(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Machine-learning-based prediction of disability progression in multiple sclerosis: An observational, international, multi-center study [1]

['Edward De Brouwer', 'Esat-Stadius', 'Ku Leuven', 'Thijs Becker', 'I-Biostat', 'Hasselt University', 'Data Science Institute', 'Lorin Werthen-Brabants', 'Sumo', 'Idlab']

Date: 2024-08

Abstract Background Disability progression is a key milestone in the disease evolution of people with multiple sclerosis (PwMS). Prediction models of the probability of disability progression have not yet reached the level of trust needed to be adopted in the clinic. A common benchmark to assess model development in multiple sclerosis is also currently lacking. Methods Data of adult PwMS with a follow-up of at least three years from 146 MS centers, spread over 40 countries and collected by the MSBase consortium was used. With basic inclusion criteria for quality requirements, it represents a total of 15, 240 PwMS. External validation was performed and repeated five times to assess the significance of the results. Transparent Reporting for Individual Prognosis Or Diagnosis (TRIPOD) guidelines were followed. Confirmed disability progression after two years was predicted, with a confirmation window of six months. Only routinely collected variables were used such as the expanded disability status scale, treatment, relapse information, and MS course. To learn the probability of disability progression, state-of-the-art machine learning models were investigated. The discrimination performance of the models is evaluated with the area under the receiver operator curve (ROC-AUC) and under the precision recall curve (AUC-PR), and their calibration via the Brier score and the expected calibration error. All our preprocessing and model code are available at https://gitlab.com/edebrouwer/ms_benchmark, making this task an ideal benchmark for predicting disability progression in MS. Findings Machine learning models achieved a ROC-AUC of 0⋅71 ± 0⋅01, an AUC-PR of 0⋅26 ± 0⋅02, a Brier score of 0⋅1 ± 0⋅01 and an expected calibration error of 0⋅07 ± 0⋅04. The history of disability progression was identified as being more predictive for future disability progression than the treatment or relapses history. Conclusions Good discrimination and calibration performance on an external validation set is achieved, using only routinely collected variables. This suggests machine-learning models can reliably inform clinicians about the future occurrence of progression and are mature for a clinical impact study.

Author summary Models that accurately predict disability progression in individuals with multiple sclerosis (MS) have the potential to greatly benefit both patients and medical professionals. By aiding in life planning and treatment decision-making, these predictive models can enhance the overall quality of care for people with MS. While previous academic literature has demonstrated the feasibility of predicting disability progression, recent systematic reviews have shed light on several methodological limitations within the existing research. These reviews have highlighted concerns such as the absence of probability calibration assessment, potential biases in cohort selection, and insufficient external validation. Furthermore, the datasets examined often include variables that are not routinely collected in clinical settings or readily available for digital analysis. Consequently, it remains uncertain whether the models identified in these systematic reviews can be effectively implemented in a clinical context. Compounding this issue, the lack of availability of data and analysis code makes it challenging to compare results across different publications. To address these gaps, this study endeavors to develop and validate a machine-learning-based prediction model using the largest longitudinal patient cohort ever assembled for disability progression prediction in MS. Leveraging data from MSBase, a comprehensive international data registry comprising information from multiple MS centers, we aim to create robust models capable of accurately predicting the probability of disability progression. The integration of machine learning models into routine clinical practice has the potential to greatly enhance treatment decision-making and life planning for individuals with MS. The models developed through this study could be subsequently evaluated in a clinical impact study involving MS centers participating in MSBase. This research represents a significant advancement towards the practical application of machine learning models in improving the treatment and care of individuals with MS.

