Machine learning-based imaging diagnostics has recently reached or even superseded the level of clinical experts in several clinical domains. However, classification decisions of a trained machine learning system are typically non-transparent, a major hindrance for clinical integration, error tracking or knowledge discovery. In this study, we present a transparent deep learning framework relying on convolutional neural networks (CNNs) and layer-wise relevance propagation (LRP) for diagnosing multiple sclerosis (MS), the most widespread autoimmune neuroinflammatory disease. MS is commonly diagnosed utilizing a combination of clinical presentation and conventional magnetic resonance imaging (MRI), specifically the occurrence and presentation of white matter lesions in T2-weighted images. We hypothesized that using LRP in a naive predictive model would enable us to uncover relevant image features that a trained CNN uses for decision-making. Since imaging markers in MS are well-established this would enable us to validate the respective CNN model. First, we pre-trained a CNN on MRI data from the Alzheimer's Disease Neuroimaging Initiative (n = 921), afterwards specializing the CNN to discriminate between MS patients and healthy controls (n = 147). Using LRP, we then produced a heatmap for each subject in the holdout set depicting the voxel-wise relevance for a particular classification decision. The resulting CNN model resulted in a balanced accuracy of 87.04% and an area under the curve of 96.08% in a receiver operating characteristic curve. The subsequent LRP visualization revealed that the CNN model focuses indeed on individual lesions, but also incorporates additional information such as lesion location, non-lesional white matter or gray matter areas such as the thalamus, which are established conventional and advanced MRI markers in MS. We conclude that LRP and the proposed framework have the capability to make diagnostic decisions of CNN models transparent, which could serve to justify classification decisions for clinical review, verify diagnosis-relevant features and potentially gather new disease knowledge. ability in patients [2]. The current quasi-standard for diagnosing MS, the McDonald criteria, relies on clinical presentation and the presence of lesions visible in conventional T2-weighted brain magnetic resonance imaging (MRI) data [3]. Most common are fluid-suppressed T2-weighted image sequences, which are sensitive towards MS-relevant white matter lesions, but also relatively unspecific with respect to underlying disease processes [4]. Several other imaging markers have been described including neurodegeneration, thalamic atrophy, cortical lesions, altered functional connectivity or central vein signs

IMPORTANCE Clinically isolated syndrome (CIS) describes a first clinical incident suggestive of multiple sclerosis (MS). Identifying patients with CIS who have a high risk of future disease activity and subsequent MS diagnosis is crucial for patient monitoring and the initiation of disease-modifying therapy. OBJECTIVE To investigate the association of retinal optical coherence tomography (OCT) results with future disease activity in patients with CIS. DESIGN, SETTING, AND PARTICIPANTS This prospective, longitudinal cohort study took place between January 2011 and May 2017 at 2 German tertiary referral centers. A total of 179 patients with CIS were screened (80 in Berlin and 99 in Munich). Patients underwent neurological examination, magnetic resonance imaging (MRI), and OCT. Only eyes with no previous optic neuritis were considered for OCT analysis. MAIN OUTCOMES AND MEASURES The primary outcome was not meeting the no evidence of disease activity (NEDA-3) criteria; secondary outcomes were MS diagnosis (by the 2010 McDonald criteria) and worsening of disability. The primary measure was OCT-derived ganglion cell and inner plexiform layer thickness; the secondary measures included peripapillary retinal nerve fiber layer thickness, inner nuclear layer thickness, and MRI-derived T2-weighted lesions. RESULTS A total of 97 of the 179 screened patients (54.2%) were enrolled in the study at a median of 93 (interquartile range [IQR], 62-161) days after a first demyelinating event. The median follow-up duration (Kaplan-Meier survival time) was 729 (IQR, 664-903) days. Of 97 patients with CIS (mean age 33.6 [7.9] years; 61 [62.9%] female), 58 (59%) did not meet NEDA-3 criteria during the follow-up period. A Kaplan-Meier analysis showed a significant probability difference in not meeting NEDA-3 criteria by ganglion cell and inner plexiform later thickness (thinnest vs thickest tertile: hazard ratio [HR], 3.33 [95% CI, 1.70-6.55; P < .001; log-rank P = .001). A follow-up diagnosis of MS was more likely for patients with low ganglion cell and inner plexiform layer thickness (thinnest vs thickest tertile: HR, 4.05 [95% CI, 1.93-8.50]; P < .001). Low peripapillary retinal nerve fiber layer thickness likewise indicated risk of not meeting NEDA-3 criteria (thinnest vs thickest tertile: HR, 2.46 [95% CI, 1.29-4.66]; P = .01; log-rank P = .02). Inner nuclear layer thickness and T2-weighted lesion count were not associated with not meeting NEDA-3 criteria. CONCLUSIONS AND RELEVANCE Retinal ganglion cell and inner plexiform layer thickness might prove a valuable imaging marker for anticipating future disease activity and diagnosis of MS in patients with CIS, which can potentially support patient monitoring and initiation of disease-modifying therapy.