Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2022 Oct 26:16:975862.
doi: 10.3389/fnins.2022.975862. eCollection 2022.

New multiple sclerosis lesion segmentation and detection using pre-activation U-Net

Pooya Ashtari  1   2 , Berardino Barile  1   2 , Sabine Van Huffel  1 , Dominique Sappey-Marinier  2
Affiliations

New multiple sclerosis lesion segmentation and detection using pre-activation U-Net

Pooya Ashtari et al. Front Neurosci. .

Abstract

Automated segmentation of new multiple sclerosis (MS) lesions in 3D MRI data is an essential prerequisite for monitoring and quantifying MS progression. Manual delineation of such lesions is time-consuming and expensive, especially because raters need to deal with 3D images and several modalities. In this paper, we propose Pre-U-Net, a 3D encoder-decoder architecture with pre-activation residual blocks, for the segmentation and detection of new MS lesions. Due to the limited training set and the class imbalance problem, we apply intensive data augmentation and use deep supervision to train our models effectively. Following the same U-shaped architecture but different blocks, Pre-U-Net outperforms U-Net and Res-U-Net on the MSSEG-2 dataset, achieving a Dice score of 40.3% on new lesion segmentation and an F1 score of 48.1% on new lesion detection. The codes and trained models are publicly available at https://github.com/pashtari/xunet.

Keywords: U-Net; multiple sclerosis; new lesions; pre-activation; segmentation.

PubMed Disclaimer

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

Figure 1
Figure 1
Qualitative results on new MS lesion segmentation. The three examples are from three different patients in the test set. The new lesions are shown in red in the segmentation maps. The new lesions circled in yellow (rows 2-3 and column 6) are successfully detected only by Pre-U-Net, while the new lesion circled in blue (row 3 and column 3) is not captured by any of the models, representing a very difficult case. The patient-wise Dice score for each example is displayed on the segmentation map.
Figure 2
Figure 2
The proposed encoder-decoder architecture. The two lower-resolution auxiliary maps are only used in the training phase as deep supervisions.
Figure 3
Figure 3
The proposed blocks. (A) U-Net block. (B) Res-U-Net block. (C) Pre-U-Net block.
Figure 4
Figure 4
Comparison of different models on the MSSEG-2 test set. (A–C) Show box plots of Dice score (%), F1 score (%), and Hausdorff Distance (mm), respectively. The asterisks indicate how significantly a model score differs from those of the other baselines when using a pairwise Wilcoxon signed-rank test (*p-value < 0.05, **p-value < 0.01).
See this image and copyright information in PMC

References

    1. Ashtari P., Barile B., Van Huffel S., Sappey-Marinier D. (2021a). “Longitudinal multiple sclerosis lesion segmentation using pre-activation U-Net,” in MSSEG-2 Challenge Proceedings: Multiple Sclerosis New Lesions Segmentation Challenge Using a Data Management and Processing Infrastructure, (Strasbourg), 45–51. Available online at: https://hal.inria.fr/hal-03358968v1/document#page=54
    1. Ashtari P., Maes F., Van Huffel S. (2021b). Low-rank convolutional networks for brain tumor segmentation, in International MICCAI Brainlesion Workshop: BrainLes 2020. Lecture Notes in Computer Science, Vol. 12658, eds Crimi A., Bakas S. (Cham: Springer; ), 470–480. 10.1007/978-3-030-72084-1_42 - DOI
    1. Aslani S., Dayan M., Storelli L., Filippi M., Murino V., Rocca M. A., et al. . (2019). Multi-branch convolutional neural network for multiple sclerosis lesion segmentation. NeuroImage 196, 1–15. 10.1016/j.neuroimage.2019.03.068 - DOI - PubMed
    1. Çiçek Ö., Abdulkadir A., Lienkamp S. S., Brox T., Ronneberger O. (2016). 3D U-Net: learning dense volumetric segmentation from sparse annotation, in Medical Image Computing and Computer-Assisted Intervention-MICCAI 2016, eds Ourselin S., Joskowicz L., Sabuncu M. R., Unal G., Wells W. (Cham: Springer; ), 424–432. 10.1007/978-3-319-46723-8_49 - DOI
    1. Commowick O., Istace A., Kain M., Laurent B., Leray F., Simon M., et al. . (2018). Objective evaluation of multiple sclerosis lesion segmentation using a data management and processing infrastructure. Sci. Rep. 8, 1–17. 10.1038/s41598-018-31911-7 - DOI - PMC - PubMed

LinkOut - more resources