Contributions
Abstract: P801
Type: Poster
Abstract Category: Clinical aspects of MS - Diagnosis and differential diagnosis
Introduction: A shorter interval between onset of clinically isolated syndrome (CIS) and a second relapse (onset of clinically definite MS (CDMS)) is associated with faster disease progression and therefore is of interest. We have previously shown that machine learning classification can provide an individualised prediction of early conversion to CDMS from subjects´ baseline MRI characteristics. However, it is unknown whether the contribution to the prediction comes from regional or more global MRI measures.
Aim: We aim to identify the global and regional MRI parameters and clinical measures that predict conversion to CDMS within 1 and 3 years.
Methods: 296 CIS patients studied within 3 months from onset in three MAGNIMS centres (London, Barcelona and Siena) were included in this study. Structural MRI, white matter (WM) lesion masks, and demographic and clinical information at baseline and at 1- and 3-year follow-up (FU) were collected. 66/296 patients (22.3%) converted to CDMS at 1-year FU and 107/248 (43.1%) at 3 years. The available clinical and MRI measures were grouped as follows:
1. global measures (age, gender, EDSS, CIS type, WM lesion load and count, grey matter (GM) and WM volume, brain volume),
2. lobar measures (mean MRI intensities, lesion count and load, mean cortical thickness (CT), mean GM and WM density, region-of-interest (ROI) volume of each lobe),
3. regional measures (same as in 2. but calculated in 142 ROIs).
These groups were used as inputs to random forests, which provide a likelihood of conversion for each patient as a result.
Classification performance was measured using area under curve (AUC), which is calculated from sensitivity and specificity at varying cut-off thresholds for the likelihood.
The numbers of CIS and CDMS patients were balanced to avoid bias. This was repeated 1000 times with 10-fold cross-validation to allow for generalisation.
Results: Lobar MRI measures performed best at predicting CDMS with an AUC of 62% (range 47-76%) and p=0.005 at 1 year, and an AUC of 61% (41-75%) with p=0.023 at 3 years. They are followed by regional MRI measures with AUCs ~58% at both FUs but p-values just above 0.05. Global measures had AUCs ~54% and p>0.05.
Conclusion: Random forests can be used to predict conversion to CDMS at 1- and 3-year FU using lobar features. Very small regions or global MRI measures seem to reduce accuracy due to redundancies and noise. Future work will focus on particularly predictive brain regions.
Disclosure: This study was supported by a UCL SLMS Grand Challenge grant and MAGNIMS.
V. Wottschel, D.C. Alexander, C. Enzinger and J.L. Frederiksen have nothing to disclose.
D.T. Chard has received honoraria (paid to his employer) from Ismar Healthcare NV, Swiss MS Society, Excemed (previously Serono Symposia International Foundation), Merck, Bayer and Teva for faculty-led education work; Teva for advisory board work; meeting expenses from Merck, Teva, Novartis, the MS Trust and National MS Society; and has previously held stock in GlaxoSmithKline.
M. Filippi is Editor-in-Chief of the Journal of Neurology; serves on scientific advisory boards for Teva Pharmaceutical Industries; has received compensation for consulting services and/or speaking activities from Biogen Idec, Excemed, Novartis, and Teva Pharmaceutical Industries; and receives research support from Biogen Idec, Teva Pharmaceutical Industries, Novartis, Italian Ministry of Health, Fondazione Italiana Sclerosi Multipla, Cure PSP, Alzheimer´s Drug Discovery Foundation (ADDF), the Jacques and Gloria Gossweiler Foundation (Switzerland), and ARiSLA (Fondazione Italiana di Ricerca per la SLA).
C. Gasperini received fee as speaker for Bayer-Schering Pharma, Sanofi-Aventis, Genzyme, Biogen, Teva, Novartis, Merck Serono, and a grant for research by Teva.
A. Giorgio has received honoraria for lecturing, travel expenses for attending meetings and financial support for research from Bayer Schering, Biogen Idec, Merck Serono, Novartis and TEVA Neurosciences.
M.A. Rocca received speakers honoraria from Biogen Idec, Novartis, Genzyme, Sanofi-Aventis and Excemed and receives research support from the Italian Ministry of Health and Fondazione Italiana Sclerosi Multipla.
A. Rovira serves on scientific advisory boards for Biogen Idec, Novartis, Genzyme, and OLEA Medical, has received speaker honoraria from Bayer, Genzyme, Sanofi-Aventis, Bracco, Merck-Serono, Teva Pharmaceutical Industries Ltd, OLEA Medical, Stendhal, Novartis and Biogen Idec, and has research agreements with Siemens AG.
N. De Stefano has received honoraria from Schering, Biogen Idec, Teva Pharmaceutical Industries, Novartis, Genzyme, and Merck Serono SA for consulting services, speaking, and travel support. He serves on advisory boards for Biogen Idec, Merck Serono SA, Novartis and Genzyme.
M. Tintore has received compensation for consulting services and speaking honoraria from Bayer, Biogen, Merk-Serono, Teva, Novartis, Sanofi-Aventis, Genzyme and Roche.
D.H. Miller has received honoraria, through payments to UCL Institute of Neurology, for Advisory Committee and/or Consultancy advice in multiple sclerosis studies from Biogen Idec, GlaxoSmithKline, Novartis, Merck, Chugai, Mitsubishi Pharma Europe and Bayer Schering Pharma and has received compensation through payments to UCL Institute of Neurology for performing central MRI analysis of multiple sclerosis trials from GlaxoSmithKline, Biogen Idec, Novartis and Merck.
O. Ciccarelli is an Associate Editor of Neurology and serves as a consultant for GE Healthcare, Novartis, Roche, Biogen, Genzyme and Teva.
