DeepGRAI (Deep Gray Rating via Artificial Intelligence): Fast, feasible, and clinically relevant thalamic atrophy measurement on clinical quality T2-FLAIR MRI in multiple sclerosis

Clinical applications of deep learning in neuroinflammatory diseases: A scoping review

BACKGROUND

Deep learning (DL) is an artificial intelligence technology that has aroused much excitement for predictive medicine due to its ability to process raw data modalities such as images, text, and time series of signals.

OBJECTIVES

Here, we intend to give the clinical reader elements to understand this technology, taking neuroinflammatory diseases as an illustrative use case of clinical translation efforts. We reviewed the scope of this rapidly evolving field to get quantitative insights about which clinical applications concentrate the efforts and which data modalities are most commonly used.

METHODS

We queried the PubMed database for articles reporting DL algorithms for clinical applications in neuroinflammatory diseases and the radiology.healthairegister.com website for commercial algorithms.

RESULTS

The review included 148 articles published between 2018 and 2024 and five commercial algorithms. The clinical applications could be grouped as computer-aided diagnosis, individual prognosis, functional assessment, the segmentation of radiological structures, and the optimization of data acquisition. Our review highlighted important discrepancies in efforts. The segmentation of radiological structures and computer-aided diagnosis currently concentrate most efforts with an overrepresentation of imaging. Various model architectures have addressed different applications, relatively low volume of data, and diverse data modalities. We report the high-level technical characteristics of the algorithms and synthesize narratively the clinical applications. Predictive performances and some common a priori on this topic are finally discussed.

CONCLUSION

The currently reported efforts position DL as an information processing technology, enhancing existing modalities of paraclinical investigations and bringing perspectives to make innovative ones actionable for healthcare.

Thalamic atrophy and dysconnectivity are associated with cognitive impairment in a multi-center, clinical routine, real-word study of people with relapsing-remitting multiple sclerosis

BACKGROUND

Prior research has established a link between thalamic pathology and cognitive impairment (CI) in people with multiple sclerosis (pwMS). However, the translation of these findings to pwMS in everyday clinical settings has been insufficient.

OBJECTIVE

To assess which global and/or thalamic imaging biomarkers can be used to identify pwMS at risk for CI and cognitive worsening (CW) in a real-world setting.

METHODS

This was an international, multi-center (11 centers), longitudinal, retrospective, real-word study of people with relapsing-remitting MS (pwRRMS). Brain MRI exams acquired at baseline and follow-up were collected. Cognitive status was evaluated using the Symbol Digit Modalities Test (SDMT). Thalamic volume (TV) measurement was performed on T2-FLAIR, as well as on T1-WI, when available. Thalamic dysconnectivity, T2-lesion volume (T2-LV), and volumes of gray matter (GM), whole brain (WB) and lateral ventricles (LVV) were also assessed.

RESULTS

332 pwMS were followed for an average of 2.8 years. At baseline, T2-LV, LVV, TV and thalamic dysconnectivity on T2-FLAIR (p < 0.016), and WB, GM and TV volumes on T1-WI (p < 0.039) were significantly worse in 90 (27.1 %) CI vs. 242 (62.9 %) non-CI pwRRMS. Greater SDMT decline over the follow-up was associated with lower baseline TV on T2-FLAIR (standardized β = 0.203, p = 0.002) and greater thalamic dysconnectivity (standardized β = -0.14, p = 0.028) in a linear regression model.

CONCLUSIONS

PwRRMS with thalamic atrophy and worse thalamic dysconnectivity present more frequently with CI and experience greater CW over mid-term follow-up in a real-world setting.

Dynamic disability measures decrease the clinico-radiological gap in people with severely affected multiple sclerosis

BACKGROUND

Expanded Disability Status Scale (EDSS) is limited when utilized in highly disabled people with multiple sclerosis (pwMS).

OBJETIVE

To explore the relationship between disability measures and MRI outcomes in severely-affected pwMS.

