Automated Detection and Characterization of Multiple Sclerosis Lesions in Brain MR Images

doi:10.1016/S0730-725X(97)00300-7

Magnetic Resonance Imaging

Volume 16, Issue 3, April 1998, Pages 311-318

https://doi.org/10.1016/S0730-725X(97)00300-7 Get rights and content

Abstract

In the present study an automatic algorithm for detection and contouring of multiple sclerosis (MS) lesions in brain magnetic resonance (MR) images is introduced. This algorithm automatically detects MS lesions in axial proton density, T₂-weighted, gadolinium enhanced, and fast fluid attenuated inversion recovery (FLAIR) brain MR images. Automated detection consists of three main stages: (1) detection and contouring of all hyperintense signal regions within the image; (2) partial elimination of false positive segments (defined herein as artifacts) by size, shape index, and anatomical location; (3) the use of an artificial neural paradigm (Back-Propagation) for final removal of artifacts by differentiating them from true MS lesions. The algorithm was applied to 45 images acquired from 14 MS patients. The algorithm’s sensitivity was 0.87 and the specificity 0.96. In 34 images, 100% of the lesions were detected. The algorithm potentially may serve as a useful preprocessing tool for quantitative MS monitoring via magnetic resonance imaging.

Introduction

Multiple sclerosis (MS) is the most common demyelinating disease of the central nervous system white matter[1]and is one of the major causes of disability in young adults. Magnetic resonance imaging (MRI) is the most suitable technique for evaluation of demyelinating lesions within the brain and spinal cord. MRI provides a good contrast between MS lesions and normal white matter as MS lesions produce hyperintense signals in both proton density (PD) and T₂-weighted images. The addition of contrast-enhancement with gadolinium DPTA (Gd) injection, or a fast fluid attenuated inversion recovery (FLAIR) (FF), further improves the sensitivity of MRI examinations.2, 3, 4, 5

Careful detection and segmentation of MS lesions in MR images are needed for accurate evaluation of the disease burden. Quantitative characterization of the detected MS lesions is essential for assessment of disease progression and treatment efficacy. Several studies have used different measuring methods and varying definitions for assessing disease burden, such as total lesion area and volume, corpus callosum’s area and the number of enhancing lesions in various time scales.6, 7, 8, 9, 10In addition, qualitative or semi-quantitative methods have also been applied. For example, Kappos et al.,[10]have analyzed factors such as the occurrence of new lesions, disappearance of lesions, and changes in lesion intensities and/or size. Other studies have used computerized database systems to evaluate the number, size, and location of areas of increased signals.11, 12

Most of the recent studies have used semiautomatic algorithms that employ segmentation techniques such as thresholding strategies, region growing, and interslice connectivity criteria.13, 14, 15Several algorithms that have a minimal interaction with the operator have also been developed. For example, Pannizzo et al.,[16]first used a supervised automatic procedure to discriminate between the brain tissue and the fat and skull bone, and then an algorithm based on a histogram analysis that detected MS lesions. Miki et al.[17]have developed a computer assisted program which is based on “fuzzy connectedness” which automatically marked potential lesion sites. The user accepted or rejected these lesions by means of yes/no response to the program query. This group has also reported that 0% inter- and intraobserver variability were obtained with this procedure, with no false positive and very low false negative results.[18]

Using a different strategy, information from two image types was used to mark lesions. Kapouleas,[19]by using both PD and T₂-weighted images of each slice, first located the brain in each slice (with PD), and then located anatomical landmarks (such as the interhemispherical fissure) and suspected lesions (with T₂) within the brain. Then using a geometric model and anatomical landmarks, he defined the regions where MS lesions are unlikely to appear and eliminated false positive segments. Bedell et al.[20]have combined fast spin echo data with information from flow images to minimize false lesion classifications. A general review presenting conclusions and recommendations regarding various processing strategies for MRI detection of MS was recently presented by Evans et al.[21]

In this study, an algorithm that automatically detects and contours MS lesions in axial MR brain images is introduced. The algorithm utilizes regional signal intensity, lesions’ shape, anatomical location and an artificial neural network for removing artifacts (i.e., non MS regions marked as MS lesions).

