Automatic detection of lesion load change in Multiple Sclerosis using convolutional neural networks with segmentation confidence

Abstract

The detection of new or enlarged white-matter lesions in multiple sclerosis is a vital task in the monitoring of patients undergoing disease-modifying treatment for multiple sclerosis. However, the definition of 'new or enlarged' is not fixed, and it is known that lesion-counting is highly subjective, with high degree of inter-and intra-rater variability. Automated methods for lesion quantification, if accurate enough, hold the potential to make the detection of new and enlarged lesions consistent and repeatable. However, the majority of lesion segmentation algorithms are not evaluated for their ability to separate radiologically progressive from radiologically stable patients, despite this being a pressing clinical use-case. In this paper we demonstrate that change in volumetric measurements of lesion load alone is not a good method for performing this separation, even for highly performing segmentation methods. Instead, we propose a method for identifying lesion changes of high certainty, and establish on an internal dataset of longitudinal multiple sclerosis cases that this method is able to separate progressive from stable timepoints with a very high level of discrimination (AUC = 0.999), while changes in lesion volume are much less able to perform this separation (AUC = 0.71). Validation of the method on a second external dataset confirms that the method is able to generalize beyond the setting in which it was trained, achieving an accuracy of 83 % in separating stable and progressive timepoints.Both lesion volume and lesion count have previously been shown to be, together with clinical covariates, strong predictors of disease course across a population. However, in this paper we demonstrate that for individual patients, changes in these measures are not an adequate means of establishing no evidence of disease activity. Meanwhile, directly detecting tissue which changes, with high confidence, from non-lesion to lesion is a feasible methodology for identifying radiologically active patients.

Simultaneous lesion and brain segmentation in multiple sclerosis using deep neural networks

“…A more useful comparison for the healthy-appearing tissue labels would be between our automated tool and expert manual labels (although these, too, will suffer from substantial inter- and intra-rater variability): alternatively, as in Cerri et al 31 a comparison could be made to more robust tools, incorporating lesion filling. Finally, we only examined cross-sectional performance: in a companion study 26 , we demonstrate the applicability of the DeepSCAN classifier for detecting changes in lesion loads, making use of the segmentation uncertainty, but it would also be of merit to validate longitudinal performance of a range of classifiers.…”

“…1 . Following 26 – 28 , the network was trained using a combination of multi-class cross-entropy loss and a label-flip loss for each individual tissue class which quantifies label uncertainty. Training of the classifier was performed using the ADAM optimiser, with a cosine-annealing learning rate schedule applied over each epoch 29 .…”

Robustness of Simultaneous Lesion and Neuroanatomy Segmentation in Multiple Sclerosis Using Deep Neural Networks

“…Another limitation of the current study is our attention only to cross-sectional data. In a companion study [30], we propose to look at longitudinal imaging of the same Insel32 dataset, in order to demonstrate the viability of automated methods for reliably detecting changes in lesion loads. Improved labelling (weak 29 or human-derived) of the healthy-appearing tissue, in particular the cortex, would allow such methods to be applied also to atrophy biomarkers.…”

“…Initial attempts relied on the small numbers (30) of hand labelled brains, while later attempts have leveraged the availability of large cohorts of imaging data by training on the outputs of existing (non-learning-based) automated tools such as Freesurfer and FSL-FIRST [16,17]. Segmentation of deep-white matter structures is highly relevant in multiple sclerosis, since a recent study of 1,417 MS patients has shown links between deep grey matter volume loss and worsening condition in multiple sclerosis.…”

Lesion probability mapping in MS patients using a regression network on MR Fingerprinting

“…[33] To overcome these limitations of manual segmentation, different DL architectures and networks have been used, yielding dice coefficients ranging from 0.48 to 0.95 for WM lesion segmentation. [33][34][35][36] Therefore, in a recent publication, it was shown that this processing step can be improved by regression by also generating distance maps of the lesions. [37] This could provide more information about lesion geometry, structure, and changes similar to lesion probability mapping.…”