Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Dec 13, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

Confidence Calibration and Predictive Uncertainty Estimation for Deep Medical Image Segmentation.

Alireza Mehrtash, William M Wells, Clare M Tempany

IEEE Transactions on Medical Imaging

|August 4, 2020

Summary

Fully convolutional neural networks (FCNs) in medical image segmentation are often overconfident. Model ensembling calibrates these networks, improving reliability and out-of-distribution detection for better clinical interpretation.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Can Crowdsourced Annotations Improve AI-based Congestion Scoring For Bedside Lung Ultrasound?

Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention·2026

Same author

EchoAgent: guideline-centric reasoning agent for echocardiography measurement and interpretation.

International journal of computer assisted radiology and surgery·2026

Same author

Point tracking as a temporal Cue for robust myocardial segmentation in echocardiography videos.

International journal of computer assisted radiology and surgery·2026

Same author

TREAT-Netv2: regional wall motion-informed video-tabular fusion for ACS treatment prediction.

International journal of computer assisted radiology and surgery·2026

Same author

Optimizing point-of-care ultrasound video acquisition for probabilistic multi-task heart failure detection.

International journal of computer assisted radiology and surgery·2026

Same author

Shift happens: a fairness-oriented framework for medical classification under hidden bias.

International journal of computer assisted radiology and surgery·2026

Area of Science:

Medical imaging
Artificial intelligence
Computer vision

Background:

Fully convolutional neural networks (FCNs), particularly U-Nets, excel in medical image segmentation.
Batch normalization and Dice loss aid training but lead to poorly calibrated, overconfident predictions.
Unreliable predictions hinder clinical interpretation and trust in AI models.

Purpose of the Study:

To investigate predictive uncertainty estimation in FCNs for medical image segmentation.
To compare cross-entropy and Dice loss for segmentation quality and uncertainty estimation.
To propose and evaluate model ensembling for confidence calibration.

Main Methods:

Systematic comparison of cross-entropy and Dice loss for FCNs.
Implementation of model ensembling to calibrate FCNs trained with batch normalization and Dice loss.

More Related Videos

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images

Published on: April 13, 2013

Related Experiment Videos

Last Updated: Dec 13, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images

Published on: April 13, 2013

Experimental validation across brain, heart, and prostate segmentation tasks.

Main Results:

Model ensembling consistently enhances confidence calibration in FCNs.
Calibrated FCNs demonstrate improved prediction of segmentation quality.
Out-of-distribution test examples are more effectively detected by calibrated models.

Conclusions:

Model ensembling offers a practical solution for confidence calibration in medical image segmentation FCNs.
Calibrated FCNs increase reliability and interpretability in clinical applications.
This study provides valuable insights into uncertainty estimation and OOD detection for AI in medicine.