Localization with Limited Annotation for Chest X-rays

Eyal Rozenberg, Daniel Freedman, Alex Bronstein
Proceedings of the Machine Learning for Health NeurIPS Workshop, PMLR 116:52-65, 2020.

Abstract

Localization of an object within an image is a common task in medical imaging. Learning to localize or detect objects typically requires the collection of data which has been labelled with bounding boxes or similar annotations, which can be very time consuming and expensive. A technique which could perform such learning with much less annotation would, therefore, be quite valuable. We present such a technique for localization with limited annotation, in which the number of images with bounding boxes can be a small fraction of the total dataset (e.g. less than 1{%}); all other images only possess a whole image label and no bounding box. We propose a novel loss function for tackling this problem; the loss is a continuous relaxation of a well-defined discrete formulation of weakly supervised learning and is numerically well-posed. Furthermore, we propose a new architecture which accounts for both patch dependence and shift-invariance, through the inclusion of CRF layers and anti-aliasing filters, respectively. We apply our technique to the localization of thoracic diseases in chest X-ray images and demonstrate state-of-the-art localization performance on the ChestX-ray14 dataset.

Cite this Paper

BibTeX
@InProceedings{pmlr-v116-rozenberg20a, title = {{Localization with Limited Annotation for Chest X-rays}}, author = {Rozenberg, Eyal and Freedman, Daniel and Bronstein, Alex}, booktitle = {Proceedings of the Machine Learning for Health NeurIPS Workshop}, pages = {52--65}, year = {2020}, editor = {Dalca, Adrian V. and McDermott, Matthew B.A. and Alsentzer, Emily and Finlayson, Samuel G. and Oberst, Michael and Falck, Fabian and Beaulieu-Jones, Brett}, volume = {116}, series = {Proceedings of Machine Learning Research}, month = {13 Dec}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v116/rozenberg20a/rozenberg20a.pdf}, url = {https://proceedings.mlr.press/v116/rozenberg20a.html}, abstract = {Localization of an object within an image is a common task in medical imaging. Learning to localize or detect objects typically requires the collection of data which has been labelled with bounding boxes or similar annotations, which can be very time consuming and expensive. A technique which could perform such learning with much less annotation would, therefore, be quite valuable. We present such a technique for localization with limited annotation, in which the number of images with bounding boxes can be a small fraction of the total dataset (e.g. less than 1{%}); all other images only possess a whole image label and no bounding box. We propose a novel loss function for tackling this problem; the loss is a continuous relaxation of a well-defined discrete formulation of weakly supervised learning and is numerically well-posed. Furthermore, we propose a new architecture which accounts for both patch dependence and shift-invariance, through the inclusion of CRF layers and anti-aliasing filters, respectively. We apply our technique to the localization of thoracic diseases in chest X-ray images and demonstrate state-of-the-art localization performance on the ChestX-ray14 dataset.} }
Endnote
%0 Conference Paper %T Localization with Limited Annotation for Chest X-rays %A Eyal Rozenberg %A Daniel Freedman %A Alex Bronstein %B Proceedings of the Machine Learning for Health NeurIPS Workshop %C Proceedings of Machine Learning Research %D 2020 %E Adrian V. Dalca %E Matthew B.A. McDermott %E Emily Alsentzer %E Samuel G. Finlayson %E Michael Oberst %E Fabian Falck %E Brett Beaulieu-Jones %F pmlr-v116-rozenberg20a %I PMLR %P 52--65 %U https://proceedings.mlr.press/v116/rozenberg20a.html %V 116 %X Localization of an object within an image is a common task in medical imaging. Learning to localize or detect objects typically requires the collection of data which has been labelled with bounding boxes or similar annotations, which can be very time consuming and expensive. A technique which could perform such learning with much less annotation would, therefore, be quite valuable. We present such a technique for localization with limited annotation, in which the number of images with bounding boxes can be a small fraction of the total dataset (e.g. less than 1{%}); all other images only possess a whole image label and no bounding box. We propose a novel loss function for tackling this problem; the loss is a continuous relaxation of a well-defined discrete formulation of weakly supervised learning and is numerically well-posed. Furthermore, we propose a new architecture which accounts for both patch dependence and shift-invariance, through the inclusion of CRF layers and anti-aliasing filters, respectively. We apply our technique to the localization of thoracic diseases in chest X-ray images and demonstrate state-of-the-art localization performance on the ChestX-ray14 dataset.
APA
Rozenberg, E., Freedman, D. & Bronstein, A.. (2020). Localization with Limited Annotation for Chest X-rays. Proceedings of the Machine Learning for Health NeurIPS Workshop, in Proceedings of Machine Learning Research 116:52-65 Available from https://proceedings.mlr.press/v116/rozenberg20a.html.

Related Material