Improving automated multiple sclerosis lesion segmentation with a cascaded 3D convolutional neural network approach

ABSTRACTBackground and PurposeAssessment of brain lesions on magnetic resonance imaging (MRI) is crucial for research in multiple sclerosis (MS). Manual segmentation is time‐consuming and inconsistent. We aimed to develop an automated MS lesion segmentation algorithm for T2‐weighted fluid‐attenuated inversion recovery (FLAIR) MRI.MethodsWe developed FLAIR Lesion Analysis in Multiple Sclerosis (FLAMeS), a deep learning‐based MS lesion segmentation algorithm based on the nnU‐Net 3D full‐resolution U‐Net and trained on 668 FLAIR 1.5 and 3 tesla scans from persons with MS. FLAMeS was evaluated on three external datasets: MSSEG‐2 (n = 14), MSLesSeg (n = 51), and a clinical cohort (n = 10), and compared to SAMSEG, LST‐LPA, and LST‐AI. Performance was assessed qualitatively by two blinded experts and quantitatively by comparing automated and ground truth lesion masks using standard segmentation metrics.ResultsIn a blinded qualitative review of 20 scans, both raters selected FLAMeS as the most accurate segmentation in 15 cases, with one rater favoring FLAMeS in two additional cases. Across all testing datasets, FLAMeS achieved a mean Dice score of 0.74, a true positive rate of 0.84, and an F1 score of 0.78, consistently outperforming the benchmark methods. For other metrics, including positive predictive value, relative volume difference, and false positive rate, FLAMeS performed similarly to or better than benchmark methods. Most lesions missed by FLAMeS were smaller than 10 mm3, whereas the benchmark methods missed larger lesions in addition to smaller ones.ConclusionsFLAMeS is an accurate, robust method for MS lesion segmentation that outperforms other publicly available methods.

We present the application of limited one-time sampling irregularity map (LOTS-IM): a fully automatic unsupervised approach to extract brain tissue irregularities in magnetic resonance images (MRI), for quantitatively assessing white matter hyperintensities (WMH) of presumed vascular origin, and multiple sclerosis (MS) lesions and their progression. LOTS-IM generates an irregularity map (IM) that represents all voxels as irregularity values with respect to the ones considered ”normal”. Unlike probability values, IM represents both regular and irregular regions in the brain based on the original MRI's texture information. We evaluated and compared the use of IM for WMH and MS lesions segmentation on T2-FLAIR MRI with the state-of-the-art unsupervised lesions’ segmentation method, Lesion Growth Algorithm from the public toolbox Lesion Segmentation Toolbox (LST-LGA), with several well established conventional supervised machine learning schemes and with state-of-the-art supervised deep learning methods for WMH segmentation. In our experiments, LOTS-IM outperformed unsupervised method LST-LGA on WMH segmentation, both in performance and processing speed, thanks to the limited one-time sampling scheme and its implementation on GPU. Our method also outperformed supervised conventional machine learning algorithms (i.e., support vector machine (SVM) and random forest (RF)) and deep learning algorithms (i.e., deep Boltzmann machine (DBM) and convolutional encoder network (CEN)), while yielding comparable results to the convolutional neural network schemes that rank top of the algorithms developed up to date for this purpose (i.e., UResNet and UNet). LOTS-IM also performed well on MS lesions segmentation, performing similar to LST-LGA. On the other hand, the high sensitivity of IM on depicting signal change deems suitable for assessing MS progression, although care must be taken with signal changes not reflective of a true pathology.

The diagnosis of multiple sclerosis (MS) is based on accurate detection of lesions on magnetic resonance imaging (MRI) which also provides ongoing essential information about the progression and status of the disease. Manual detection of lesions is very time consuming and lacks accuracy. Most of the lesions are difficult to detect manually, especially within the grey matter. This paper proposes a novel and fully automated convolution neural network (CNN) approach to segment lesions. The proposed system consists of two 2D patchwise CNNs which can segment lesions more accurately and robustly. The first CNN network is implemented to segment lesions accurately, and the second network aims to reduce the false positives to increase efficiency. The system consists of two parallel convolutional pathways, where one pathway is concatenated to the second and at the end, the fully connected layer is replaced with CNN. Three routine MRI sequences T1-w, T2-w and FLAIR are used as input to the CNN, where FLAIR is used for segmentation because most lesions on MRI appear as bright regions and T1-w & T2-w are used to reduce MRI artifacts. We evaluated the proposed system on two challenge datasets that are publicly available from MICCAI and ISBI. Quantitative and qualitative evaluation has been performed with various metrics like false positive rate (FPR), true positive rate (TPR) and dice similarities, and were compared to current state-of-the-art methods. The proposed method shows consistent higher precision and sensitivity than other methods. The proposed method can accurately and robustly segment MS lesions from images produced by different MRI scanners, with a precision up to 90%.

