Abstract

Machine learning (ML) based imaging diagnostics has recently reached or even surpassed the level of clinical experts in several clinical domains. In MS, ML techniques have been developed with two main objectives: 1) detection and segmentation of brain lesions and 2) classification of clinical forms and prediction of patient disability based on conventional or/and advanced MRI methods.

In this presentation, we review the recent developments in ML approaches including deep learning (DL) techniques based on conventional and/or advanced MRI biomarkers for clinical classification and monitoring as well as disability prediction.

Classical ML techniques such as Support Vector Machine (SVM), Random-Forest (RF), k-nearest neighbor (kNN), Linear Discriminant Analysis (LDA), initially used for lesion detection and segmentation, were compared with recent DL techniques. These methods include different types of neural networks such as Convolutional Neural Networks (CNN) and graph-based neural networks (GNN) that are designed for image and graph MRI data, respectively. In one hand, conventional MRI provides multispectral images such as pre- and post-gadolinium-contrast T1-weighted (T1w), T2-weighted (T2w), proton-density-weighted (PDw) or fluid-attenuated inversion recovery (FLAIR) for lesion detection and segmentation. On the other hand, advanced MRI techniques such tensor diffusion imaging (DTI) combined with graph models provide brain structural connectivity data allowing the extraction of global and local metrics for white matter (WM) characterization.

Recent developments of CNN based on conventional MRI images out performed ML techniques and provide excellent performances for lesion detection as well as for prediction of patient disability. Best results were obtained by measuring new and enlarging lesions (volume). Furthermore, GNN based on structural connectivity data out performed previous results for disability prediction. Finally, CNN and others ML methods are developed to improve the performances of classification or prediction, and further, to better visualize and interpret the CNN decision.

Deep learning frameworks based on 3D CNN or other advanced ML techniques provide to-day excellent performances for lesion segmentation and disability prediction on conventional MRI, and even better results, on WM fiber bundles and structural connectivity data. New CNN are improved by taking in account several parameters such data characteristics, measures uncertainty and by providing information on decision.

Abstract

Machine learning (ML) based imaging diagnostics has recently reached or even surpassed the level of clinical experts in several clinical domains. In MS, ML techniques have been developed with two main objectives: 1) detection and segmentation of brain lesions and 2) classification of clinical forms and prediction of patient disability based on conventional or/and advanced MRI methods.

In this presentation, we review the recent developments in ML approaches including deep learning (DL) techniques based on conventional and/or advanced MRI biomarkers for clinical classification and monitoring as well as disability prediction.

Classical ML techniques such as Support Vector Machine (SVM), Random-Forest (RF), k-nearest neighbor (kNN), Linear Discriminant Analysis (LDA), initially used for lesion detection and segmentation, were compared with recent DL techniques. These methods include different types of neural networks such as Convolutional Neural Networks (CNN) and graph-based neural networks (GNN) that are designed for image and graph MRI data, respectively. In one hand, conventional MRI provides multispectral images such as pre- and post-gadolinium-contrast T1-weighted (T1w), T2-weighted (T2w), proton-density-weighted (PDw) or fluid-attenuated inversion recovery (FLAIR) for lesion detection and segmentation. On the other hand, advanced MRI techniques such tensor diffusion imaging (DTI) combined with graph models provide brain structural connectivity data allowing the extraction of global and local metrics for white matter (WM) characterization.

Recent developments of CNN based on conventional MRI images out performed ML techniques and provide excellent performances for lesion detection as well as for prediction of patient disability. Best results were obtained by measuring new and enlarging lesions (volume). Furthermore, GNN based on structural connectivity data out performed previous results for disability prediction. Finally, CNN and others ML methods are developed to improve the performances of classification or prediction, and further, to better visualize and interpret the CNN decision.

Deep learning frameworks based on 3D CNN or other advanced ML techniques provide to-day excellent performances for lesion segmentation and disability prediction on conventional MRI, and even better results, on WM fiber bundles and structural connectivity data. New CNN are improved by taking in account several parameters such data characteristics, measures uncertainty and by providing information on decision.