Abstract: P801
Type: Poster
Abstract Category: Clinical aspects of MS - Diagnosis and differential diagnosis
Introduction: A shorter interval between onset of clinically isolated syndrome (CIS) and a second relapse (onset of clinically definite MS (CDMS)) is associated with faster disease progression and therefore is of interest. We have previously shown that machine learning classification can provide an individualised prediction of early conversion to CDMS from subjects´ baseline MRI characteristics. However, it is unknown whether the contribution to the prediction comes from regional or more global MRI measures.
Aim: We aim to identify the global and regional MRI parameters and clinical measures that predict conversion to CDMS within 1 and 3 years.
Methods: 296 CIS patients studied within 3 months from onset in three MAGNIMS centres (London, Barcelona and Siena) were included in this study. Structural MRI, white matter (WM) lesion masks, and demographic and clinical information at baseline and at 1- and 3-year follow-up (FU) were collected. 66/296 patients (22.3%) converted to CDMS at 1-year FU and 107/248 (43.1%) at 3 years. The available clinical and MRI measures were grouped as follows:
1. global measures (age, gender, EDSS, CIS type, WM lesion load and count, grey matter (GM) and WM volume, brain volume),
2. lobar measures (mean MRI intensities, lesion count and load, mean cortical thickness (CT), mean GM and WM density, region-of-interest (ROI) volume of each lobe),
3. regional measures (same as in 2. but calculated in 142 ROIs).
These groups were used as inputs to random forests, which provide a likelihood of conversion for each patient as a result.
Classification performance was measured using area under curve (AUC), which is calculated from sensitivity and specificity at varying cut-off thresholds for the likelihood.
The numbers of CIS and CDMS patients were balanced to avoid bias. This was repeated 1000 times with 10-fold cross-validation to allow for generalisation.
Results: Lobar MRI measures performed best at predicting CDMS with an AUC of 62% (range 47-76%) and p=0.005 at 1 year, and an AUC of 61% (41-75%) with p=0.023 at 3 years. They are followed by regional MRI measures with AUCs ~58% at both FUs but p-values just above 0.05. Global measures had AUCs ~54% and p>0.05.
Conclusion: Random forests can be used to predict conversion to CDMS at 1- and 3-year FU using lobar features. Very small regions or global MRI measures seem to reduce accuracy due to redundancies and noise. Future work will focus on particularly predictive brain regions.
Disclosure: This study was supported by a UCL SLMS Grand Challenge grant and MAGNIMS.
V. Wottschel, D.C. Alexander, C. Enzinger and J.L. Frederiksen have nothing to disclose.
D.T. Chard has received honoraria (paid to his employer) from Ismar Healthcare NV, Swiss MS Society, Excemed (previously Serono Symposia International Foundation), Merck, Bayer and Teva for faculty-led education work; Teva for advisory board work; meeting expenses from Merck, Teva, Novartis, the MS Trust and National MS Society; and has previously held stock in GlaxoSmithKline.
M. Filippi is Editor-in-Chief of the Journal of Neurology; serves on scientific advisory boards for Teva Pharmaceutical Industries; has received compensation for consulting services and/or speaking activities from Biogen Idec, Excemed, Novartis, and Teva Pharmaceutical Industries; and receives research support from Biogen Idec, Teva Pharmaceutical Industries, Novartis, Italian Ministry of Health, Fondazione Italiana Sclerosi Multipla, Cure PSP, Alzheimer´s Drug Discovery Foundation (ADDF), the Jacques and Gloria Gossweiler Foundation (Switzerland), and ARiSLA (Fondazione Italiana di Ricerca per la SLA).
C. Gasperini received fee as speaker for Bayer-Schering Pharma, Sanofi-Aventis, Genzyme, Biogen, Teva, Novartis, Merck Serono, and a grant for research by Teva.
A. Giorgio has received honoraria for lecturing, travel expenses for attending meetings and financial support for research from Bayer Schering, Biogen Idec, Merck Serono, Novartis and TEVA Neurosciences.
M.A. Rocca received speakers honoraria from Biogen Idec, Novartis, Genzyme, Sanofi-Aventis and Excemed and receives research support from the Italian Ministry of Health and Fondazione Italiana Sclerosi Multipla.
A. Rovira serves on scientific advisory boards for Biogen Idec, Novartis, Genzyme, and OLEA Medical, has received speaker honoraria from Bayer, Genzyme, Sanofi-Aventis, Bracco, Merck-Serono, Teva Pharmaceutical Industries Ltd, OLEA Medical, Stendhal, Novartis and Biogen Idec, and has research agreements with Siemens AG.
N. De Stefano has received honoraria from Schering, Biogen Idec, Teva Pharmaceutical Industries, Novartis, Genzyme, and Merck Serono SA for consulting services, speaking, and travel support. He serves on advisory boards for Biogen Idec, Merck Serono SA, Novartis and Genzyme.
M. Tintore has received compensation for consulting services and speaking honoraria from Bayer, Biogen, Merk-Serono, Teva, Novartis, Sanofi-Aventis, Genzyme and Roche.
D.H. Miller has received honoraria, through payments to UCL Institute of Neurology, for Advisory Committee and/or Consultancy advice in multiple sclerosis studies from Biogen Idec, GlaxoSmithKline, Novartis, Merck, Chugai, Mitsubishi Pharma Europe and Bayer Schering Pharma and has received compensation through payments to UCL Institute of Neurology for performing central MRI analysis of multiple sclerosis trials from GlaxoSmithKline, Biogen Idec, Novartis and Merck.
O. Ciccarelli is an Associate Editor of Neurology and serves as a consultant for GE Healthcare, Novartis, Roche, Biogen, Genzyme and Teva.