METHODS

PwMS recruited from The Boston Home (TBH), a specialized residential facility for severly-affected pwMS and University at Buffalo (UB) MS Center were assessed using EDSS, MS Severity Scale, age-related MSS, Scripps Neurological Rating Scale (SNRS) and Combinatorial Weight-Adjusted Disability Score (CombiWISE). In all scores except SNRS, higher score indicates greater disability. MRI measures of T1, T2-lesion volume (LV), whole brain, gray matter, medulla oblongata and thalamic volumes (WBV, GMV, MOV, TV) and thalamic dysconnectivity were obtained.

RESULTS

Greatest disability differences between the TBH and UB pwMS were in SNRS (24.4 vs 71.9, p < 0.001, Cohen's d = 4.05) and CombiWISE (82.3 vs. 38.9, p < 0.001, Cohen's d = 4.02). In combined analysis of all pwMS, worse SNRS scores were correlated with worse MRI pathology in 8 out of 9 outcomes. EDSS only with 3 measures (GMV, MOV and TV). In severely-affected pwMS, SNRS was associated with T1-LV, T2-LV and WBV (not surviving false discovery rate (FDR) correction for multiple comparisons) whereas EDSS did not.

CONCLUSION

Granular and dynamic disability measures may bridge the clinico-radiologcal gap present in severely affected pwMS.

Differentiating multiple sclerosis from non-specific white matter changes using a convolutional neural network image classification model

BACKGROUND

The diagnosis of multiple sclerosis (MS) relies heavily on neuroimaging with magnetic resonance imaging (MRI) and exclusion of mimics. This can be a challenging task due to radiological overlap in several disorders and may require ancillary testing or longitudinal follow up. One of the most common radiological MS mimickers is non-specific white matter disease (NSWMD). We aimed to develop and evaluate models leveraging machine learning algorithms to help distinguish MS and NSWMD.

METHODS

All adult patients who underwent MRI brain using a demyelinating protocol with available electronic medical records between 2015 and 2019 at Cleveland Clinic affiliated facilities were included. Diagnosis of MS and NSWMD were assessed from clinical documentation. Those with a diagnosis of MS and NSWMD were matched using total T2 lesion volume (T2LV) and used to train models with logistic regression and convolutional neural networks (CNN). Performance metrices were reported for each model.

RESULTS

A total of 250 NSWMD MRI scans were identified, and 250 unique MS MRI scans were matched on T2LV. Cross validated logistic regression model was able to use 20 variables (including spinal cord area, regional volumes, and fractions) to predict MS compared to NSWMD with 68.0% accuracy while the CNN model was able to classify MS compared to NSWMD in two independent validation and testing cohorts with 77% and 78% accuracy on average.

CONCLUSION

Automated methods can be used to differentiate MS compared to NSWMD. These methods can be used to supplement currently available diagnostic tools for patients being evaluated for MS.

MRI features and disability in multiple sclerosis: A systematic review and meta-analysis

BACKGROUND

In this systematic review and meta-analysis, we aimed to investigate the correlation between disability in patients with Multiple sclerosis (MS) measured by the Expanded Disability Status Scale (EDSS) and brain Magnetic Resonance Imaging (MRI) features to provide reliable results on which characteristics in the MRI can predict disability and prognosis of the disease.

METHODS

A systematic literature search was performed using three databases including PubMed, Scopus, and Web of Science. The selected peer-reviewed studies must report a correlation between EDSS scores and MRI features. The correlation coefficients of included studies were converted to the Fisher's z scale, and the results were pooled.

RESULTS

Overall, 105 studies A total of 16,613 patients with MS entered our study. We found no significant correlation between total brain volume and EDSS assessment (95 % CI: -0.37 to 0.08; z-score: -0.15). We examined the potential correlation between the volume of T1 and T2 lesions and the level of disability. A positive significant correlation was found (95 % CI: 0.19 to 0.43; z-score: 0.31), (95 % CI: 0.17 to 0.33; z-score: 0.25). We observed a significant correlation between white matter volume and EDSS score in patients with MS (95 % CI: -0.37 to -0.03; z-score: -0.21). Moreover, there was a significant negative correlation between gray matter volume and disability (95 % CI: -0.025 to -0.07; z-score: -0.16).