Abstract

Machine learning (ML) based imaging diagnostics has recently reached or even surpassed the level of clinical experts in several clinical domains. In MS, ML techniques have been developed with two main objectives: 1) detection and segmentation of brain lesions and 2) classification of clinical forms and prediction of patient disability based on conventional or/and advanced MRI methods.

In this presentation, we review the recent developments in ML approaches including deep learning (DL) techniques based on conventional and/or advanced MRI biomarkers for clinical classification and monitoring as well as disability prediction.

Classical ML techniques such as Support Vector Machine (SVM), Random-Forest (RF), k-nearest neighbor (kNN), Linear Discriminant Analysis (LDA), initially used for lesion detection and segmentation, were compared with recent DL techniques. These methods include different types of neural networks such as Convolutional Neural Networks (CNN) and graph-based neural networks (GNN) that are designed for image and graph MRI data, respectively. In one hand, conventional MRI provides multispectral images such as pre- and post-gadolinium-contrast T1-weighted (T1w), T2-weighted (T2w), proton-density-weighted (PDw) or fluid-attenuated inversion recovery (FLAIR) for lesion detection and segmentation. On the other hand, advanced MRI techniques such tensor diffusion imaging (DTI) combined with graph models provide brain structural connectivity data allowing the extraction of global and local metrics for white matter (WM) characterization.

Recent developments of CNN based on conventional MRI images out performed ML techniques and provide excellent performances for lesion detection as well as for prediction of patient disability. Best results were obtained by measuring new and enlarging lesions (volume). Furthermore, GNN based on structural connectivity data out performed previous results for disability prediction. Finally, CNN and others ML methods are developed to improve the performances of classification or prediction, and further, to better visualize and interpret the CNN decision.

Deep learning frameworks based on 3D CNN or other advanced ML techniques provide to-day excellent performances for lesion segmentation and disability prediction on conventional MRI, and even better results, on WM fiber bundles and structural connectivity data. New CNN are improved by taking in account several parameters such data characteristics, measures uncertainty and by providing information on decision.

Background

Treatment response in multiple sclerosis (MS) is frequently suboptimal and, in many cases, different disease-modifying therapies need to be tested. The expanded disability status scale (EDSS) score and other markers of disease activity (such as relapses, new lesions, brain atrophy, etc.) are crucial for treatment decision making. However, EDSS suffers from poor reliability, repeatability, and high inter-rater variability. Therefore, automatic and objective disability scoring using MRI information could help to monitor disease progression reliably and optimize treatment.

Objectives

Develop a machine-learning model that learns relations between MRI-based brain volumes and clinical disability measured by EDSS.

Methods

Multi-center FLAIR and T1 MRI data from 325 MS patients were used. Individuals were rated in each center using EDSS within 0-89 days before or after the MRI scan. Automated image analysis was performed using icobrain, providing volumetric quantification of gray matter, white matter, whole brain, lateral ventricles, T1 hypointense and FLAIR hyperintense lesions. Moreover, other features, such as age, sex, and center, were available. A machine-learning model based on random forest regression was built for estimating EDSS automatically from these features. The model’s performance was assessed by means of mean squared error (MSE) and mean absolute error (MAE) evaluated overall, and on two EDSS subgroups, <=4 and >4,in a 100 repetition 10-fold-cross-validation fashion. Subsequently, the percentage of cases for which the automatic EDSS was within 1.5, 1 or 0.5 points, respectively, from the clinically reported EDSS was computed.

Results

The proposed automatic EDSS estimation model obtained MSE=2.36±0.03, MAE=1.24±0.89 for the overall interval, with MSE=1.83±0.04, MAE=1.11±0.76 for EDSS<=4 (N=200) and MSE=3.26±0.06, MAE=1.46±1.05 for EDSS>4 (N=118). The percentage of cases with absolute error strictly below 1.5, 1 and 0.5 EDSS points was 67%, 46% and 24%, respectively.

Conclusions

A good match between the automatic EDSS and the measured EDSS is only possible up to a certain extent, suggesting that the MRI-based EDSS score might also capture complementary information on disease activity compared to the clinically measured EDSS. Understanding such differences is a prerequisite for predicting future disability progression in MS.