Dense Segmentation in Selected Dimensions: Application to Retinal Optical Coherence Tomography

Bart Liefers, Cristina González-Gonzalo, Caroline Klaver, Bram van Ginneken, Clara I. Sánchez
Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning, PMLR 102:337-346, 2019.

Abstract

We present a novel convolutional neural network architecture designed for dense segmentation in a subset of the dimensions of the input data. The architecture takes an N-dimensional image as input, and produces a label for every pixel in M output dimensions, where 0 < M < N. Large context is incorporated by an encoder-decoder structure, while funneling shortcut subnetworks provide precise localization. We demonstrate applicability of the architecture on two problems in retinal optical coherence tomography: segmentation of geographic atrophy and segmentation of retinal layers. Performance is compared against two baseline methods, that leave out either the encoder-decoder structure or the shortcut subnetworks. For segmentation of geographic atrophy, an average Dice score of 0.49 ± 0.21 was obtained, compared to 0.46 ± 0.22 and 0.28 ± 0.19 for the baseline methods, respectively. For the layer-segmentation task, the proposed architecture achieved a mean absolute error of 1.305 ± 0.547 pixels compared to 1.967 ± 0.841 and 2.166 ± 0.886± for the baseline methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v102-liefers19a, title = {Dense Segmentation in Selected Dimensions: Application to Retinal Optical Coherence Tomography}, author = {Liefers, Bart and {Gonz\'alez-Gonzalo}, Cristina and Klaver, Caroline and {van Ginneken}, Bram and {S\'anchez}, Clara I.}, booktitle = {Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning}, pages = {337--346}, 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/liefers19a/liefers19a.pdf}, url = {https://proceedings.mlr.press/v102/liefers19a.html}, abstract = {We present a novel convolutional neural network architecture designed for dense segmentation in a subset of the dimensions of the input data. The architecture takes an N-dimensional image as input, and produces a label for every pixel in M output dimensions, where 0 < M < N. Large context is incorporated by an encoder-decoder structure, while funneling shortcut subnetworks provide precise localization. We demonstrate applicability of the architecture on two problems in retinal optical coherence tomography: segmentation of geographic atrophy and segmentation of retinal layers. Performance is compared against two baseline methods, that leave out either the encoder-decoder structure or the shortcut subnetworks. For segmentation of geographic atrophy, an average Dice score of 0.49 ± 0.21 was obtained, compared to 0.46 ± 0.22 and 0.28 ± 0.19 for the baseline methods, respectively. For the layer-segmentation task, the proposed architecture achieved a mean absolute error of 1.305 ± 0.547 pixels compared to 1.967 ± 0.841 and 2.166 ± 0.886± for the baseline methods.} }
Endnote
%0 Conference Paper %T Dense Segmentation in Selected Dimensions: Application to Retinal Optical Coherence Tomography %A Bart Liefers %A Cristina González-Gonzalo %A Caroline Klaver %A Bram van Ginneken %A Clara I. Sánchez %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-liefers19a %I PMLR %P 337--346 %U https://proceedings.mlr.press/v102/liefers19a.html %V 102 %X We present a novel convolutional neural network architecture designed for dense segmentation in a subset of the dimensions of the input data. The architecture takes an N-dimensional image as input, and produces a label for every pixel in M output dimensions, where 0 < M < N. Large context is incorporated by an encoder-decoder structure, while funneling shortcut subnetworks provide precise localization. We demonstrate applicability of the architecture on two problems in retinal optical coherence tomography: segmentation of geographic atrophy and segmentation of retinal layers. Performance is compared against two baseline methods, that leave out either the encoder-decoder structure or the shortcut subnetworks. For segmentation of geographic atrophy, an average Dice score of 0.49 ± 0.21 was obtained, compared to 0.46 ± 0.22 and 0.28 ± 0.19 for the baseline methods, respectively. For the layer-segmentation task, the proposed architecture achieved a mean absolute error of 1.305 ± 0.547 pixels compared to 1.967 ± 0.841 and 2.166 ± 0.886± for the baseline methods.
APA
Liefers, B., González-Gonzalo, C., Klaver, C., van Ginneken, B. & Sánchez, C.I.. (2019). Dense Segmentation in Selected Dimensions: Application to Retinal Optical Coherence Tomography. Proceedings of The 2nd International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 102:337-346 Available from https://proceedings.mlr.press/v102/liefers19a.html.

Related Material