Section snippets

Basic Assumptions

To develop the detection algorithm, the following assumptions were made:

a) In PD, T₂-weighted, gadolinium enhanced, and FF-MR images, MS lesions appear much brighter than the rest of the brain.

b) Non-MS regions in the brain, which also produce high signal intensity, (especially in T₂-weighted MR images) such as blood vessels, and cerebrospinal fluid within the ventricles, have either a relatively very small or very large (in the case of the ventricles) area (see quantitative definitions in the

Results

The developed algorithm was tested by processing 45 images acquired from 14 different MS patients. Only 10 images out of the 45 were included in the set which was used for training the ANN. The selected images were acquired, using one of the following scanners: Elscint (Gyrex), Philips (Gyroscan) or Seimens (Magnetom). PD, T₂-weighted, Gd enhanced and FF images were processed. The analyzed images were selected from a larger data set of arbitrarily selected images. Brain stem images, and images

Discussion

An automatic algorithm for the detection and characterization of MS lesions on MR images is a highly desirable tool for assessment of disease progression and treatment efficacy. A computerized algorithm is objective, more consistent and potentially faster than manual tracing. Several approaches for automatization of MS lesion detection and tracing have been suggested.10, 14, 15, 16, 17, 18, 19These studies have demonstrated that automatization can yield good results of lesion detection and may

Acknowledgements

We thank TEVA-Pharmaceutical Industries LTD. for their financial support. The “Sociedad Venezolana Amigos del Technion” (D.G.Z.), the Technion V.P.R. Fund 130-320- Montreal Biomedical Research Fund and the Irving and Adela Rosenberg Foundation Inc. (H.A.) are also gratefully acknowledged.

References (26)

I. Kapouleas
Automatic detection of white matter lesions in magnetic resonance brain images
Computer Methods and Programs in Biomedicine
(1990)
M. Kamber et al.
Knowledge-based interpretation of magnetic resonance imagesDetecting multiple sclerosis lesions
Artificial Intelligence in Medicine
(1993)
L. Honig et al.
Magnetic resonance imaging of spinal cord lesions in multiple sclerosis
J. Neurol. Neurosurg. Psychiatry
(1989)
J.O. Harris et al.
Serial gadolinium-enhanced magnetic resonance imaging scans in patients with early, relapsing-remitting multiple sclerosisimplications for clinical trials and natural history
Ann. Neurol.
(1991)
D.H. Miller
Magnetic resonance in monitoring the treatment of multiple sclerosis
Ann. Neurol.
(1994)
Fillipi, M.; Rovaris, M.; Baratti, C.; Mammi, S.; Horsefield, M.A.; Yousry, T.; Comi, G. Intra- and inter-observer...
D.H. Miller et al.
Magnetic resonance imaging in monitoring the treatment of multiple sclerosisConcerted action guidelines
J. Neurol. Neurosurg. Psychiatry
(1991)
S.M. Rao et al.
Correlation of magnetic resonance imaging with neuropsychological testing in multiple sclerosis
Neurology
(1989)
T. Swirsky-Sacchetti et al.
Neuropsychological and structural brain lesions in multiple sclerosisA regional analysis
Neurology
(1992)
A.D. Gean-Marton et al.
Abnormal corpus callosumA sensitive and specific indicator of multiple sclerosis
Radiology
(1991)

L. Kappos et al.

Magnetic resonance imaging in the evaluation of treatment in multiple sclerosis

Neuroradiology

(1990)

M. Fillipi et al.

Correlations between changes in disability and T₂-weighted brain MRI activity in multiple sclerosisA follow-up study

Neurology

(1995)

D.W. Paty et al.

Interferon beta-1 is effective in relapsing remitting multiple sclerosis. MRI analysis results of a multi center, randomized, double-blind, placebo-controlled trial. UBC MS/MRI study group and the IFNB Multiple Sclerosis study group

Neurology

(1993)

Cited by (60)

