2024-01-23-valanarasu24a.md

title	abstract	layout	series	publisher	issn	id	month	tex_title	firstpage	lastpage	page	order	cycles	bibtex_author	author	date	address	container-title	volume	genre	issued	pdf	extras
On-the-Fly Test-time Adaptation for Medical Image Segmentation	One major problem in deep learning-based solutions for medical imaging is the drop in performance when a model is tested on a data distribution different from the one that it is trained on. Adapting the source model to target data distribution at test-time is an efficient solution for the data-shift problem. Previous methods solve this by adapting the model to target distribution by using techniques like entropy minimization or regularization. In these methods, the models are still updated by back-propagation using an unsupervised loss on complete test data distribution. In real-world clinical settings, it makes more sense to adapt a model to a new test image on-the-fly and avoid model update during inference due to privacy concerns and lack of computing resource at deployment. To this end, we propose a new setting - On-the-Fly Adaptation which is zero-shot and episodic i.e., the model is adapted to a single image at a time and also does not perform any back-propagation during test-time). To achieve this, we propose a new framework called Adaptive UNet where each convolutional block is equipped with an adaptive batch normalization layer to adapt the features with respect to a domain code. The domain code is generated using a pre-trained encoder trained on a large corpus of medical images. During test-time, the model takes in just the new test image and generates a domain code to adapt the features of source model according to the test data. We validate the performance on both 2D and 3D data distribution shifts where we get a better performance compared to previous test-time adaptation methods.	inproceedings	Proceedings of Machine Learning Research	PMLR	2640-3498	valanarasu24a	0	On-the-Fly Test-time Adaptation for Medical Image Segmentation	586	598	586-598	586	false	Valanarasu, Jeya Maria Jose and Guo, Pengfei and VS, Vibashan and Patel, Vishal M.	given family Jeya Maria Jose Valanarasu given family Pengfei Guo given family Vibashan VS given family Vishal M. Patel	2024-01-23		Medical Imaging with Deep Learning	227	inproceedings	date-parts 2024 1 23	https://proceedings.mlr.press/v227/valanarasu24a/valanarasu24a.pdf