Change detection and progression assessment in multiple sclerosis (MS) by serial magnetic resonance imaging (MRI) are important, yet challenging tasks. Analysis algorithms such as Voxel-Guided Morphometry (VGM) enable detection and quantification of even minor changes of the brain at different time points. To shorten computation times and ameliorate clinical applicability, we developed a convolutional neural network based VGM (Deep VGM) providing a fast solution for intra-individual serial volume change analysis in MS.

April 24, 2018April 10, 2018Free AccessINSPIRATION: Towards individual determination of NEDA 4 - Qualitative visual determination of short term stable brain volumes from standardized MRI acquisition in daily clinical routine of MS patients (P3.374)Achim Gass, Luc Bracoud, Johannes Gregori, Stefan Hoffmann, and Christian CornelissenAuthors Info & AffiliationsApril 10, 2018 issue90 (15_supplement) Letters to the Editor

April 24, 2018April 10, 2018Free AccessINSPIRATION: An approach to brain volume and quantitative lesion load assessments from standardized MRI acquisition in daily clinical routine of MS patients (P3.364)Achim Gass, Stefan Hoffmann, Johannes Gregori, Luc Bracoud, and Christian CornelissenAuthors Info & AffiliationsApril 10, 2018 issue90 (15_supplement) Letters to the Editor

Objective: MRI has become an integral part of MS-patient management. However, quantitative analysis of lesion-load is not trivial and has mainly been realized in clinical trials. We investigate whether additional quantitative information and visualization of lesion-load is regarded useful in the daily management of RRMS-patients. Background: INSPIRATION-MRI is a non-interventional study, conducted in Germany, to validate the feasibility and potential benefit of standardized MRI-acquisition and central-quantitative MRI-reading in clinical practice for RRMS-patients. Methods: This NIS included 253 patients, in 18 centers. Sites underwent expert training and sequence implementation employing standardized MRI-sequences. In addition to routine local diagnostic reading, a centralized quantitative MRI-data analysis is performed (volume of T2-lesions, T1-hypointense and contrast-enhancing lesions, percentage of brain volume change via boundary shift integral). The results are visualized and provided to the physicians. Results: 251 of 253 (99.2[percnt]) baseline data sets passed MRI-data quality analysis (consistency, artifacts, signal-to-noise-ratio, contrast-to-noise-ratio). 35.1[percnt] of the patients were treated with fingolimod upon study inclusion (21.5[percnt] interferons, 17.5[percnt] dimethylfumarate, 6.8 [percnt] natalizumab, 6.8[percnt] copaxone, 12.3[percnt] no/other). The mean number (±SD)/ml volume (±SD) of T2/FLAIR lesions at baseline was 30.1 (±2.8)/11033.1 (±1578.9), of black holes 4.0 (±0.9)/490.3 (±135.5) and of CM enhancing lesions 0.4 (±0.2)/31.1 (±20.3). The lesion count of T2/FLAIR lesions differed between neurologists (lesion count ≥9: 48.2[percnt]), radiologists (64.9[percnt]) and central reading (80.9[percnt]). Conclusions: More sophisticated, additional quantitative MRI-analysis is provided in a real world situation. A common standardized analysis of MRI data for several centers revealed differences between the individual estimation of lesion numbers by the centers and the quantitative approach of the central reading centers. A common analysis might improve the comparability of individual MRI scans. The quantitation of lesion load and volumes and visualization of MRI-abnormalities may facilitate MRI-data use by the responsible neurologist to support patient management. Disclosure: Prof. Gass received personal compensation from activities with Biogen Idec and Novartis. Dr. Gregori has received personal compensation for activities with Mediri as an employee. Dr. Hoffmann has received personal compensation for activities with Mediri GmbH as an employee. Dr. Fuchs has received personal compensation for activities with Novartis. Dr. Cornelissen has received personal compensation for activities with Novartis as an employee.

Magnetic resonance imaging (MRI) is the primary clinical tool to examine inflammatory brain lesions in Multiple Sclerosis (MS). Disease progression and inflammatory activities are examined by longitudinal image analysis to support diagnosis and treatment decision. Automated lesion segmentation methods based on deep convolutional neural networks (CNN) have been proposed, but are not yet applied in the clinical setting. Typical CNNs working on cross-sectional single time-point data have several limitations: changes to the image characteristics between single examinations due to scanner and protocol variations have an impact on the segmentation output, while at the same time the additional temporal correlation using pre-examinations is disregarded. In this work, we investigate approaches to overcome these limitations. Within a CNN architectural design, we propose convolutional Long Short-Term Memory (C-LSTM) networks to incorporate the temporal dimension. To reduce scanner- and protocol dependent variations between single MRI exams, we propose a histogram normalization technique as pre-processing step. The ISBI 2015 challenge data was used for network training and cross-validation. We demonstrate that the combination of the longitudinal normalization and CNN architecture increases the performance and the inter-time-point stability of the lesion segmentation. In the combined solution, the dice coefficient was increased and made more consistent for each subject. The proposed methods can therefore be used to increase the performance and stability of fully automated lesion segmentation applications in the clinical routine or in clinical trials.