GOAL: Gist-set Online Active Learning for Efficient Chest X-ray Image Annotation

Chanh Nguyen, Minh Thanh Huynh, Minh Quan Tran, Ngoc Hoang Nguyen, Mudit Jain, Van Doan Ngo, Tan Duc Vo, Trung Bui, Steven Quoc Hung Truong
Proceedings of the Fourth Conference on Medical Imaging with Deep Learning, PMLR 143:545-553, 2021.

Abstract

Deep learning in medical image analysis often requires an extensive amount of high-quality labeled data for training to achieve Human-level accuracy. We propose Gist-set Online Active Learning (GOAL), a novel solution for limited high-quality labeled data in medical imaging analysis. Our approach advances the existing active learning methods in three aspects. Firstly, we improve the classification performance with fewer manual annotations by presenting a sample selection strategy called gist set selection. Secondly, unlike traditional methods focusing only on random uncertain samples of low prediction confidence, we propose a new method in which only informative uncertain samples are selected for human annotation. Thirdly, we propose an application of online learning where high-confidence samples are automatically selected, iteratively assigned, and pseudo-labels are updated. We validated our approach on two private and one public dataset. The experimental results show that, by applying GOAL, we can reduce required labeled data up to 88% while maintaining the same F1 scores compared to the models trained on full datasets

Cite this Paper

BibTeX
@InProceedings{pmlr-v143-nguyen21a, title = {{GOAL}: Gist-set Online Active Learning for Efficient Chest X-ray Image Annotation}, author = {Nguyen, Chanh and Huynh, Minh Thanh and Tran, Minh Quan and Nguyen, Ngoc Hoang and Jain, Mudit and Ngo, Van Doan and Vo, Tan Duc and Bui, Trung and Truong, Steven Quoc Hung}, booktitle = {Proceedings of the Fourth Conference on Medical Imaging with Deep Learning}, pages = {545--553}, year = {2021}, editor = {Heinrich, Mattias and Dou, Qi and de Bruijne, Marleen and Lellmann, Jan and Schläfer, Alexander and Ernst, Floris}, volume = {143}, series = {Proceedings of Machine Learning Research}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v143/nguyen21a/nguyen21a.pdf}, url = {https://proceedings.mlr.press/v143/nguyen21a.html}, abstract = {Deep learning in medical image analysis often requires an extensive amount of high-quality labeled data for training to achieve Human-level accuracy. We propose Gist-set Online Active Learning (GOAL), a novel solution for limited high-quality labeled data in medical imaging analysis. Our approach advances the existing active learning methods in three aspects. Firstly, we improve the classification performance with fewer manual annotations by presenting a sample selection strategy called gist set selection. Secondly, unlike traditional methods focusing only on random uncertain samples of low prediction confidence, we propose a new method in which only informative uncertain samples are selected for human annotation. Thirdly, we propose an application of online learning where high-confidence samples are automatically selected, iteratively assigned, and pseudo-labels are updated. We validated our approach on two private and one public dataset. The experimental results show that, by applying GOAL, we can reduce required labeled data up to 88% while maintaining the same F1 scores compared to the models trained on full datasets} }
Endnote
%0 Conference Paper %T GOAL: Gist-set Online Active Learning for Efficient Chest X-ray Image Annotation %A Chanh Nguyen %A Minh Thanh Huynh %A Minh Quan Tran %A Ngoc Hoang Nguyen %A Mudit Jain %A Van Doan Ngo %A Tan Duc Vo %A Trung Bui %A Steven Quoc Hung Truong %B Proceedings of the Fourth Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2021 %E Mattias Heinrich %E Qi Dou %E Marleen de Bruijne %E Jan Lellmann %E Alexander Schläfer %E Floris Ernst %F pmlr-v143-nguyen21a %I PMLR %P 545--553 %U https://proceedings.mlr.press/v143/nguyen21a.html %V 143 %X Deep learning in medical image analysis often requires an extensive amount of high-quality labeled data for training to achieve Human-level accuracy. We propose Gist-set Online Active Learning (GOAL), a novel solution for limited high-quality labeled data in medical imaging analysis. Our approach advances the existing active learning methods in three aspects. Firstly, we improve the classification performance with fewer manual annotations by presenting a sample selection strategy called gist set selection. Secondly, unlike traditional methods focusing only on random uncertain samples of low prediction confidence, we propose a new method in which only informative uncertain samples are selected for human annotation. Thirdly, we propose an application of online learning where high-confidence samples are automatically selected, iteratively assigned, and pseudo-labels are updated. We validated our approach on two private and one public dataset. The experimental results show that, by applying GOAL, we can reduce required labeled data up to 88% while maintaining the same F1 scores compared to the models trained on full datasets
APA
Nguyen, C., Huynh, M.T., Tran, M.Q., Nguyen, N.H., Jain, M., Ngo, V.D., Vo, T.D., Bui, T. & Truong, S.Q.H.. (2021). GOAL: Gist-set Online Active Learning for Efficient Chest X-ray Image Annotation. Proceedings of the Fourth Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 143:545-553 Available from https://proceedings.mlr.press/v143/nguyen21a.html.

Related Material