CONCLUSION

In conclusion, this systematic review and meta-analysis revealed that disability in patients with MS is linked to extensive changes in different brain regions, encompassing gray and white matter, as well as T1 and T2 weighted MRI lesions.

Feasibility of detecting atrophy relevant for disability and cognition in multiple sclerosis using 3D-FLAIR

BACKGROUND AND OBJECTIVES

Disability and cognitive impairment are known to be related to brain atrophy in multiple sclerosis (MS), but 3D-T1 imaging required for brain volumetrics is often unavailable in clinical protocols, unlike 3D-FLAIR. Here our aim was to investigate whether brain volumes derived from 3D-FLAIR images result in similar associations with disability and cognition in MS as do those derived from 3D-T1 images.

METHODS

3T-MRI scans of 329 MS patients and 76 healthy controls were included in this cross-sectional study. Brain volumes were derived using FreeSurfer on 3D-T1 and compared with brain volumes derived with SynthSeg and SAMSEG on 3D-FLAIR. Relative agreement was evaluated by calculating the intraclass correlation coefficient (ICC) of the 3D-T1 and 3D-FLAIR volumes. Consistency of relations with disability and average cognition was assessed using linear regression, while correcting for age and sex. The findings were corroborated in an independent validation cohort of 125 MS patients.

RESULTS

The ICC between volume measured with FreeSurfer and those measured on 3D-FLAIR for brain, ventricle, cortex, total deep gray matter and thalamus was above 0.74 for SAMSEG and above 0.91 for SynthSeg. Worse disability and lower average cognition were similarly associated with brain (adj. R2 = 0.24-0.27, p < 0.01; adj. R2 = 0.26-0.29, p < 0.001) ventricle (adj. R2 = 0.27-0.28, p < 0.001; adj. R2 = 0.19-0.20, p < 0.001) and deep gray matter volumes (adj. R2 = 0.24-0.28, p < 0.001; adj. R2 = 0.27-0.28, p < 0.001) determined with all methods, except for cortical volumes derived from 3D-FLAIR.

DISCUSSION

In this cross-sectional study, brain volumes derived from 3D-FLAIR and 3D-T1 show similar relationships to disability and cognitive dysfunction in MS, highlighting the potential of these techniques in clinical datasets.

Artificial Intelligence and Multiple Sclerosis: Up-to-Date Review

Multiple sclerosis (MS) remains a challenging neurological disorder for the clinician in terms of diagnosis and management. The growing integration of AI-based algorithms in healthcare offers a golden opportunity for clinicians and patients with MS. AI models are based on statistical analyses of large quantities of data from patients including "demographics, genetics, clinical and radiological presentation." These approaches are promising in the quest for greater diagnostic accuracy, tailored management plans, and better prognostication of disease. The use of AI in multiple sclerosis represents a paradigm shift in disease management. With ongoing advancements in AI technologies and the increasing availability of large-scale datasets, the potential for further innovation is immense. As AI continues to evolve, its integration into clinical practice will play a vital role in improving diagnostics, optimizing treatment strategies, and enhancing patient outcomes for MS. This review is about conducting a literature review to identify relevant studies on AI applications in MS. Only peer-reviewed studies published in the last four years have been selected. Data related to AI techniques, advancements, and implications are extracted. Through data analysis, key themes and tendencies are identified. The review presents a cohesive synthesis of the current state of AI and MS, highlighting potential implications and new advancements.

Therapy effect on AI-derived thalamic atrophy using clinical routine MRI protocol: A longitudinal, multi-center, propensity-matched multiple sclerosis study

BACKGROUND

The effect of disease modifying therapies (DMTs) on brain atrophy in persons with multiple sclerosis (pwMS) is typically investigated in highly standardized clinical trial settings or single-center academic institutions. We aimed at utilizing artificial intelligence (AI)-based volumetric analysis on routine unstandardized T2-FLAIR scans in determining the effect of DMTs on lateral ventricular volume (LVV) and thalamic volume (TV) changes in pwMS.