Citation: De Brouwer E, Becker T, Werthen-Brabants L, Dewulf P, Iliadis D, Dekeyser C, et al. (2024) Machine-learning-based prediction of disability progression in multiple sclerosis: An observational, international, multi-center study. PLOS Digit Health 3(7): e0000533. https://doi.org/10.1371/journal.pdig.0000533 Editor: Ryan S. McGinnis, Wake Forest University School of Medicine, UNITED STATES OF AMERICA Received: June 29, 2023; Accepted: May 14, 2024; Published: July 25, 2024 Copyright: © 2024 De Brouwer et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: The data set used in this study is available upon request to the MSBase principal investigators included in the study. MSBase operates as a single point of contact to facilitate the data sharing agreements with the individual data custodians. Inquiries should be addressed at [email protected]. Data is restricted behind a request to ensure a controlled usage of patients data and to stay inline with specific data ownership requirements. The data processing and training scripts to reproduce all experiments are publicly available at https://gitlab.com/edebrouwer/ms_benchmark. Funding: This study was funded by the Research Foundation Flanders (FWO) and the Flemish government through the Onderzoeksprogramma Artificiele Intelligentie (AI) Vlaanderen program (https://www.flandersairesearch.be/en). This funding was awarded to YM, LB, TD, DD, WW, and BDB and funded EBD, TB, LWB, PD, DI, MS, YM, LB, TD, DD, WW, and BDB. EDB was also concomitantly funded by a FWO-SB fellowship (1S98821N - https://fwo.be). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors declare no competing non-financial interests but the following competing financial interests: - Dana Horakova received speaker honoraria and consulting fees from Biogen, Merck, Teva, Roche, Sanofi Genzyme, and Novartis, as well as support for research activities from Biogen and Czech Minsitry of Education [project Progres Q27/LF1]. - Francesco Patti received speaker honoraria and advisory board fees from Almirall, Bayer, Biogen, Celgene, Merck, Novartis, Roche, Sanofi-Genzyme and TEVA. He received research funding from Biogen, Merck, FISM (Fondazione Italiana Sclerosi Multipla), Reload Onlus Association and University of Catania. - Guillermo Izquierdo received speaking honoraria from Biogen, Novartis, Sanofi, Merck, Roche, Almirall and Teva. - Sara Eichau received speaker honoraria and consultant fees from Biogen Idec, Novartis, Merck, Bayer, Sanofi Genzyme, Roche and Teva. - Marc Girard received consulting fees from Teva Canada Innovation, Biogen, Novartis and Genzyme Sanofi; lecture payments from Teva Canada Innovation, Novartis and EMD. He has also received a research grant from Canadian Institutes of Health Research. - Alessandra Lugaresi has served as a Biogen, Bristol Myers Squibb, Merck Serono, Novartis, Roche, Sanofi/ Genzyme and Teva Advisory Board Member. She received congress and travel/accommodation expense compensations or speaker honoraria from Biogen, Merck, Mylan, Novartis, Roche, Sanofi/Genzyme, Teva and Fondazione Italiana Sclerosi Multipla (FISM). Her institutions received research grants from Novartis and Sanofi Genzyme. - Pierre Grammond has served in advisory boards for Novartis, EMD Serono, Roche, Biogen idec, Sanofi Genzyme, Pendopharm and has received grant support from Genzyme and Roche, has received research grants for his institution from Biogen idec, Sanofi Genzyme, EMD Serono. - Tomas Kalincik served on scientific advisory boards for BMS, Roche, Janssen, Sanofi Genzyme, Novartis, Merck and Biogen, steering committee for Brain Atrophy Initiative by Sanofi Genzyme, received conference travel support and/or speaker honoraria from WebMD Global, Eisai, Novartis, Biogen, Sanofi-Genzyme, Teva, BioCSL and Merck and received research or educational event support from Biogen, Novartis, Genzyme, Roche, Celgene and Merck. - Raed Alroughani received honoraria as a speaker and for serving on scientific advisory boards from Bayer, Biogen, GSK, Merck, Novartis, Roche and Sanofi-Genzyme. - Francois Grand’Maison received honoraria or research funding from Biogen, Genzyme, Novartis, Teva Neurosciences, Mitsubishi and ONO Pharmaceuticals. - Murat Terzi received travel grants from Novartis, Bayer-Schering, Merck and Teva; has participated in clinical trials by Sanofi Aventis, Roche and Novartis. - Jeannette Lechner-Scott travel compensation from Novartis, Biogen, Roche and Merck. Her institution receives the honoraria for talks and advisory board commitment as well as research grants from Biogen, Merck, Roche, TEVA and Novartis. - Samia J. Khoury received compensation for participation in the Novartis Maestro program. - Vincent van Pesch has received travel grants from Merck, Biogen, Sanofi, Bristol Myers Squibb, Almirall and Roche; his institution receives honoraria for consultancy and lectures and research grants from