Multiple Sclerosis lesions detection by a hybrid Watershed-Clustering algorithm
2021, Clinical Imaging
Citation Excerpt :
Moreover, we calculated the DICE coefficient to highlight the difference between manual and automated segmentation in according to Carass et al. study [15]. Automated methods of detection, identification and quantification of MS lesions were presented by many researchers [16–18]. Golberg et al. [16] used the automatic method and they obtained a specificity of 87% and a specificity of 96%.
Computer Aided Diagnosis (CAD) systems have been developing in the last years with the aim of helping the diagnosis and monitoring of several diseases. We present a novel CAD system based on a hybrid Watershed-Clustering algorithm for the detection of lesions in Multiple Sclerosis.
Magnetic Resonance Imaging scans (FLAIR sequences without gadolinium) of 20 patients affected by Multiple Sclerosis with hyperintense lesions were studied. The CAD system consisted of the following automated processing steps: images recording, automated segmentation based on the Watershed algorithm, detection of lesions, extraction of both dynamic and morphological features, and classification of lesions by Cluster Analysis.
The investigation was performed on 316 suspect regions including 255 lesion and 61 non-lesion cases. The Receiver Operating Characteristic analysis revealed a highly significant difference between lesions and non-lesions; the diagnostic accuracy was 87% (95% CI: 0.83–0.90), with an appropriate cut-off of 192.8; the sensitivity was 77% and the specificity was 87%.
In conclusion, we developed a CAD system by using a modified algorithm for automated image segmentation which may discriminate MS lesions from non-lesions. The proposed method generates a detection out-put that may be support the clinical evaluation.
icobrain ms 5.1: Combining unsupervised and supervised approaches for improving the detection of multiple sclerosis lesions
2021, NeuroImage: Clinical
Citation Excerpt :
There is also a wide range of metrics different studies have used to quantify the obtained results. Many of those are voxel-based similarity metrics, whereas others focused on lesion-level metrics (Goldberg-Zimring et al., 1998; Sajja et al., 2006; Styner et al., 2008; Yamamoto et al., 2010), such as lesion-wise Dice score. The latter are less sensitive to the manual delineation variabilities, which was one of the reasons we opted for these metrics in our experiments.
Multiple sclerosis (MS) is a chronic autoimmune, inflammatory neurological disease of the central nervous system. Its diagnosis nowadays commonly includes performing an MRI scan, as it is the most sensitive imaging test for MS. MS plaques are commonly identified from fluid-attenuated inversion recovery (FLAIR) images as hyperintense regions that are highly varying in terms of their shapes, sizes and locations, and are routinely classified in accordance to the McDonald criteria. Recent years have seen an increase in works that aimed at development of various semi-automatic and automatic methods for detection, segmentation and classification of MS plaques. In this paper, we present an automatic combined method, based on two pipelines: a traditional unsupervised machine learning technique and a deep-learning attention-gate 3D U-net network. The deep-learning network is specifically trained to address the weaker points of the traditional approach, namely difficulties in segmenting infratentorial and juxtacortical plaques in real-world clinical MRIs. It was trained and validated on a multi-center multi-scanner dataset that contains 159 cases, each with T1 weighted (T1w) and FLAIR images, as well as manual delineations of the MS plaques, segmented and validated by a panel of raters. The detection rate was quantified using lesion-wise Dice score. A simple label fusion is implemented to combine the output segmentations of the two pipelines. This combined method improves the detection of infratentorial and juxtacortical lesions by 14% and 31% respectively, in comparison to the unsupervised machine learning pipeline that was used as a performance assessment baseline.
Robust, atlas-free, automatic segmentation of brain MRI in health and disease
2019, Heliyon
Citation Excerpt :
While this technique is robust to morphological variations [31], it is highly sensitive to noise and radiofrequency bias, which can cause disconnections or holes within segmented regions. Adaptive thresholding algorithms [32, 33] performed better by eliminating many of these artifacts, but they suffer from inconsistency and often require significant manual input for lesion segmentation. C-DEF overcomes these issues by using multiple image contrasts, correcting bias fields, and eliminating the need for an atlas through incorporating image features, thereby resulting in a more robust disease-independent classification.
Brain- and lesion-volumes derived from magnetic resonance images (MRI) serve as important imaging markers of disease progression in neurodegenerative diseases and aging. While manual segmentation of these volumes is both tedious and impractical in large cohorts of subjects, automated segmentation methods often fail in accurate segmentation of brains with severe atrophy or high lesion loads. The purpose of this study was to develop an atlas-free brain Classification using DErivative-based Features (C-DEF), which utilizes all scans that may be acquired during the course of a routine MRI study at any center.
Proton-density, T₂-weighted, T₁-weighted, brain-free water, 3D FLAIR, 3D T₂-weighted, and 3D T₂*-weighted images, collected routinely on patients with neuroinflammatory diseases at the NIH, were used to optimize the C-DEF algorithm on healthy volunteers and HIV + subjects (cohort 1). First, manually marked lesions and eroded FreeSurfer brain segmentation masks (compiled into gray and white matter, globus pallidus, CSF labels) were used in training. Next, the optimized C-DEF was applied on a separate cohort of HIV + subjects (cohort two), and the results were compared with that of FreeSurfer and Lesion-TOADS. Finally, C-DEF segmentation was evaluated on subjects clinically diagnosed with various other neurological diseases (cohort three).
C-DEF algorithm was optimized using leave-one-out cross validation on five healthy subjects (age 36 ± 11 years), and five subjects infected with HIV (age 57 ± 2.6 years) in cohort one. The optimized C-DEF algorithm outperformed FreeSurfer and Lesion-TOADS segmentation in 49 other subjects infected with HIV (cohort two, age 54 ± 6 years) in qualitative and quantitative comparisons. Although trained only on HIV brains, sensitivity to detect lesions using C-DEF increased by 45% in HTLV-I-associated myelopathy/tropical spastic paraparesis (n = 5; age 58 ± 7 years), 33% in multiple sclerosis (n = 5; 42 ± 9 years old), and 4% in subjects with polymorphism of the cytotoxic T-lymphocyte-associated protein 4 gene (n = 5; age 24 ± 12 years) compared to Lesion-TOADS.
