Collaborative study in deep learning for predicting disease activity in multiple sclerosis
ECTRIMS Online Library. Chien C. 09/13/19; 278704; P1665
Claudia Chien
Claudia Chien
Contributions
Abstract

Abstract: P1665

Type: Poster Sessions

Abstract Category: Poster Session 3

C. Chien1,2, F. Eitel3, A. Brandt1,2,4, F. Paul1,2,5, K. Ritter3

1Experimental and Clinical Research Center, Max Delbrück Center for Molecular Medicine & Charité - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, 2NeuroCure Clinical Research Center, 3Department of Psychiatry and Psychotherapy, Bernstein Center for Computational Neuroscience, Berlin Center for Advanced Neuroimaging, Charite - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany, 4Department of Neurology, University of California, Irvine, Irvine, CA, United States, 5Department of Neurology, Charite - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany

Introduction: We propose using state-of-the-art machine (deep) learning techniques, in particular convolutional neural networks (CNNs), for the purpose of extracting individual multiple sclerosis (MS) disease severity and activity markers from structural magnetic resonance imaging (MRI) data.
Objectives: Collecting and procuring an MRI and clinical data repository from centres world-wide for collaborative, expert scientific contributions in the interpretation of CNN-identified biomarkers. This repository will serve to predict disease activity in MS patients from around the world and will validate previous pilot project findings.
Aims: We aim to develop novel techniques for the identification of new MRI-derived clinically applicable biomarkers for the complex task of personalised prediction of disease activity in MS patients.
Methods: In a pilot study, MS patients (n=76, age=43±12, F:M=42:34) and HC (n=71, age=38±13, F:M=46:25) MRIs from the NeuroCure Clinical Research Center were evaluated by CNNs that were pre-trained with Alzheimer's data (n=921 MRI scans) from the Alzheimer's Disease Neuroimaging Initiative to classify patient groups.
To further validate and visualise how CNNs might help in predicting disease activity based on structural MRI data, we will collect MS T1- and T2-weighted cerebral scans including the upper cervical cord, in addition to clinical follow-up data for a prospective study, fine-tuned for MS MRI biomarker detection.
Results: Pilot study results of transfer learning from CNNs trained with Alzheimer patient data gave a classification accuracy of 87%, where the algorithm independently learned patterns in MRIs to separate MS from HC. Subsequent visualisation analysis revealed when the CNN was fine-tuned to the MS dataset, it focused not only on lesion voxels, but also incorporates information from lesion location, as well as non-lesional white and grey matter areas, such as the thalamus.
Conclusions: Initial results show immense promise in using deep learning CNNs to evaluate MS patient MRIs, which can automatically detect differences between MS and HC, with the possibility for personalised disease activity tracking and treatment evaluation. For robust, reliable, and clinically relevant future applications, further data contribution is required. Together, with neurologists, neuroradiologists, and neuroscience researchers, the data will be analysed in a collaborative, expert-led, and world-wide manner.
Disclosure: Funding This project is supported by the Deutsche Multiple Sklerose Gesellschaft Bundesverband e.V.
Conflicts of Interest: CC has nothing to disclose.
FE has nothing to disclose.
AUB is cofounder and holds shares of Motognosis GmbH and Nocturne GmbH. He is named as inventor on several patent applications describing multiple sclerosis serum biomarkers, perceptive visual computing-based motion analysis and retinal image analysis; none of this is related to the present abstract.
FP has received research support from Bayer, Novartis, Biogen Idec, Teva, Sanofi-Aventis/Genzyme, Merck Serono, Alexion, Chugai, Arthur Arnstein Foundation Berlin, Guthy Jackson Charitable Foundation and the US National Multiple Sclerosis Society; has received travel funding and/or speaker honoraria from Bayer, Novartis, Biogen Idec, Teva, Sanofi-Aventis/Genzyme and Merck Serono; and has consulted for Sanofi Genzyme, Biogen Idec and MedImmune; none of this is related to the present abstract.
KR has nothing to disclose.

