Local Gamma Augmentation for Ischemic Stroke Lesion Segmentation on MRI

Jon Middleton, Marko Bauer, Kaining Sheng, Jacob Johansen, Mathias Perslev, Silvia Ingala, Mads Nielsen, Akshay Pai
Proceedings of the 5th Northern Lights Deep Learning Conference ({NLDL}), PMLR 233:158-164, 2024.

Abstract

The identification and localisation of pathological tissues in medical images continues to command much attention among deep learning practitioners. When trained on abundant datasets, deep neural networks can match or exceed human performance. However, the scarcity of annotated data complicates the training of these models. Data augmentation techniques can compensate for a lack of training samples. However, many commonly used augmentation methods can fail to provide meaningful samples during model fitting. We present local gamma augmentation, a technique for introducing new instances of intensities in pathological tissues. We leverage local gamma augmentation to compensate for a bias in intensities corresponding to ischemic stroke lesions in human brain MRIs. On three datasets, we show how local gamma augmentation can improve the image-level sensitivity of a deep neural network tasked with ischemic lesion segmentation on magnetic resonance images.

Cite this Paper

BibTeX
@InProceedings{pmlr-v233-middleton24a, title = {Local Gamma Augmentation for Ischemic Stroke Lesion Segmentation on {MRI}}, author = {Middleton, Jon and Bauer, Marko and Sheng, Kaining and Johansen, Jacob and Perslev, Mathias and Ingala, Silvia and Nielsen, Mads and Pai, Akshay}, booktitle = {Proceedings of the 5th Northern Lights Deep Learning Conference ({NLDL})}, pages = {158--164}, year = {2024}, editor = {Lutchyn, Tetiana and Ramírez Rivera, Adín and Ricaud, Benjamin}, volume = {233}, series = {Proceedings of Machine Learning Research}, month = {09--11 Jan}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v233/middleton24a/middleton24a.pdf}, url = {https://proceedings.mlr.press/v233/middleton24a.html}, abstract = {The identification and localisation of pathological tissues in medical images continues to command much attention among deep learning practitioners. When trained on abundant datasets, deep neural networks can match or exceed human performance. However, the scarcity of annotated data complicates the training of these models. Data augmentation techniques can compensate for a lack of training samples. However, many commonly used augmentation methods can fail to provide meaningful samples during model fitting. We present local gamma augmentation, a technique for introducing new instances of intensities in pathological tissues. We leverage local gamma augmentation to compensate for a bias in intensities corresponding to ischemic stroke lesions in human brain MRIs. On three datasets, we show how local gamma augmentation can improve the image-level sensitivity of a deep neural network tasked with ischemic lesion segmentation on magnetic resonance images.} }
Endnote
%0 Conference Paper %T Local Gamma Augmentation for Ischemic Stroke Lesion Segmentation on MRI %A Jon Middleton %A Marko Bauer %A Kaining Sheng %A Jacob Johansen %A Mathias Perslev %A Silvia Ingala %A Mads Nielsen %A Akshay Pai %B Proceedings of the 5th Northern Lights Deep Learning Conference ({NLDL}) %C Proceedings of Machine Learning Research %D 2024 %E Tetiana Lutchyn %E Adín Ramírez Rivera %E Benjamin Ricaud %F pmlr-v233-middleton24a %I PMLR %P 158--164 %U https://proceedings.mlr.press/v233/middleton24a.html %V 233 %X The identification and localisation of pathological tissues in medical images continues to command much attention among deep learning practitioners. When trained on abundant datasets, deep neural networks can match or exceed human performance. However, the scarcity of annotated data complicates the training of these models. Data augmentation techniques can compensate for a lack of training samples. However, many commonly used augmentation methods can fail to provide meaningful samples during model fitting. We present local gamma augmentation, a technique for introducing new instances of intensities in pathological tissues. We leverage local gamma augmentation to compensate for a bias in intensities corresponding to ischemic stroke lesions in human brain MRIs. On three datasets, we show how local gamma augmentation can improve the image-level sensitivity of a deep neural network tasked with ischemic lesion segmentation on magnetic resonance images.
APA
Middleton, J., Bauer, M., Sheng, K., Johansen, J., Perslev, M., Ingala, S., Nielsen, M. & Pai, A.. (2024). Local Gamma Augmentation for Ischemic Stroke Lesion Segmentation on MRI. Proceedings of the 5th Northern Lights Deep Learning Conference ({NLDL}), in Proceedings of Machine Learning Research 233:158-164 Available from https://proceedings.mlr.press/v233/middleton24a.html.

Related Material