C-DEF outperformed other segmentation algorithms in the various neurological diseases explored herein, especially in lesion segmentation. While the results reported are from routine images acquired at the NIH, the algorithm can be easily trained and optimized for any set of contrasts and protocols for wider application. We are currently exploring various technical aspects of optimal implementation of CDEF in a clinical setting and evaluating a larger cohort of patients with other neurological diseases. Improving the accuracy of brain segmentation methodology will help better understand the relationship of imaging abnormalities to clinical and neuropsychological markers in disease.
Survey of automated multiple sclerosis lesion segmentation techniques on magnetic resonance imaging
2018, Computerized Medical Imaging and Graphics
Citation Excerpt :
All of these measures are based on comparing the result of the automated segmentation against the ground truth. In order to comprehend the evaluation measures, four basic retrieval terms and their clinical meaning should be understood (Goldberg-Zimring et al., 1998): True positive (TP): Refers to correctly segmented MS lesions areas.
Multiple sclerosis (MS) is a chronic disease. It affects the central nervous system and its clinical manifestation can variate. Magnetic Resonance Imaging (MRI) is often used to detect, characterize and quantify MS lesions in the brain, due to the detailed structural information that it can provide. Manual detection and measurement of MS lesions in MRI data is time-consuming, subjective and prone to errors. Therefore, multiple automated methodologies for MRI-based MS lesion segmentation have been proposed. Here, a review of the state-of-the-art of automatic methods available in the literature is presented.
The current survey provides a categorization of the methodologies in existence in terms of their input data handling, their main strategy of segmentation and their type of supervision. The strengths and weaknesses of each category are analyzed and explicitly discussed. The positive and negative aspects of the methods are highlighted, pointing out the future trends and, thus, leading to possible promising directions for future research. In addition, a further clustering of the methods, based on the databases used for their evaluation, is provided. The aforementioned clustering achieves a reliable comparison among methods evaluated on the same databases.
Despite the large number of methods that have emerged in the field, there is as yet no commonly accepted methodology that has been established in clinical practice. Future challenges such as the simultaneous exploitation of more sophisticated MRI protocols and the hybridization of the most promising methods are expected to further improve the performance of the segmentation.
BOOST: A supervised approach for multiple sclerosis lesion segmentation
2014, Journal of Neuroscience Methods
Citation Excerpt :
One of the advantages of these strategies is the implicit feature selection applied to the dataset, since they give a stronger impact to the most discriminating features while underscoring or even discarding the less discriminating ones. This feature pool used can comprise common features used for classification such as image intensities (Kamber et al., 1995; Boer et al., 2009), spatial locations (Anbeek et al., 2004), atlas probabilities (Goldberg-Zimring et al., 1998) or new features such as meta-features that exploit contextual information (Geremia et al., 2011). In this paper, we present BOOST, a new classification approach based on an ensemble classifier of regression stumps, with a feature pool that combines multi-channel image intensities, probability maps, contextual meta-features, and an outlier map based on image intensities.
Automatic multiple sclerosis lesion segmentation is a challenging task. An extensive analysis of the most recent techniques indicates an improvement of the results obtained when using prior knowledge and contextual information.
We present BOOST, a knowledge-based approach to automatically segment multiple sclerosis lesions through a voxel by voxel classification. We used the Gentleboost classifier and a set of features, including contextual features, registered atlas probability maps and an outlier map.
Results are computed on a set of 45 cases from three different hospitals (15 of each), obtaining a moderate agreement between the manual annotations and the automatically segmented results.
We quantitatively compared our results with three public state-of-the-art approaches obtaining competitive results and a better overlap with manual annotations. Our approach tends to better segment those cases with high lesion load, while cases with small lesion load are more difficult to accurately segment.
We believe BOOST has potential applicability in the clinical practice, although it should be improved in those cases with small lesion load.
OASIS is Automated Statistical Inference for Segmentation, with applications to multiple sclerosis lesion segmentation in MRI
2013, NeuroImage: Clinical
Magnetic resonance imaging (MRI) can be used to detect lesions in the brains of multiple sclerosis (MS) patients and is essential for diagnosing the disease and monitoring its progression. In practice, lesion load is often quantified by either manual or semi-automated segmentation of MRI, which is time-consuming, costly, and associated with large inter- and intra-observer variability. We propose OASIS is Automated Statistical Inference for Segmentation (OASIS), an automated statistical method for segmenting MS lesions in MRI studies. We use logistic regression models incorporating multiple MRI modalities to estimate voxel-level probabilities of lesion presence. Intensity-normalized T1-weighted, T2-weighted, fluid-attenuated inversion recovery and proton density volumes from 131 MRI studies (98 MS subjects, 33 healthy subjects) with manual lesion segmentations were used to train and validate our model. Within this set, OASIS detected lesions with a partial area under the receiver operating characteristic curve for clinically relevant false positive rates of 1% and below of 0.59% (95% CI; [0.50%, 0.67%]) at the voxel level. An experienced MS neuroradiologist compared these segmentations to those produced by LesionTOADS, an image segmentation software that provides segmentation of both lesions and normal brain structures. For lesions, OASIS out-performed LesionTOADS in 74% (95% CI: [65%, 82%]) of cases for the 98 MS subjects.
To further validate the method, we applied OASIS to 169 MRI studies acquired at a separate center. The neuroradiologist again compared the OASIS segmentations to those from LesionTOADS. For lesions, OASIS ranked higher than LesionTOADS in 77% (95% CI: [71%, 83%]) of cases. For a randomly selected subset of 50 of these studies, one additional radiologist and one neurologist also scored the images. Within this set, the neuroradiologist ranked OASIS higher than LesionTOADS in 76% (95% CI: [64%, 88%]) of cases, the neurologist 66% (95% CI: [52%, 78%]) and the radiologist 52% (95% CI: [38%, 66%]).
OASIS obtains the estimated probability for each voxel to be part of a lesion by weighting each imaging modality with coefficient weights. These coefficients are explicit, obtained using standard model fitting techniques, and can be reused in other imaging studies. This fully automated method allows sensitive and specific detection of lesion presence and may be rapidly applied to large collections of images.

