge, Hills Rd, CB2 0QQ, Cambridge, UK sug: subj: Image Processing, Computer Assisted Methods Neural Networks (Computer) Models, Statistical Calibration Human Prediction Models Diagnostic Errors Prevention and Control Imaging, Three-Dimensional Precision Recall Bias Deep Learning ab: The Dice similarity coefficient (DSC) is both a widely used metric and loss function for biomedical image segmentation due to its robustness to class imbalance. However, it is well known that the DSC loss is poorly calibrated, resulting in overconfident predictions that cannot be usefully interpreted in biomedical and clinical practice. Performance is often the only metric used to evaluate segmentations produced by deep neural networks, and calibration is often neglected. However, calibration is important for translation into biomedical and clinical practice, providing crucial contextual information to model predictions for interpretation by scientists and clinicians. In this study, we provide a simple yet effective extension of the DSC loss, named the DSC++ loss, that selectively modulates the penalty associated with overconfident, incorrect predictions. As a standalone loss function, the DSC++ loss achieves significantly improved calibration over the conventional DSC loss across six well-validated open-source biomedical imaging datasets, including both 2D binary and 3D multi-class segmentation tasks. Similarly, we observe significantly improved calibration when integrating the DSC++ loss into four DSC-based loss functions. Finally, we use softmax thresholding to illustrate that well calibrated outputs enable tailoring of recall-precision bias, which is an important post-processing technique to adapt the model predictions to suit the biomedical or clinical task. The DSC++ loss overcomes the major limitation of the DSC loss, providing a suitable loss function for training deep learning segmentation models for use in biomedical and clinical practice. Source code is available at https://github.com/mlyg/DicePlusPlus. pubtype: Academic Journal doctype: diagnostic images equations & formulas pictorial research tables/charts Journal Article ougenre: Article language: English refInfo: holdings: @attributes: islocal: N |
|---|