Deep Reinforcement Learning for Subpixel Neural Tracking

Tianhong Dai, Magda Dubois, Kai Arulkumaran, Jonathan Campbell, Cher Bass, Benjamin Billot, Fatmatulzehra Uslu, Vincenzo de Paola, Claudia Clopath, Anil Anthony Bharath
Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning, PMLR 102:130-150, 2019.

Abstract

Automatically tracing elongated structures, such as axons and blood vessels, is a challenging problem in the field of biomedical imaging, but one with many downstream applications. Real, labelled data is sparse, and existing algorithms either lack robustness to different datasets, or otherwise require significant manual tuning. Here, we instead learn a tracking algorithm in a synthetic environment, and apply it to tracing axons. To do so, we formulate tracking as a reinforcement learning problem, and apply deep reinforcement learning techniques with a continuous action space to learn how to track at the subpixel level. We train our model on simple synthetic data and test it on mouse cortical two-photon microscopy images. Despite the domain gap, our model approaches the performance of a heavily engineered tracker from a standard analysis suite for neuronal microscopy. We show that fine-tuning on real data improves performance, allowing better transfer when real labelled data is available. Finally, we demonstrate that our model’s uncertainty measure—a feature lacking in hand-engineered trackers—corresponds with how well it tracks the structure.

Cite this Paper

BibTeX
@InProceedings{pmlr-v102-dai19a, title = {Deep Reinforcement Learning for Subpixel Neural Tracking}, author = {Dai, Tianhong and Dubois, Magda and Arulkumaran, Kai and Campbell, Jonathan and Bass, Cher and Billot, Benjamin and Uslu, Fatmatulzehra and {de Paola}, Vincenzo and Clopath, Claudia and Bharath, {Anil Anthony}}, booktitle = {Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning}, pages = {130--150}, year = {2019}, editor = {Cardoso, M. Jorge and Feragen, Aasa and Glocker, Ben and Konukoglu, Ender and Oguz, Ipek and Unal, Gozde and Vercauteren, Tom}, volume = {102}, series = {Proceedings of Machine Learning Research}, month = {08--10 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v102/dai19a/dai19a.pdf}, url = {https://proceedings.mlr.press/v102/dai19a.html}, abstract = {Automatically tracing elongated structures, such as axons and blood vessels, is a challenging problem in the field of biomedical imaging, but one with many downstream applications. Real, labelled data is sparse, and existing algorithms either lack robustness to different datasets, or otherwise require significant manual tuning. Here, we instead learn a tracking algorithm in a synthetic environment, and apply it to tracing axons. To do so, we formulate tracking as a reinforcement learning problem, and apply deep reinforcement learning techniques with a continuous action space to learn how to track at the subpixel level. We train our model on simple synthetic data and test it on mouse cortical two-photon microscopy images. Despite the domain gap, our model approaches the performance of a heavily engineered tracker from a standard analysis suite for neuronal microscopy. We show that fine-tuning on real data improves performance, allowing better transfer when real labelled data is available. Finally, we demonstrate that our model’s uncertainty measure—a feature lacking in hand-engineered trackers—corresponds with how well it tracks the structure.} }
Endnote
%0 Conference Paper %T Deep Reinforcement Learning for Subpixel Neural Tracking %A Tianhong Dai %A Magda Dubois %A Kai Arulkumaran %A Jonathan Campbell %A Cher Bass %A Benjamin Billot %A Fatmatulzehra Uslu %A Vincenzo de Paola %A Claudia Clopath %A Anil Anthony Bharath %B Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2019 %E M. Jorge Cardoso %E Aasa Feragen %E Ben Glocker %E Ender Konukoglu %E Ipek Oguz %E Gozde Unal %E Tom Vercauteren %F pmlr-v102-dai19a %I PMLR %P 130--150 %U https://proceedings.mlr.press/v102/dai19a.html %V 102 %X Automatically tracing elongated structures, such as axons and blood vessels, is a challenging problem in the field of biomedical imaging, but one with many downstream applications. Real, labelled data is sparse, and existing algorithms either lack robustness to different datasets, or otherwise require significant manual tuning. Here, we instead learn a tracking algorithm in a synthetic environment, and apply it to tracing axons. To do so, we formulate tracking as a reinforcement learning problem, and apply deep reinforcement learning techniques with a continuous action space to learn how to track at the subpixel level. We train our model on simple synthetic data and test it on mouse cortical two-photon microscopy images. Despite the domain gap, our model approaches the performance of a heavily engineered tracker from a standard analysis suite for neuronal microscopy. We show that fine-tuning on real data improves performance, allowing better transfer when real labelled data is available. Finally, we demonstrate that our model’s uncertainty measure—a feature lacking in hand-engineered trackers—corresponds with how well it tracks the structure.
APA
Dai, T., Dubois, M., Arulkumaran, K., Campbell, J., Bass, C., Billot, B., Uslu, F., de Paola, V., Clopath, C. & Bharath, A.A.. (2019). Deep Reinforcement Learning for Subpixel Neural Tracking. Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 102:130-150 Available from https://proceedings.mlr.press/v102/dai19a.html.

Related Material