Exploration of machine learning techniques in predicting multiple sclerosis disease course

To explore the value of machine learning methods for predicting multiple sclerosis disease course. 1693 CLIMB study patients were classified as increased EDSS≥1.5 (worsening) or not (non-worsening) at up to five years after baseline visit. Support vector machines (SVM) were used to build the classif...

Full description

Saved in:
Bibliographic Details
Published in:PloS one Vol. 12; no. 4; p. e0174866
Main Authors: Zhao, Yijun, Healy, Brian C., Rotstein, Dalia, Guttmann, Charles R. G., Bakshi, Rohit, Weiner, Howard L., Brodley, Carla E., Chitnis, Tanuja
Format: Journal Article
Language:English
Published: United States Public Library of Science 05.04.2017
Public Library of Science (PLoS)
Subjects:
Adult
Analysis
Artificial intelligence
Autoimmune diseases
Biology and Life Sciences
Brain
Brain - diagnostic imaging
Brain - pathology
Classification
Computer and Information Sciences
Diagnosis
Disease Progression
Female
Genetics
Humans
Learning
Learning algorithms
Logistic Models
Machine Learning
Magnetic Resonance Imaging
Male
Medicine and Health Sciences
Multiple sclerosis
Multiple Sclerosis - pathology
Neuroimaging
Neurology
Patients
People and Places
Physical Sciences
Predictions
Research and Analysis Methods
Sensitivity
Sensitivity and Specificity
Short term
Statistical analysis
Support Vector Machine
Support vector machines
United States > US
Massachusetts
ISSN:1932-6203, 1932-6203
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To explore the value of machine learning methods for predicting multiple sclerosis disease course. 1693 CLIMB study patients were classified as increased EDSS≥1.5 (worsening) or not (non-worsening) at up to five years after baseline visit. Support vector machines (SVM) were used to build the classifier, and compared to logistic regression (LR) using demographic, clinical and MRI data obtained at years one and two to predict EDSS at five years follow-up. Baseline data alone provided little predictive value. Clinical observation for one year improved overall SVM sensitivity to 62% and specificity to 65% in predicting worsening cases. The addition of one year MRI data improved sensitivity to 71% and specificity to 68%. Use of non-uniform misclassification costs in the SVM model, weighting towards increased sensitivity, improved predictions (up to 86%). Sensitivity, specificity, and overall accuracy improved minimally with additional follow-up data. Predictions improved within specific groups defined by baseline EDSS. LR performed more poorly than SVM in most cases. Race, family history of MS, and brain parenchymal fraction, ranked highly as predictors of the non-worsening group. Brain T2 lesion volume ranked highly as predictive of the worsening group. SVM incorporating short-term clinical and brain MRI data, class imbalance corrective measures, and classification costs may be a promising means to predict MS disease course, and for selection of patients suitable for more aggressive treatment regimens.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
Competing Interests: Yijun Zhao has no disclosures. Brian Healy has received grant support from Merck-Serono and Novartis. Dalia Rotstein has no disclosures. Charles Guttmann has no disclosures. Rohit Bakshi has received consulting fees and/or research support from Abbvie, Alkermes, Biogen Idec, Genzyme/Sanofi, Novartis, Teva, and Questcor in the past 12 months. Howard Weiner has served as a consultant for Biogen-Idec, Nasvax, Novartis, Merck Serono, and Teva Neurosciences, and has received grant support from Merck Serono. Dr. Brodley has no disclosures. Dr. Chitnis has served as an advisor for Biogen-Idec, Novartis, Sanofi-Aventis, Teva Neurosciences, and has received grant support from National MS Society, NIH, Guthy-Jackson Charitable Foundation, Merck-Serono and Novartis. The disclosures listed for authors do not alter our adherence to PLOS ONE policies on sharing data and materials.
Conceptualization: YZ BCH CEB TC.Data curation: BCH TC.Formal analysis: YZ BCH CEB TC.Funding acquisition: HLW CEB TC.Investigation: YZ BCH CEB TC.Methodology: YZ BCH CEB TC.Project administration: TC.Resources: CRGG RB HLW CEB TC.Software: YZ BCH.Supervision: CEB TC.Validation: DR CRGG RB HLW TC.Visualization: YZ.Writing – original draft: YZ TC.Writing – review & editing: BCH DR CRGG RB HLW CEB.
ISSN:1932-6203
1932-6203
DOI:10.1371/journal.pone.0174866