Machine Learning based Analysis of Longitudinal Dynamics of Brain Lesions in Multiple Sclerosis

Master Practical Training Project/ Master’s thesis

Supervision

  • Roxane Licandro, Medical University of Vienna, Department of Biomedical Imaging and Image-guided Therapy, Computational Imaging Research, roxane.licandro@meduniwien.ac.at

Cooperation Partner

  • Department of Neurology, Medical University of Vienna, Vienna, Austria
  • Division of Neuropathology and Neurochemistry, Department of Neurology, Medical University of Vienna, Vienna, Austria
  • Comprehensive Center for Clinical Neurosciences & Mental Health, Medical University of Vienna, Vienna, Austria
  • Biomedical Imaging and Image-guided Therapy, High Field MR Center, Medical University of Vienna, Austria.
  • Center for Brain Research, Medical University of Vienna, Austria.

Problem Statement

Multiple sclerosis (MS) is a chronic inflammatory autoimmune disease of the central nervous system (CNS) causing demyelination in the white and gray matter leading to brain atrophy. To date, the cause of MS has not been identified. Typically diagnosed between the ages of 20 and 40, MS patients present with a wide range of symptoms such as visual impairment, paresthesia, muscle weakness, bladder dysfunction, and fatigue. The neuroinflammatory phase predominates initially – characterized by relapses, but is already accompanied by background progression, which gradually becomes more and more dominant within time -characterized by relapse-independent clinical deterioration. This lifelong chronic disease leads to multiple processes in the human brain. Therefore, main focus of this project is to create computational methods to assess underlying pathogenic mechanisms and to predict progression patterns based on longitudinal 7T Magnetic Resonance Imaging Data. Understanding the processes that drive neurodegeneration in MS patients could ultimately contribute to the development of new treatment strategies and help to assess a potential patient’s treatment response [1].

Performing computational analysis entails the following challenges:

  • Imaging Dynamics: Per patient MR images are acquired over a long time period (< 12 years) resulting in morphometric brain changes: enlargement of ventricles, appearance, expansion shrinkage and disappearance of MS lesions and atrophy of brain tissue [2].
  • Lesion Dynamics: Brain lesions observed in MR images over time show changes in shape and size, but also a reorganization of tissue (remyelination) occurs, resulting in intensity variations.
  • Lack of algorithms for the reliable tracking and classification of longitudinal observations of MS related lesions.

Goal
The aim of this practical training project/master’s thesis is to develop a patient specific approach for MS lesion tracking and longitudinal analysis. Registration (rigid and non-rigid) [3] and machine learning [4-6] will be used in combination with tracking strategies to learn lesion dynamics.

Figure 1: Longitudinal 7T MR scans over 12 years of one MS patient. Corresponding lesion segmentations are displayed for timepoint year 0 and year 11 as overlay. Image courtesy Medical University of Vienna.

Workflow

  • Literature research on existing brain registration and processing approaches
  • Get familiar with MRI data (imaging protocols, reading routines, artefacts, etc.)
  • Development of a processing and lesion tracking pipeline in Python, Tensorflow/Keras or PyTorch.
  • Visualisation of extracted lesion dynamics
  • Evaluation of the system proposed against baseline approaches.
  • Written report/thesis (in English) and final presentation.

For more information please contact Roxane Licandro roxane.licandro@meduniwien.ac.at

References
[1] D. H. Mahad, B. D. Trapp, H. Lassmann, Pathological mechanisms in progressive multiple sclerosis, The Lancet Neurology, Volume 14, Issue 2, 2015, Pages 183-193, ISSN 1474-4422, https://doi.org/10.1016/S1474-4422(14)70256-X.

[2] A. Dal-Bianco, G. Grabner, C. Kronnerwetter, M. Weber, B. Kornek, G. Kasprian, T. Berger, F. Leutmezer, P.S. Rommer, S. Trattnig, H. Lassmann and S. Hametner, Long-term evolution of multiple sclerosis iron rim lesions in 7 T MRI. Brain: a journal of neurology, 144(3), 2015, Pages 833–847, https://doi.org/10.1093/brain/awaa436

[3] https://antspy.readthedocs.io/en/latest/registration.html

[4] M.M. Weeda, I. Brouwer, M.L. de Vos, M.S. de Vries, F. Barkhof, P.J.W. Pouwels, H. Vrenken, Comparing lesion segmentation methods in multiple sclerosis: Input from one manually delineated subject is sufficient for accurate lesion segmentation, NeuroImage: Clinical, Volume 24, 2019, 102074, ISSN 2213-1582, https://doi.org/10.1016/j.nicl.2019.102074.

[5] S. Valverde, M. Cabezas, E. Roura, S. González-Villà, D. Pareto, J. C. Vilanova, L.Ramió-Torrentà, À. Rovira, A. Oliver, X. Lladó, Improving automated multiple sclerosis lesion segmentation with a cascaded 3D convolutional neural network approach, NeuroImage, Volume 155, 2017, Pages 159-168, ISSN 1053-8119, https://doi.org/10.1016/j.neuroimage.2017.04.034.

[6] R. Licandro, J. Hofmanninger, M. Perkonigg, S. Röhrich, M.-A. Weber, M. Wennmann, L. Kintzele, M. Piraud, B. Menze, G. Langs, Asymmetric Cascade Networks for Focal Bone Lesion Prediction in Multiple Myeloma, International Conference on Medical Imaging with Deep Learning (MIDL), London, July 2019. https://arxiv.org/abs/1907.13539

Data Privacy Statement