Diagnosing Optic Neuritis in Neuromyelitis Optica Spectrum Disorders (NMOSD) Using 6 Machine Learning Models with MRI

Neuromyelitis Optica Spectrum Disorders (NMOSD) is an inflammatory disease in the human central nervous system that causes severe optic neuritis (ON). ON is not necessarily present in all NMOSD patients. Thus far, no study has distinguished NMOSD patients with and without ON. Therefore, this study aims to diagnose ON in NMOSD patients using 6 prominent machine learning (ML) models: decision trees, linear discriminant analysis, logistic regression, support vector machines (SVM), Naïve Bayes, and k-nearest-neighbor. This study measured three non-texture (quantitative) features: area, volume, and signal intensity; and five texture (qualitative) features: energy, entropy, homogeneity, contrast, and correlation, of the optic nerves on MR images. This is the first study that used the texture features of the optic nerve for the diagnosis of ON in NMOSD patients. All these features and the age of patients are used for training and testing the ML models. There is a significant difference ( p