Background and Objectives:Multiple sclerosis (MS) lesion segmentation is important in estimating the progress of the disease and measuring the impact of new clinical treatments. Manual lesion delineation for the segmentation of lesions is time-consuming and suffers from observer variability. Therefore, a fully automated MS lesion segmentation method is considerable important in clinical practice.Subjects and Methods:In this study, we present a multilabel fusion embedded level set method for white matter lesion segmentation from MS patient images. Specifically, we focus on the validation of the variational level set method. Lesion segmentation is achieved by extending the level set contour which consists of an intensity-constrained term, an image data term, and a regularization term.Results:To compare the performance of our method with other state-of-the-art methods, we evaluated the methods with 25 magnetic resonance imaging datasets of MS patients. The dice score reaches an average of 0.55 for the proposed method. The sensitivity value and specificity value reach an average of 0.89 and 0.14, respectively.Conclusions:Experimental results demonstrate that our method is robust to parameter setting and outperforms other methods. The intensity-constrained term plays a key role in improving the segmentation accuracy. The experimental results show that our approach is effective and robust for lesion segmentation, which might simplify the quantification of lesions in basic research and even clinical trials.

Segmentation of multiple sclerosis lesions in brain MRI: A review of automated approaches

The Segmentation is a fundamental technique used in image processing to extract suspicious regions from a given image. Quantitative assessment of Magnetic Resonance Imaging (MRI) lesion load of patients with multiple sclerosis (MS) is the most objective approach for a better understanding of the history of the pathology, either natural or modified by therapies. Recently, numerous methods have been introduced for the segmentation of MS lesions in MR images. In this work, we propose to segment the images based on a constrained Gaussian Mixture Model (GMM) and using Genetic Algorithm to determine the optimal parameters of this highly non-linear model. We use this method for automatic segmentation of MS lesions in fluid attenuated inversion recovery (FLAIR) of magnetic resonance images. Moreover, a preprocessing step is also performed to suppress artifacts and to remove the unwanted skull portions from brain MRI. The proposed method has been tested on real patients' MR images. The experimental results show that the proposed method can effectively segment MS lesions in the brain MR images.

MRI FLAIR lesion segmentation in multiple sclerosis: Does automated segmentation hold up with manual annotation?

Magnetic Resonance Imaging (MRI), allowing in-vivo detection of lesions, is today a crucial tool for diagnosis of Multiple Sclerosis (MS). Although the detection of lesions are not sufficient for a diagnosis of MS because of similarity with patterns detected in other neurological diseases, taking into account different radiological informations, MRI findings can often yield a high degree of confidence. We used a snake based procedure for segmentation of lesion then proposing a method based on Cluster Analysis to support clinicians in the diagnosis of MS. By identifying a minimum set of significant descriptors, our algorithm can help neurologist and neuroimaging expert to distinguish MS plaques from other kinds of lesions.

In this work we present BIANCA‐MS, a novel tool for brain white matter lesion segmentation in multiple sclerosis (MS), able to generalize across both the wide spectrum of MRI acquisition protocols and the heterogeneity of manually labeled data. BIANCA‐MS is based on the original version of BIANCA and implements two innovative elements: a harmonized setting, tested under different MRI protocols, which avoids the need to further tune algorithm parameters to each dataset; and a cleaning step developed to improve consistency in automated and manual segmentations, thus reducing unwanted variability in output segmentations and validation data. BIANCA‐MS was tested on three datasets, acquired with different MRI protocols. First, we compared BIANCA‐MS to other widely used tools. Second, we tested how BIANCA‐MS performs in separate datasets. Finally, we evaluated BIANCA‐MS performance on a pooled dataset where all MRI data were merged. We calculated the overlap using the DICE spatial similarity index (SI) as well as the number of false positive/negative clusters (nFPC/nFNC) in comparison to the manual masks processed with the cleaning step. BIANCA‐MS clearly outperformed other available tools in both high‐ and low‐resolution images and provided comparable performance across different scanning protocols, sets of modalities and image resolutions. BIANCA‐MS performance on the pooled dataset (SI: 0.72 ± 0.25, nFPC: 13 ± 11, nFNC: 4 ± 8) were comparable to those achieved on each individual dataset (median across datasets SI: 0.72 ± 0.28, nFPC: 14 ± 11, nFNC: 4 ± 8). Our findings suggest that BIANCA‐MS is a robust and accurate approach for automated MS lesion segmentation.

