Improving Identically Distributed and Out-of-Distribution Medical Image Classification with Segmentation-Guided Attention in Small Dataset Scenarios

Mariia Rizhko, Lauren Erdman, Mandy Rickard, Kunj Sheth, Daniel Alvarez, Kyla N Velaer, Megan A. Bonnett, Christopher S. Cooper, Gregory E. Tasian, John Weaver, Alice Xiang, Armando J. Lorenzo, Anna Goldenberg
Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, PMLR 250:1282-1296, 2024.

Abstract

We propose a new approach for training medical image classification models using segmentation masks, particularly effective in small dataset scenarios. By guiding the model’s attention with segmentation masks toward relevant features, we significantly improve accuracy for diagnosing Hydronephrosis. Evaluation of our model on identically distributed data showed either the same or better performance with improvement up to 0.28 in AUROC and up to 0.33 in AUPRC. Our method showed better generalization ability than baselines, improving from 0.02 to 0.75 in AUROC and from 0.09 to 0.47 in AUPRC for four different out-of-distribution datasets. The results show that models trained on smaller datasets using our approach can achieve comparable results to those trained on datasets 25 times larger.

Cite this Paper

BibTeX
@InProceedings{pmlr-v250-rizhko24a, title = {Improving Identically Distributed and Out-of-Distribution Medical Image Classification with Segmentation-Guided Attention in Small Dataset Scenarios}, author = {Rizhko, Mariia and Erdman, Lauren and Rickard, Mandy and Sheth, Kunj and Alvarez, Daniel and Velaer, Kyla N and Bonnett, Megan A. and Cooper, Christopher S. and Tasian, Gregory E. and Weaver, John and Xiang, Alice and Lorenzo, Armando J. and Goldenberg, Anna}, booktitle = {Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning}, pages = {1282--1296}, year = {2024}, editor = {Burgos, Ninon and Petitjean, Caroline and Vakalopoulou, Maria and Christodoulidis, Stergios and Coupe, Pierrick and Delingette, Hervé and Lartizien, Carole and Mateus, Diana}, volume = {250}, series = {Proceedings of Machine Learning Research}, month = {03--05 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v250/main/assets/rizhko24a/rizhko24a.pdf}, url = {https://proceedings.mlr.press/v250/rizhko24a.html}, abstract = {We propose a new approach for training medical image classification models using segmentation masks, particularly effective in small dataset scenarios. By guiding the model’s attention with segmentation masks toward relevant features, we significantly improve accuracy for diagnosing Hydronephrosis. Evaluation of our model on identically distributed data showed either the same or better performance with improvement up to 0.28 in AUROC and up to 0.33 in AUPRC. Our method showed better generalization ability than baselines, improving from 0.02 to 0.75 in AUROC and from 0.09 to 0.47 in AUPRC for four different out-of-distribution datasets. The results show that models trained on smaller datasets using our approach can achieve comparable results to those trained on datasets 25 times larger.} }
Endnote
%0 Conference Paper %T Improving Identically Distributed and Out-of-Distribution Medical Image Classification with Segmentation-Guided Attention in Small Dataset Scenarios %A Mariia Rizhko %A Lauren Erdman %A Mandy Rickard %A Kunj Sheth %A Daniel Alvarez %A Kyla N Velaer %A Megan A. Bonnett %A Christopher S. Cooper %A Gregory E. Tasian %A John Weaver %A Alice Xiang %A Armando J. Lorenzo %A Anna Goldenberg %B Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2024 %E Ninon Burgos %E Caroline Petitjean %E Maria Vakalopoulou %E Stergios Christodoulidis %E Pierrick Coupe %E Hervé Delingette %E Carole Lartizien %E Diana Mateus %F pmlr-v250-rizhko24a %I PMLR %P 1282--1296 %U https://proceedings.mlr.press/v250/rizhko24a.html %V 250 %X We propose a new approach for training medical image classification models using segmentation masks, particularly effective in small dataset scenarios. By guiding the model’s attention with segmentation masks toward relevant features, we significantly improve accuracy for diagnosing Hydronephrosis. Evaluation of our model on identically distributed data showed either the same or better performance with improvement up to 0.28 in AUROC and up to 0.33 in AUPRC. Our method showed better generalization ability than baselines, improving from 0.02 to 0.75 in AUROC and from 0.09 to 0.47 in AUPRC for four different out-of-distribution datasets. The results show that models trained on smaller datasets using our approach can achieve comparable results to those trained on datasets 25 times larger.
APA
Rizhko, M., Erdman, L., Rickard, M., Sheth, K., Alvarez, D., Velaer, K.N., Bonnett, M.A., Cooper, C.S., Tasian, G.E., Weaver, J., Xiang, A., Lorenzo, A.J. & Goldenberg, A.. (2024). Improving Identically Distributed and Out-of-Distribution Medical Image Classification with Segmentation-Guided Attention in Small Dataset Scenarios. Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 250:1282-1296 Available from https://proceedings.mlr.press/v250/rizhko24a.html.

Related Material