METHODS

The DeepGRAI (Deep Gray Rating via Artificial Intelligence) registry is a multi-center, longitudinal, observational, real-word study with a convenience sample of 1002 relapsing-remitting (RR) pwMS from 30 United States sites. Brain MRI exams acquired as part of the routine clinical management were collected at baseline and on average at 2.6-years follow-up. The MRI scans were acquired either on 1.5T or 3T scanners with no prior harmonization. TV was determined using the DeepGRAI tool and lateral ventricular volume LVV was measured using NeuroSTREAM software.

RESULTS

After propensity matching based on baseline age, disability and time of follow-up, untreated pwRRMS had significantly greater TV change when compared to treated pwRRMS (-1.2% vs. -0.3%, p = 0.044). PwRRMS treated with high-efficacy DMTs had significant and two-fold lower% LVV change when compared to pwRRMS treated on moderate-efficacy DMTs (3.5% vs. 7.0%, p = 0.001). PwRRMS who stopped DMT during the follow-up had significantly greater annualized% TV change compared to pwRRMS who remained on their DMT (-0.73% vs. -0.14%, p = 0.012) and significantly greater annualized% LVV change (3.4% vs. 1.7%, p = 0.047). These findings were also observed in a propensity analysis that additionally incorporated matching for scanner model at both baseline and follow-up visits.

CONCLUSIONS

LVV and TV measured on T2-FLAIR scans can detect treatment-induced short-term neurodegenerative changes measured in a real-word unstandardized, multicenter, clinical routine setting.

Thalamic atrophy measured by artificial intelligence in a multicentre clinical routine real-word study is associated with disability progression

BACKGROUND

The thalamus is a key grey matter structure, and sensitive marker of neurodegeneration in multiple sclerosis (MS). Previous reports indicated that thalamic volumetry using artificial intelligence (AI) on clinical-quality T2-fluid-attenuated inversion recovery (FLAIR) images alone is fast and reliable.

OBJECTIVE

To investigate whether thalamic volume (TV) loss, measured longitudinally by AI, is associated with disability progression (DP) in patients with MS, participating in a large multicentre study.

METHODS

The DeepGRAI (Deep Grey Rating via Artificial Intelligence) Registry is a multicentre (30 USA sites), longitudinal, observational, retrospective, real-word study of relapsing-remitting (RR) MS patients. Each centre enrolled between 30 and 35 patients. Brain MRI exams acquired at baseline and follow-up on 1.5T or 3T scanners with no prior standardisation were collected. TV measurement was performed on T2-FLAIR using DeepGRAI, and on two dimensional (D)-weighted and 3D T1-weighted images (WI) by using FMRIB's Integrated Registration and Segmentation Tool software where possible.

RESULTS

1002 RRMS patients were followed for an average of 2.6 years. Longitudinal TV analysis was more readily available on T2-FLAIR (96.1%), compared with 2D-T1-WI (61.8%) or 3D-T1-WI (33.2%). Over the follow-up, DeepGRAI TV loss was significantly higher in patients with DP, compared with those with disability improvement (DI) or disease stability (-1.35% in DP, -0.87% in DI and -0.57% in Stable, p=0.045, Bonferroni-adjusted, age-adjusted and follow-up time-adjusted analysis of covariance). In a regression model including MRI scanner change, age, sex, disease duration and follow-up time, DP was associated with DeepGRAI TV loss (p=0.022).

CONCLUSIONS

Thalamic atrophy measured by AI in a multicentre clinical routine real-word setting is associated with DP over mid-term follow-up.

MVnet: automated time-resolved tracking of the mitral valve plane in CMR long-axis cine images with residual neural networks: a multi-center, multi-vendor study

BACKGROUND

Mitral annular plane systolic excursion (MAPSE) and left ventricular (LV) early diastolic velocity (e') are key metrics of systolic and diastolic function, but not often measured by cardiovascular magnetic resonance (CMR). Its derivation is possible with manual, precise annotation of the mitral valve (MV) insertion points along the cardiac cycle in both two and four-chamber long-axis cines, but this process is highly time-consuming, laborious, and prone to errors. A fully automated, consistent, fast, and accurate method for MV plane tracking is lacking. In this study, we propose MVnet, a deep learning approach for MV point localization and tracking capable of deriving such clinical metrics comparable to human expert-level performance, and validated it in a multi-vendor, multi-center clinical population.

