Test Time Adaptation in Medical Image Segmentation

Test Time Adaptation (TTA) also known as Source-free Domain Adaptation(SFDA), Unsupervised Model Adaptation (UMA).

This is a curated list of TTA in medical image analysis, which also contains the generic TTA methodology. Welcome to add papers by pulling request or raising issues.

Overview

• Tasks
  • Segmentation
  • Classification
  • Detection
• Methodology
  • Information Entropy
  • Pseudo Labeling
  • Batch Normalization
• Dataset
• Miscellaneous

Tasks

Segmentation

Date	First & Last Author	Title	Paper & Code
2020.12	NeeravKarani & EnderKonukoglu	Test-time adaptable neural networks for robust medical image segmentation	MIA, code
2021.06	YufanHea & JerryL.Prince	Autoencoder based self-supervised test-time adaptation for medical image analysis	MIA, code
2021.09	Minhao Hu & Shaoting Zhang	Fully Test-Time Adaptation for Image Segmentation	MICCAI2021
2022.09	Mathilde Bateson & Ismail Ben Ayed	Test-Time Adaptation with Shape Moments for Image Segmentation	MICCAI2022, code
2022.10	Hao Li & Ipek Oguz	Self-supervised Test-Time Adaptation for Medical Image Segmentation	MLCN2022, code
2022.04	ChenYang & YixuanYuan	Source free domain adaptation for medical image segmentation with fourier style mining	MIA, code
2022.05	Yang Hongzheng & Dou Qi	DLTTA: Dynamic Learning of Test-Time Adaptation for Cross-domain Medical Images	TMI2022,code
2022.06	Devavrat Tomar & Behzad Bozorgtabar	OptTTA: Learnable Test-Time Augmentation for Source-Free Medical Image Segmentation Under Domain Shift	MIDL2022, code
2022.02	Neerav Karani & Ender Konukoglu	A Field of Experts Prior for Adapting Neural Networks at Test Time	arxiv

Classification

Date	First & Last Author	Title	Paper & Code
2022.09	Wenao Ma & Qi Dou	Test-Time Adaptation with Calibration of Medical Image Classification Nets for Label Distribution Shift	MICCAI2022,code

Detection

Date	First & Last Author	Title	Paper & Code

Methodology

Information Entropy

Date	First & Last Author	Title	Paper & Code
2021.03	Dequan Wang & Trevor Darrell	Tent: Fully test-time adaptation by entropy minimization	ICLR2021, code

Pseudo Labeling

Batch Normalization

Date	First & Last Author	Title	Paper & Code
2020.06	Zachary Nado & Jasper Snoek	Evaluating Prediction-Time Batch Normalization for Robustness under Covariate Shift	arxiv, code
2021.03	Dequan Wang & Trevor Darrell	Tent: Fully test-time adaptation by entropy minimization	ICLR2021, code

Dataset

Dataset	Task	Modality & No	Description
M&Ms	Cardiac Segmentation	CMR & 375 volumes & 3 classes	Four scanners in six centers
SCGM	Spinal Cord Grey Matter Segmentation	MRI & 2 classes	Four scanners in four centers
SAML	Prostate MRI Segmentation	MRI & 116 volumes & 1 classes	Various scanners in six centers

Miscellaneous

• Awesome Source-free Test-time Adaptation