Abstract
Automated methods for segmentation of ischemic stroke lesions could significantly reduce the workload of radiologists and speed up the beginning of patient treatment. In this paper, we present a method for subacute ischemic stroke lesion segmentation from multispectral magnetic resonance images (MRI). The method involves classification of voxels with a Random Forest algorithm and subsequent classification refinement with contextual clustering. In addition, we utilize the training data to build statistical group-specific templates and use them for calculation of individual voxel-wise differences from the global mean. Our method achieved a Dice coefficient of 0.61 for the leave-one-out cross-validated training data and 0.47 for the testing data of the ISLES challenge 2015.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Pexman, J.H., Barber, P.A., Hill, M.D., Sevick, R.J., Demchuk, A.M., Hudon, M.E., Hu, W.Y., Buchan, A.M.: Use of the Alberta Stroke Program Early CT Score (ASPECTS) for assessing CT scans in patients with acute stroke. Am. J. Neuroradiol. 22(8), 1534–1542 (2001)
Maier, O., Wilms, M., von der Gablentz, J., Krämer, U.M., Münte, T.F., Handels, H.: Extra Tree forests for sub-acute ischemic stroke lesion segmentation in MR sequences. J. Neurosci. Methods 240, 89–100 (2015)
Kabir, Y., Dojat, M., Scherrer, B., Garbay, C., Forbes, F.: Multimodal MRI segmentation of ischemic stroke lesions. In: 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 2007. EMBS 2007, pp. 1595–1598. IEEE, August 2007
Avants, B.B., Tustison, N.J., Song, G., Cook, P.A., Klein, A., Gee, J.C.: A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage 54(3), 2033–2044 (2011)
Breiman, L.: Random forests. Mach. Learn. 45(1), 5–32 (2001)
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., Duchesnay, E.: Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011)
Besag, J.: On the statistical analysis of dirty pictures. J. R. Stat. Soc. Series B (Methodological) 48(3), 259–302 (1986)
Salli, E., Aronen, H.J., Savolainen, S., Korvenoja, A., Visa, A.: Contextual clustering for analysis of functional MRI data. IEEE Trans. Med. Imaging 20(5), 403–414 (2001)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix
Appendix
Generation of Templates
The common template was done in two phases. First, the initial template was formed:
buildtemplateparallel.sh -d 3 -m 1 \(\mathtt {\times }\) 0 \(\mathtt {\times }\) 0 -n 0 -r 1 -t GR -s CC -o [initial template image] -c 0 -j 1 [T1 images]
After that, the final template was built using the initial template:
buildtemplateparallel.sh -d 3 -m 30 \(\mathtt {\times }\) 90 \(\mathtt {\times }\) 20 -n 0 -r 0 -t GR -s CC -o [template image] -z [initial template image] -c 0 [T1 images]
Warping of T1 images to common template was done using antsRegistration tool and the following parameters:
--metric MI[template image, T1 image,1,32] --transform affine[0.25] --convergence 10000 \(\mathtt {\times }\) 10000 \(\mathtt {\times }\) 10000 \(\mathtt {\times }\) 10000 \(\mathtt {\times }\) 10000 --shrink factors 5 \(\mathtt {\times }\) 4 \(\mathtt {\times }\) 3 \(\mathtt {\times }\) 2 \(\mathtt {\times }\) 1 --smoothing-sigmas 4 \(\mathtt {\times }\) 3 \(\mathtt {\times }\) 2 \(\mathtt {\times }\) 1 \(\mathtt {\times }\) 0 --metric CC[template image, T1 image,1,5] --transform SyN[0.25,3.0,0.0] --convergence 50 \(\mathtt {\times }\) 35 \(\mathtt {\times }\) 15 --shrink factors 3 \(\mathtt {\times }\) 2 \(\mathtt {\times }\) 1 --smoothing-sigmas 2 \(\mathtt {\times }\) 1 \(\mathtt {\times }\) 0 --use-histogram-matching 1 --x [lesion image]
Parameters for Random Forest Classifier
Scikit-learn’s function sklearn.ensemble.RandomForestClassifier was used with the following parameters:
\({\texttt {n\_estimators=300, criterion='gini', max\_depth=None, min\_samples\_}}\) \({\texttt {split=2, min\_samples\_leaf=1, min\_weight\_fraction\_leaf=0.0, max\_}}\) \({\texttt {features=4, max\_leaf\_nodes=None, bootstrap=True, oob\_score=False,}}\) \({\texttt {n\_jobs=1, random\_state=None, verbose=0, warm\_start=False,}}\) \({\texttt {class\_weight=None}}\)
Rights and permissions
Copyright information
© 2016 Springer International Publishing Switzerland
About this paper
Cite this paper
Halme, HL., Korvenoja, A., Salli, E. (2016). ISLES (SISS) Challenge 2015: Segmentation of Stroke Lesions Using Spatial Normalization, Random Forest Classification and Contextual Clustering. In: Crimi, A., Menze, B., Maier, O., Reyes, M., Handels, H. (eds) Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. BrainLes 2015. Lecture Notes in Computer Science(), vol 9556. Springer, Cham. https://doi.org/10.1007/978-3-319-30858-6_18
Download citation
DOI: https://doi.org/10.1007/978-3-319-30858-6_18
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-30857-9
Online ISBN: 978-3-319-30858-6
eBook Packages: Computer ScienceComputer Science (R0)