Author Of 1 Presentation
PS16.03 - Use of machine learning classifiers based on structural and functional visual metrics to predict diagnosis in children with acquired demyelination.
Abstract
Background
Predicting diagnosis in youth at the first episode of demyelination is feasible in some but not all cases. Machine learning classifiers (MLC) can be trained to identify relationships between numerous multimodal input features and disease classifications to provide highly accurate predictions.
Objectives
To assess performance of machine learning classifiers for early disease diagnosis based on visual metrics in youth with demyelination.
Methods
Standardized clinical and visual data was prospectively collected at disease onset from 141 pediatric subjects with acquired demyelinating syndromes (ADS) and 75 healthy controls (HC). Participants were recruited through The Hospital for Sick Children (Toronto, Ontario (2010-2020)) and University of Calgary (2010-2017). Patients were classified using consensus definitions of demyelinating disorders and serum antibody testing for myelin oligodendrocyte glycoprotein (MOG) and aquaporin 4 (AQP4). Twenty-two auto-segmented Optical Coherence Tomography (OCT) features, 4 functional visual and 4 clinical features were used in a stratified manner alone or in combination to identify which combination of features provided the highest predictive accuracy. These input features were analyzed using 9 supervised MLC (Random Forest (RF), AdaBoost, XGBoost, Decision Tree (DT), Logistic Regression, Support Vector Machines (SVM), k-Nearest Neighbors, Stochastic Gradient Descent, Multinomial Naive Bayes). Data was split 80/20 between training and test sets. Backward feature selection was performed to re-run the analysis with best scoring predictor features in the MLC with highest predictive accuracy.
Results
AdaBoost, SVM, and DT were the best performing MLC with a test set accuracy between 82-88% in distinguishing between ADS and HC eyes. Multiple sclerosis (MS) was distinguished from HC with 92% accuracy. In descending order, fovea thickness, inferotemporal ganglion cell layer (GCL) thickness, low contrast visual acuity, outer inferior macular thickness, temporal peripapillary retinal nerve fiber layer and superior GCL thicknesses were the most important contributors to disease classification.
Conclusions
MLC can be used to combine visual metrics and clinical parameters to distinguish ADS from HC, and to predict MS. In addition to commonly used clinical metrics, we identified other structural and functional metrics that contribute importantly to classification. Among the machine learning algorithms tested, AdaBoost, SVM and DT performed best for this model.
Moderator Of 1 Session
Author Of 2 Presentations
P0691 - Autologous Hematopoietic Stem Cell Transplantation in Neuromyelitis Optica – Year 5 Update (ID 1042)
Abstract
Background
Neuromyelitis Optica Spectrum Disorder (NMO/NMOSD) is an immune astrocytopathy characterized by disabling attacks of optic nerves, spine, the area postrema, hypothalamus and other CNS regions. Although new, targeted therapies have recently been approved, they require indefinite use, with data limited essentially to positive aquaporin-4 (AQP4) patients. As well, many patients may not have access to such agents and still rely on older, harsher immunosuppressants. The immunological features of NMO/NMOSD, coupled with its severity, make it an ideal candidate for trials of autologous hematopoietic stem cell transplantation (AHSCT), with the goal of disease remission and freedom from long-term treatment. Several small studies, with a variety of transplant regimens, have been presented thus far, with mixed results.
Objectives
To determine if NMO/NMOSD patients who have failed standard immune maintenance therapy, experience a reduction in relapse and disability without need for immune therapy after AHSCT.
Methods
Starting in 2010, patients were eligible and enrolled if they met Wingerchuk 2006 criteria for NMO, aged 18-65, with => 1 relapse in 12 months or => 2 in 24 months despite immunotherapy and EDSS < 6.5. Patients underwent non-ablative stem cell mobilization and infusion using cyclophosphamide (200mg/kg - divided as 50mg/kg/day over days -5 to -2), ATG and rituximab. The primary endpoint was a 50% reduction in relapse rate at year 3 (secondary at year 5). Additional outcomes included annualized relapse rate (ARR), EDSS, MRI, AQP4 serostatus, and optical coherence tomography over 5 years. Ten subjects were to be enrolled to provide 80% power.
Results
Between 2010-2015, 3 patients were enrolled and underwent AHSCT. A 28F, AQP4-, was transplanted in 2011. Her ARR dropped from 5 to 0 at both year 3 and 5, with her EDSS dropping from 4.0 to 2.0. A 36F, AQP4+, transplanted in 2012, had a reduction in her ARR from 4 to 0.67 at year 3, and 0.5 at year 5, with her EDSS dropping from 4.5 to 3.0 throughout the trial. She was treated with MMF after her two mild post-transplant relapses. The final patient, a 39M, AQP4+ was transplanted in 2014. He had a precipitous decline with an ARR increasing from 1.3 to 2 at year 3 and 5, and an EDSS increasing from 3.5 to 7.5 at year 3, with death from NMO at year 3.5. As well, both seropositive patients remained seropositive after transplant. The trial was closed in 2016 due to challenges in recruitment.
Conclusions
While the small cohort size limits interpretation, two of three patients had a marked improvement in their NMO activity and disability after AHSCT. Regimen selection and patient features may speak to the success and failures in this trial and other published studies, but it appears that AHSCT in NMO/NMOSD is a viable option worthy of further study and refinement.
P0765 - CRION versus Atypical MOG in a 61 Year-Old Woman (ID 1035)
Abstract
Background
Chronic relapsing inflammatory optic neuropathy (CRION) is defined as recurrent optic neuritis (ON) with relatively good response to steroids. Typical cases show optic nerve enhancement on MRI, and negative aquaporin-4 (AQP4) serology is a suggested criterion. Thus CRION phenotypes not uncommonly overlap with myelin oligodendrocyte glycoprotein (MOG) and MOG-associated disease (MOGAD). Several studies have shown a relatively high rate of positive MOG antibodies in CRION patients, begging the question of how to best investigate and treat CRION patients in the age of MOGAD.
Objectives
We present an atypical case of a 61 year-old woman with CRION to illustrate the behaviour, investigations, and treatment issues in distinguishing between CRION and MOGAD.
Methods
A full outline and timeline of the patient’s history, exam features, investigations and treatment response are presented, as well as data supporting the potential link between CRION and MOGAD.
Results
A 61 year-old woman presented in fall 2019 with a second R ON, her first episode having occurred 18 years prior with full spontaneous recovery. This was followed by another ON in the same eye one month later, and two severe bilateral ON attacks in a matter of a few months, resulting in light perception acuity only, and highly suggestive of MOGAD. Between 2019 and the present, her MRI has shown unchanging right optic nerve T2 hyperintense signal and gadolinium enhancement, with no other demyelinating features. Extensive immune investigations, including screening for SLE and Sjogren’s with an ENA panel, paraneoplastic antibody testing, and both local and Mayo Clinical Laboratories testing for AQP4 and MOG were unremarkable or negative, as was screening for malignancy and sarcoidosis. After her first bilateral ON episode, she had no meaningful response to high-dose steroid therapy with a prolonged taper, but had her subsequent, severe bilateral ON attack to OU LS occurred as she weaned off steroids, and required both high dose corticosteroids and plasma exchange to achieve OD HM and OS 20/60. Maintenance steroids and the initiation of a steroid-sparing agent have prevented further ON relapses thus far.
Conclusions
This case illustrates the similarities between CRION, MOGAD, but with several atypical features. Being aware of the possibility of MOGAD in addition to other entities, and ensuring a thorough workup with consideration of preventative immune therapy in steroid-responsive patients is essential.