Multiple sclerosis lesions segmentation is an important step in the diagnosis and tracking in the evolution of the disease. Convolutional Neural Networks (CNN) have been obtaining successful results in the task of lesion segmentation in recent years, but still present problem segmenting boundaries of the lesions. In this work we focus the learning process on hard voxels close to the boundaries of the lesions by means of a stratified sampling and the use of focal loss function that dynamically increase the penalization on this kind of voxels. This approach was applied on the 2015 Longitudinal MS Lesion Segmentation Challenge dataset (ISBI2015 (https://smart-stats-tools.org/lesion-challenge)), obtaining better results than approaches using binary cross entropy loss and focal loss functions with uniform sampling.

Background: Multiple Sclerosis (MS) diagnosis and evaluation is often a challenging task due to its growing need for multimodal MRI acquisition protocol. Recently, the scientific community offers several computational alternatives to the time-consuming and subjective task of manual MS lesion segmentation. Although there is an increasing number of MS lesion segmentation methods, a controlled and realistic simulation environment can benefit the community for a reliable evaluation procedure. Methods: This study proposes an automatic parametric MS lesion simulation framework (MS-MIST) with the objective to emulate real MS-like pattern on MRI data of healthy individuals. The voxel gray-level patterns, spatial location, and shapes extracted from MS patient allow consistent simulation features. We used both visual evaluation from an expert radiologist in the field of MS diagnosis and SPM Lesion Segmentation Tool (LST) for qualitative and quantitative simulation quality, respectively. Results: Our results show that both the agreement between the automatic segmentation with the simulated lesions (Pearson’s correlation R = 0.977) and the segmentation quality scores, i.e., sensitivity (mean = 0.9050), specificity (mean = 0.9992), dice similarity (mean = 0.8972), and accuracy (mean = 0.9984), are consistent between MS-MIST and real clinical settings. Conclusions: MS-MIST proposes a practical solution to common issues discussed in the literature, such as inconsistency with real lesions geometry, imprecise lesion spatial and signal variability, lack of multiple MRI modalities and restrictions to simulate different lesion loads. It is worth noting that the simulation platform is freely available as an open-source code to the community.

LST-AI: A deep learning ensemble for accurate MS lesion segmentation

Multiple sclerosis (MS) one of the chronic Idiopathic disease needs a precise segmentation method to analyze the MS White Matter Lesion (WML) variation with respect the Extended Disability scale (EDSS) value of MS MR brain images. In this work, we use the Active contour segmentation method to segment the MS brain tissue from T2W axial view MS preprocessed MR images. The Active contour segmentation to segment MS affected area using the ITK-SNAP 3D segmentation tool. The segmented MS WML lesion image region to measure the MS lesion affected area volume and calculate the image shape and statistical feature value using Gray Level Co-occurrence Matrix (GLCM). The MS WML volume, shape, and statistical feature are correlated with the different EDSS value of the MS patient data set. The shape, statistical feature parameter, and lesion volume correlation study are very useful to monitoring changes associated with improvement of disability stage from 1–9 category of MS progression. Reproducibility and sensitivity must be enhanced to use the analysis for consecutive monitoring in individuals.

Automatic segmentation of different-sized white matter lesions by voxel probability estimation

When using convolutional neural networks (CNNs) for segmentation of organs and lesions in medical images, the conventional approach is to work with inputs and outputs either as single slice [two-dimensional (2D)] or whole volumes [three-dimensional (3D)]. One common alternative, in this study denoted as pseudo-3D, is to use a stack of adjacent slices as input and produce a prediction for at least the central slice. This approach gives the network the possibility to capture 3D spatial information, with only a minor additional computational cost. In this study, we systematically evaluate the segmentation performance and computational costs of this pseudo-3D approach as a function of the number of input slices, and compare the results to conventional end-to-end 2D and 3D CNNs, and to triplanar orthogonal 2D CNNs. The standard pseudo-3D method regards the neighboring slices as multiple input image channels. We additionally design and evaluate a novel, simple approach where the input stack is a volumetric input that is repeatably convolved in 3D to obtain a 2D feature map. This 2D map is in turn fed into a standard 2D network. We conducted experiments using two different CNN backbone architectures and on eight diverse data sets covering different anatomical regions, imaging modalities, and segmentation tasks. We found that while both pseudo-3D methods can process a large number of slices at once and still be computationally much more efficient than fully 3D CNNs, a significant improvement over a regular 2D CNN was only observed with two of the eight data sets. triplanar networks had the poorest performance of all the evaluated models. An analysis of the structural properties of the segmentation masks revealed no relations to the segmentation performance with respect to the number of input slices. A post hoc rank sum test which combined all metrics and data sets yielded that only our newly proposed pseudo-3D method with an input size of 13 slices outperformed almost all methods. In the general case, multislice inputs appear not to improve segmentation results over using 2D or 3D CNNs. For the particular case of 13 input slices, the proposed novel pseudo-3D method does appear to have a slight advantage across all data sets compared to all other methods evaluated in this work.