METHODS

The proposed pipeline first performs a coarse MV point annotation in a given cine accurately enough to apply an automated linear transformation task, which standardizes the size, cropping, resolution, and heart orientation, and second, tracks the MV points with high accuracy. The model was trained and evaluated on 38,854 cine images from 703 patients with diverse cardiovascular conditions, scanned on equipment from 3 main vendors, 16 centers, and 7 countries, and manually annotated by 10 observers. Agreement was assessed by the intra-class correlation coefficient (ICC) for both clinical metrics and by the distance error in the MV plane displacement. For inter-observer variability analysis, an additional pair of observers performed manual annotations in a randomly chosen set of 50 patients.

RESULTS

MVnet achieved a fast segmentation (<1 s/cine) with excellent ICCs of 0.94 (MAPSE) and 0.93 (LV e') and a MV plane tracking error of -0.10 ± 0.97 mm. In a similar manner, the inter-observer variability analysis yielded ICCs of 0.95 and 0.89 and a tracking error of -0.15 ± 1.18 mm, respectively.

CONCLUSION

A dual-stage deep learning approach for automated annotation of MV points for systolic and diastolic evaluation in CMR long-axis cine images was developed. The method is able to carefully track these points with high accuracy and in a timely manner. This will improve the feasibility of CMR methods which rely on valve tracking and increase their utility in a clinical setting.

Quantifying disease pathology and predicting disease progression in multiple sclerosis with only clinical routine T2-FLAIR MRI

•

We explored five brain pathology measures from clinical-quality T2-FLAIR MRI in MS.

•

These included LVV, thalamus volume, MOV, SCLV and network efficiency.

•

T2-FLAIR measures predicted a majority of the variance in research-quality MRI.

•

T2-FLAIR measures correlated with neurologic disability and cognitive function.

•

T2-FLAIR measures predicted disability progression over five-years.

•

T2-FLAIR measures can be used in legacy clinical datasets.

Background

Although quantitative measures from research-quality MRI provide a means to study multiple sclerosis (MS) pathology in vivo, these metrics are often unavailable in legacy clinical datasets.

Objective

To determine how well an automatically-generated quantitative snapshot of brain pathology, measured only on clinical routine T2-FLAIR MRI, can substitute for more conventional measures on research MRI in terms of capturing multi-factorial disease pathology and providing similar clinical relevance.

Methods

MRI with both research-quality sequences and conventional clinical T2-FLAIR was acquired for 172 MS patients at baseline, and neurologic disability was assessed at baseline and five-years later. Five measures (thalamus volume, lateral ventricle volume, medulla oblongata volume, lesion volume, and network efficiency) for quantifying disparate aspects of neuropathology from low-resolution T2-FLAIR were applied to predict standard research-quality MRI measures. They were compared in regard to association with future neurologic disability and disease progression over five years.

Results

The combination of the five T2-FLAIR measures explained most of the variance in standard research-quality MRI. T2-FLAIR measures were associated with neurologic disability and cognitive function five-years later (R² = 0.279, p < 0.001; R² = 0.382, p < 0.001), similar to standard research-quality MRI (R² = 0.279, p < 0.001; R² = 0.366, p < 0.001). They also similarly predicted disability progression over five years (%-correctly-classified = 69.8, p = 0.034), compared to standard research-quality MRI (%-correctly-classified = 72.4%, p = 0.022) in relapsing-remitting MS.

Conclusion

A set of five T2-FLAIR-only measures can substitute for standard research-quality MRI, especially in relapsing-remitting MS. When only clinical T2-FLAIR is available, it can be used to obtain substantially more quantitative information about brain pathology and disability than is currently standard practice.