Roche, Biogen, Sanofi, Merck, Bristol Myers Squibb, Janssen, Almirall and Novartis Pharma. - Radek Ampapa received conference travel support from Novartis, Teva, Biogen, Bayer and Merck and has participated in a clinical trials by Biogen, Novartis, Teva and Actelion. - Daniele Spitaleri received honoraria as a consultant on scientific advisory boards by Bayer-Schering, Novartis and Sanofi-Aventis and compensation for travel from Novartis, Biogen, Sanofi Aventis, Teva and Merck. - Claudio Solaro served on scientific advisory boards for Merck, Genzyme, Almirall, and Biogen; received honoraria and travel grants from Sanofi Aventis, Novartis, Biogen, Merck, Genzyme and Teva. - Davide Maimone served on scientific advisory boards for Bayer, Biogen, Merck, Sanofi-Genzyme, Novartis, Roche, and Almirall; received honoraria and travel grants from Sanofi Genzyme, Novartis, Biogen, Merck, and Roche. - Gerardo Iuliano (retired - no PI successor but has approved ongoing use of data) had travel/accommodations/meeting expenses funded by Bayer Schering, Biogen, Merck, Novartis, Sanofi Aventis, and Teva. - Bart Van Wijmeersch received research and travel grants, honoraria for MS-Expert advisor and Speaker fees from Bayer-Schering, Biogen, Sanofi Genzyme, Merck, Novartis, Roche and Teva. - Tamara Castillo Triviño received speaking/consulting fees and/or travel funding from Bayer, Biogen, Merck, Novartis, Roche, Sanofi-Genzyme and Teva. - Jose Luis Sanchez-Menoyo accepted travel compensation from Novartis, Merck and Biogen, speaking honoraria from Biogen, Novartis, Sanofi, Merck, Almirall, Bayer and Teva and has participated in clinical trials by Biogen, Merck and Roche - Guy Laureys received travel and/or consultancy compensation from Sanofi-Genzyme, Roche, Teva, Merck, Novartis, Celgene, Biogen. - Anneke van der Walt served on advisory boards and receives unrestricted research grants from Novartis, Biogen, Merck and Roche She has received speaker’s honoraria and travel support from Novartis, Roche, and Merck. She receives grant support from the National Health and Medical Research Council of Australia and MS Research Australia. - Jiwon Oh has received research funding from the MS Society of Canada, National MS Society, Brain Canada, Biogen, Roche, EMD Serono (an affiliate of Merck KGaA); and personal compensation for consulting or speaking from Alexion, Biogen, Celgene (BMS), EMD Serono (an affiliate of Merck KGaA), Novartis, Roche, and Sanofi-Genzyme. - Ayse Altintas received speaker honoraria from Merck, Alexion,; received travel and registration grants from Merck, Biogen - Gen Pharma, Roche, Sanofi-Genzyme. - Yara Fragoso received honoraria as a consultant on scientific advisory boards by Novartis, Teva, Roche and Sanofi-Aventis and compensation for travel from Novartis, Biogen, Sanofi Aventis, Teva, Roche and Merck. - Tunde Csepany received speaker honoraria/ conference travel support from Bayer Schering, Biogen, Merck, Novartis, Roche, Sanofi-Aventis and Teva. - Suzanne Hodgkinson received honoraria and consulting fees from Novartis, Bayer Schering and Sanofi, and travel grants from Novartis, Biogen Idec and Bayer Schering. - Norma Deri received funding from Bayer, Merck, Biogen, Genzyme and Novartis. - Bruce Taylor received funding for travel and speaker honoraria from Bayer Schering Pharma, CSL Australia, Biogen and Novartis, and has served on advisory boards for Biogen, Novartis, Roche and CSL Australia. - Fraser Moore participated in clinical trials sponsored by EMD Serono and Novartis. - Orla Gray received honoraria as consultant on scientific advisory boards for Genzyme, Biogen, Merck, Roche and Novartis; has received travel grants from Biogen, Merck, Roche and Novartis; has participated in clinical trials by Biogen and Merck. - Csilla Rozsa received speaker honoraria from Bayer Schering, Novartis and Biogen, congress and travel expense compensations from Biogen, Teva, Merck and Bayer Schering. - Allan Kermode received speaker honoraria and scientific advisory board fees from Bayer, BioCSL, Biogen, Genzyme, Innate Immunotherapeutics, Merck, Novartis, Sanofi, Sanofi-Aventis, and Teva. - Magdolna Simo received speaker honoraria from Novartis, Biogen, Bayer Schering; congress/travel compensation from Teva, Biogen, Merck, Bayer Schering. - Todd Hardy has received speaking fees or received honoraria for serving on advisory boards for Biogen, Merck, Teva, Novartis, Roche, Bristol-Myers Squibb and Sanofi-Genzyme, is Co-Editor of Advances in Clinical Neurosciences and Rehabilitation, and serves on the editorial board of Journal of Neuroimmunology and Frontiers in Neurology. - Nikolaos Grigoriadis received honoraria, consultancy/lecture fees, travel support and research grants from Biogen Idec, Biologix, Novartis, TEVA, Bayer, Merck Serono, Genesis Pharma, Sanofi – Genzyme, ROCHE, Cellgene, ELPEN and research grants from Hellenic Ministry of Development.