View all citing articles on Scopus

View full text

Original ContributionsAutomated Detection and Characterization of Multiple Sclerosis Lesions in Brain MR Images

Abstract

Introduction

Section snippets

Basic Assumptions

Results

Discussion

Acknowledgements

Computer Methods and Programs in Biomedicine

Knowledge-based interpretation of magnetic resonance imagesDetecting multiple sclerosis lesions

Artificial Intelligence in Medicine

Magnetic resonance imaging of spinal cord lesions in multiple sclerosis

J. Neurol. Neurosurg. Psychiatry

Serial gadolinium-enhanced magnetic resonance imaging scans in patients with early, relapsing-remitting multiple sclerosisimplications for clinical trials and natural history

Ann. Neurol.

Magnetic resonance in monitoring the treatment of multiple sclerosis

Ann. Neurol.

Magnetic resonance imaging in monitoring the treatment of multiple sclerosisConcerted action guidelines

J. Neurol. Neurosurg. Psychiatry

Correlation of magnetic resonance imaging with neuropsychological testing in multiple sclerosis

Neurology

Neuropsychological and structural brain lesions in multiple sclerosisA regional analysis

Neurology

Abnormal corpus callosumA sensitive and specific indicator of multiple sclerosis

Radiology

Magnetic resonance imaging in the evaluation of treatment in multiple sclerosis

Neuroradiology

Correlations between changes in disability and T2-weighted brain MRI activity in multiple sclerosisA follow-up study

Neurology

Interferon beta-1 is effective in relapsing remitting multiple sclerosis. MRI analysis results of a multi center, randomized, double-blind, placebo-controlled trial. UBC MS/MRI study group and the IFNB Multiple Sclerosis study group

Neurology

Original Contributions
Automated Detection and Characterization of Multiple Sclerosis Lesions in Brain MR Images

Correlations between changes in disability and T₂-weighted brain MRI activity in multiple sclerosisA follow-up study