Abstract: P1665

Type: Poster Sessions

Abstract Category: Poster Session 3

C. Chien1,2, F. Eitel3, A. Brandt1,2,4, F. Paul1,2,5, K. Ritter3

1Experimental and Clinical Research Center, Max Delbrück Center for Molecular Medicine & Charité - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, 2NeuroCure Clinical Research Center, 3Department of Psychiatry and Psychotherapy, Bernstein Center for Computational Neuroscience, Berlin Center for Advanced Neuroimaging, Charite - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany, 4Department of Neurology, University of California, Irvine, Irvine, CA, United States, 5Department of Neurology, Charite - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany

Introduction: We propose using state-of-the-art machine (deep) learning techniques, in particular convolutional neural networks (CNNs), for the purpose of extracting individual multiple sclerosis (MS) disease severity and activity markers from structural magnetic resonance imaging (MRI) data.
Objectives: Collecting and procuring an MRI and clinical data repository from centres world-wide for collaborative, expert scientific contributions in the interpretation of CNN-identified biomarkers. This repository will serve to predict disease activity in MS patients from around the world and will validate previous pilot project findings.
Aims: We aim to develop novel techniques for the identification of new MRI-derived clinically applicable biomarkers for the complex task of personalised prediction of disease activity in MS patients.
Methods: In a pilot study, MS patients (n=76, age=43±12, F:M=42:34) and HC (n=71, age=38±13, F:M=46:25) MRIs from the NeuroCure Clinical Research Center were evaluated by CNNs that were pre-trained with Alzheimer's data (n=921 MRI scans) from the Alzheimer's Disease Neuroimaging Initiative to classify patient groups.
To further validate and visualise how CNNs might help in predicting disease activity based on structural MRI data, we will collect MS T1- and T2-weighted cerebral scans including the upper cervical cord, in addition to clinical follow-up data for a prospective study, fine-tuned for MS MRI biomarker detection.
Results: Pilot study results of transfer learning from CNNs trained with Alzheimer patient data gave a classification accuracy of 87%, where the algorithm independently learned patterns in MRIs to separate MS from HC. Subsequent visualisation analysis revealed when the CNN was fine-tuned to the MS dataset, it focused not only on lesion voxels, but also incorporates information from lesion location, as well as non-lesional white and grey matter areas, such as the thalamus.
Conclusions: Initial results show immense promise in using deep learning CNNs to evaluate MS patient MRIs, which can automatically detect differences between MS and HC, with the possibility for personalised disease activity tracking and treatment evaluation. For robust, reliable, and clinically relevant future applications, further data contribution is required. Together, with neurologists, neuroradiologists, and neuroscience researchers, the data will be analysed in a collaborative, expert-led, and world-wide manner.
Disclosure: Funding This project is supported by the Deutsche Multiple Sklerose Gesellschaft Bundesverband e.V.
Conflicts of Interest: CC has nothing to disclose.
FE has nothing to disclose.
AUB is cofounder and holds shares of Motognosis GmbH and Nocturne GmbH. He is named as inventor on several patent applications describing multiple sclerosis serum biomarkers, perceptive visual computing-based motion analysis and retinal image analysis; none of this is related to the present abstract.
FP has received research support from Bayer, Novartis, Biogen Idec, Teva, Sanofi-Aventis/Genzyme, Merck Serono, Alexion, Chugai, Arthur Arnstein Foundation Berlin, Guthy Jackson Charitable Foundation and the US National Multiple Sclerosis Society; has received travel funding and/or speaker honoraria from Bayer, Novartis, Biogen Idec, Teva, Sanofi-Aventis/Genzyme and Merck Serono; and has consulted for Sanofi Genzyme, Biogen Idec and MedImmune; none of this is related to the present abstract.
KR has nothing to disclose.

By clicking “Accept Terms & all Cookies” or by continuing to browse, you agree to the storing of third-party cookies on your device to enhance your user experience and agree to the user terms and conditions of this learning management system (LMS).

Cookie Settings
Accept Terms & all Cookies