Introduction Multiple sclerosis (MS) is a chronic autoimmune disease of the central nervous system [1]. A recent census estimated more than 2⋅8 million people are currently living with MS [2], which causes a wide variety of symptoms such as mobility problems, cognitive impairment, pain, and fatigue. Importantly, the rate of disability progression is highly variable among people with MS (PwMS) [3]. This heterogeneity makes the personalization of care difficult and prognostic models are thus of high relevance for medical professionals, as they could contribute to better individualized treatment decisions. Indeed, a more aggressive treatment could be prescribed in case of a negative prognosis. Moreover, surveys indicate that PwMS are interested in their prognosis [4], which could help them with planning their lives. There is a large amount of literature on prognostic MS models [5–10]. Some prognostic models are or were at some point available as web tools. However, with the exception of Tintore et al. [10] that focuses on conversion to MS, none have been integrated into clinical practice and no clinical impact studies have been performed [5, 6]. Because MS is a complex chronic disease that is often treated within a multidisciplinary context, the performance of a prognostic model studied in isolation from its clinical context gives limited information on its clinical relevance [11, 12]. Recent systematic reviews have highlighted several methodological issues within the current literature [5, 6], such as the lack of calibration or a possible significant bias in the cohort selection. Moreover, the investigated datasets are rarely made available. They furthermore often contain variables that are not routinely collected within the current clinical workflow (e.g. neurofilament light chain) or are not readily available for digital analysis (e.g. Magnetic Resonance Imaging (MRI) images). In this article, we aimed at developing a model with three specific goals. Firstly, it should predict the probability (a value between 0 and 1) of disability progression for a PwMS within the next two years, instead of just a binary target (0 or 1, i.e., disease progression or no disease progression). Secondly, it should be applicable to a well-defined, relevant, and large patient population. Thirdly, all variables used in the model should be available in routine clinical care. A successful combination of these three goals would justify a clinical impact study of the model and represent a significant step towards clinical applicability. With this aim in mind, we developed and externally validated machine learning models to predict disability progression after two years for PwMS, using commonly-available clinical features. For this task we represented disability progression as a binary variable indicating if a confirmed disability progression will occur within the next two years, as defined by Kalincik et al. [13]. We trained the models using the largest longitudinal patient cohort to date for disability progression prediction in MS. The cohort was extracted from MSBase, a large international data registry containing data from multiple MS centers. We evaluated the performance, including the predicted probabilities, of different machine learning architectures and found they could achieve a ROC-AUC of 0⋅71 ± 0⋅01, an AUC-PR of 0⋅26 ± 0⋅02, a Brier score of 0⋅1 ± 0⋅01, and an expected calibration error of 0⋅07 ± 0⋅04. Importantly, and in contrast with the available literature on disease progression models for MS (except for one model to predict relapses [14]), our data pre-processing pipeline and our models check all the boxes of the Transparent Reporting for Individual Prognosis Or Diagnosis (TRIPOD) checklist. Our work therefore provides an important step towards the integration of artificial intelligence (AI) models in MS care. The outline of our approach is presented in Fig 1. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Overall layout of our approach. A: Representation of a clinical trajectory of an individual person with multiple sclerosis (PwMS). The trajectory consists of, among others, relapses, EDSS values, and treatment durations collected over time. The full list of used variables is given in the Materials and Methods. The trajectory of each patient is divided into an observation window (the available clinical history for the prediction) and the future trajectory, which is used to compute the confirmed disability progression label at two years (w c ). B: For an individual PwMS, the clinical trajectory in the observation window is extracted and used in the machine learning model to predict a well-calibrated probability of disability progression at two years. Based on the predictions, clinicians can adjust their clinical decisions accordingly. C: The MSbase dataset contains clinical data from 146 individual MS clinical centers with different clinical practice. We leveraged this feature by creating an external validation cohort of patients. We split the data per clinic, with 60% of patients used for training the model, 20% for optimizing the hyper-parameters (validation set) and 20% for external validation. The results presented in this work are all on the external validation cohort. https://doi.org/10.1371/journal.pdig.0000533.g001

Discussion The models investigated in this study provide a significant advance towards deploying AI in clinical practice in MS. After validation of the results in a clinical impact study, they have the potential to let the research in MS benefit from the advantages of advanced predictive modeling capabilities. Our work confirms that predicting the probability of disability progression of MS patients is feasible. Importantly, despite MS progression being inherently stochastic, this study shows that relevant historical clinical data, collected as part of routine clinical care, can lead to high discrimination performance and good calibration (Fig 3), which is crucial in healthcare applications. Combined with our rigorous benchmarking, external validation, and our strict adherence to the TRIPOD guidelines, this points towards a readiness of these models to be tested in a clinical impact study. Such study would evaluate the performance of these models in real-world clinical practice, compare them with the predictions of clinicians, and assess the value of such a prediction for PwMS. Over- or under-prediction of the probability of progression could indeed lead to unnecessary emotional stress or optimism. Our attained ROC-AUC scores of 0.71 are compatible with those found in the literature, which were found to range between 0.64 and 0.89 [5]. Our ROC-AUC scores are on the low end of this range. This could be explained by several factors, such as: MSBase being a large and diverse population; the use of a limited set of variables, since we constrained ourselves to variables that are collected during routine clinical care; a validation set-up where prediction is done on patients from different clinics than those in the training set. Previous work had only reported calibration graphically [20–22], with some of these models showing good calibration. The possibility to achieve well-calibrated models is empirically confirmed in our study. As previous studies used different patient populations, covariates and prediction targets, we could not directly compare our models with other models from previous studies. The models developed in this study also suffer from limitations. First of all, several countries with good quality MS registries were not included because they are not part of the MSBase initiative. Since treatment decisions can be country specific to a significant degree [23], it can result in a difference of performance of the proposed models on countries not included in this dataset. Yet, a clinical impact study in MS centers participating in MSBase would not suffer from such external validity problems. Second, our inclusion criteria required patients with good follow-up (at least one yearly visit with EDSS measurement), so stable patients that do not visit regularly, or newly diagnosed with MS cannot benefit from these models. This limits the application to patients with an already established clinical MS history. This decision was motivated by prior work [17], which showed that including clinical history as a predictor leads to more accurate prognosis, a finding that we confirm in this study. A new dedicated model would be required for disability progression in patient with shorter clinical history. Nevertheless, MS being a chronic disease, many patients would still satisfy our follow-up inclusion criteria (64% in the MSBase cohort). Third, our analysis showed that the performance of the different models varied across different patient subgroups. When segmenting the cohort by disease course or by baseline EDSS, we found that the majority subgroup (i.e. relapsing-remitting and EDSS t=0 ≤ 5⋅5) showed a better discrimination performance than subgroups with lower prevalence. We conjecture that this difference of performance is due to the lower sample size in the minority subgroups. This finding suggests a more limited value of the models for PwMS belonging to the minority subgroups. Nevertheless, the difference in calibration was not significantly different. Fourth, the progression target that we defined in this work cannot realistically fully capture the complexity of the disease and progression in MS cannot be summarized by EDSS only. EDSS itself, as an attempt to quantify progression on one-dimensional scale, lacks the expressivity to reliably encode the progression of the disease. What is more, we framed disability progression as a classification task, which is more granular than predicting future EDSS, but is more amenable for machine learning. Despite these imperfections, the confirmed disability progression label used in this work has been proven clinically useful [13], striking a good balance between abstraction and expressivity. Our work builds upon those concepts and inherits their flaws and advantages. Despite these imperfections, our models could potentially help patients in the planning of their lives and provide a baseline for further research. An emphasis on reproducibility was made, in an attempt to provide a strong benchmark for this important task. Thanks to the excellent clinically-informed pre-processing pipeline, researchers can easily extend the current models or propose their own, to continuously improve disease progression prediction. Extensions to our method could include treatment recommendation or inclusion of other biomarkers available in a specific center.

Acknowledgments The authors also wish to acknowledge the MSBase contributors for sharing the clinical data: Eva Kubala Havrdova, Charles University in Prague and General University Hospital, Prague, Czech Republic

Serkan Ozakbas, Dokuz Eylul University, Konak/Izmir, Turkey

Marco Onofrj, University G. d’Annunzio, Chieti, Italy

Raed Alroughani, Amiri Hospital, Sharq, Kuwait

Maria Pia Amato, University of Florence, Florence, Italy

Katherine Buzzard, Box Hill Hospital, Melbourne, Australia

Cavit Boz, KTU Medical Faculty Farabi Hospital, Trabzon, Turkey

Vahid Shaygannejad, Isfahan University of Medical Sciences, Isfahan, Iran

Jens Kuhle, Universitatsspital Basel, Basel, Switzerland

Bassem Yamout, American University of Beirut Medical Center, Beirut, Lebanon

Recai Turkoglu, Haydarpasa Numune Training and Research Hospital, Istanbul, Turkey

Julie Prevost, CSSS Saint-Jérôme, Saint-Jerome, Canada

Ernest Butler, Monash Medical Centre, Melbourne, Australia

Celia Oreja-Guevara, Hospital Clinico San Carlos, Madrid, Spain

Richard Macdonell, Austin Health, Melbourne, Australia

Ricardo Fernandez Bolaños, Hospital Universitario Virgen de Valme, Seville, Spain

Marie D’hooghe, Nationaal MS Centrum, Melsbroek, Belgium

Liesbeth Van Hijfte, Universitary Hospital Ghent, Ghent, Belgium

Helmut Butzkueven, The Alfred Hospital, Melbourne, Australia

Michael Barnett, Brain and Mind Centre, Sydney, Australia

Justin Garber, Westmead Hospital, Sydney, Australia

Sarah Besora, Hospital Universitari MútuaTerrassa, Barcelona, Spain

Edgardo Cristiano, Centro de Esclerosis Múltiple de Buenos Aires (CEMBA), Buenos Aires, Argentina

Magd Zakaria, Ain Shams University

Maria Laura Saladino, INEBA—Institute of Neuroscience Buenos Aires, Buenos Aires, Argentina

Shlomo Flechter, Assaf Harofeh Medical Center, Beer-Yaakov, Israel

Leontien Den braber-Moerland, Francicus Ziekenhuis, Roosendaal, Netherlands

Fraser Moore, Jewish General Hospital, Montreal, Canada

Rana Karabudak, Hacettepe University, Ankara, Turkey

Claudio Gobbi, Ospedale Civico Lugano, Lugano, Switzerland

Jennifer Massey, St Vincent’s Hospital, Sydney, Australia

Nevin Shalaby, Kasr Al Ainy MS research Unit (KAMSU), Cairo, Egypt

Jabir Alkhaboori, Royal Hospital, Muscat, Oman

Cameron Shaw, Geelong Hospital, Geelong, Australia

Jose Andres Dominguez, Hospital Universitario de la Ribera, Alzira, Spain

Jan Schepel, Waikato Hospital, Hamilton, New Zealand

Krisztina Kovacs, Péterfy Sandor Hospital, Budapest, Hungary

Pamela McCombe, Royal Brisbane and Women’s Hospital, Brisbane, Australia

Bhim Singhal, Bombay Hospital Institute of Medical Sciences, Mumbai, India

Mike Boggild, Townsville Hospital, Townsville, Australia

Imre Piroska, Veszprém Megyei Csolnoky Ferenc Kórház zrt., Veszprem, Hungary

Neil Shuey, St Vincents Hospital, Fitzroy, Melbourne, Australia

Carlos Vrech, Sanatorio Allende, Cordoba, Argentina

Tatjana Petkovska-Boskova, Clinic of Neurology Clinical Center, Skopje, Macedonia

Ilya Kister, New York University Langone Medical Center, New York, United States

Cees Zwanikken, University Hospital Nijmegen, Nijmegen, Netherlands

Jamie Campbell, Craigavon Area Hospital, Craigavon, United Kingdom

Etienne Roullet, MS Clinic, Hopital Tenon, Paris, France

Cristina Ramo-Tello, Hospital Germans Trias i Pujol, Badalona, Spain

Jose Antonio Cabrera-Gomez, Centro Internacional de Restauracion Neurologica, Havana, Cuba

Maria Edite Rio, Centro Hospitalar Universitario de Sao Joao, Porto, Portugal

Pamela McCombe, University of Queensland, Brisbane, Australia

Mark Slee, Flinders University, Adelaide, Australia

Saloua Mrabet, Razi Hospital, Manouba, Tunisia

[END]
---
[1] Url: https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0000533

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.