Applying Deep Learning to Accelerated Clinical Brain Magnetic Resonance Imaging for Multiple Sclerosis

Reliability of brain volume measures of accelerated 3D T1-weighted images with deep learning-based reconstruction

PURPOSE

The time-intensive nature of acquiring 3D T1-weighted MRI and analyzing brain volumetry limits quantitative evaluation of brain atrophy. We explore the feasibility and reliability of deep learning-based accelerated MRI scans for brain volumetry.

METHODS

This retrospective study collected 3D T1-weighted data using 3T from 42 participants for the simulated acceleration dataset and 48 for the validation dataset. The simulated acceleration dataset consists of three sets at different simulated acceleration levels (Simul-Accel) corresponding to level 1 (65% undersampling), 2 (70%), and 3 (75%). These images were then subjected to deep learning-based reconstruction (Simul-Accel-DL). Conventional images (Conv) without acceleration and DL were set as the reference. In the validation dataset, DICOM images were collected from Conv and accelerated scan with DL-based reconstruction (Accel-DL). The image quality of Simul-Accel-DL was evaluated using quantitative error metrics. Volumetric measurements were evaluated using intraclass correlation coefficients (ICCs) and linear regression analysis in both datasets. The volumes were estimated by two software, NeuroQuant and DeepBrain.

RESULTS

Simul-Accel-DL across all acceleration levels revealed comparable or better error metrics than Simul-Accel. In the simulated acceleration dataset, ICCs between Conv and Simul-Accel-DL in all ROIs exceeded 0.90 for volumes and 0.77 for normative percentiles at all acceleration levels. In the validation dataset, ICCs for volumes > 0.96, ICCs for normative percentiles > 0.89, and R2 > 0.93 at all ROIs except pallidum demonstrated good agreement in both software.

CONCLUSION

DL-based reconstruction achieves clinical feasibility of 3D T1 brain volumetric MRI by up to 75% acceleration relative to full-sampled acquisition.

Longitudinal Digital Phenotyping of Multiple Sclerosis Severity Using Passively Sensed Behaviors and Ecological Momentary Assessments

Background

Longitudinal tracking of multiple sclerosis (MS) symptoms in an individual's own environment may improve self-monitoring and clinical management for people with MS (pwMS).

Objective

We present a machine learning approach that enables longitudinal monitoring of clinically relevant patient-reported symptoms for pwMS by harnessing passively collected data from sensors in smartphones and fitness trackers.

Methods

We divide the collected data into discrete periods for each patient. For each prediction period, we first extract patient-level behavioral features from the current period (action features) and the previous period (context features). Then, we apply a machine learning (ML) approach based on Support Vector Machine with Radial Bias Function Kernel and AdaBoost to predict the presence of depressive symptoms (every two weeks) and high global MS symptom burden, severe fatigue, and poor sleep quality (every four weeks).

Results

Between November 16, 2019, and January 24, 2021, 104 pwMS (84.6% women, 93.3% non-Hispanic White, 44.0±11.8 years mean±SD age) from a clinic-based MS cohort completed 12-weeks of data collection, including a subset of 44 pwMS (88.6% women, 95.5% non-Hispanic White, 45.7±11.2 years) who completed 24-weeks of data collection. In total, we collected approximately 12,500 days of passive sensor and behavioral health data from the participants. Among the best-performing models with the least sensor data requirement, ML algorithm predicts depressive symptoms with an accuracy of 80.6% (35.5% improvement over baseline; F1-score: 0.76), high global MS symptom burden with an accuracy of 77.3% (51.3% improvement over baseline; F1-score: 0.77), severe fatigue with an accuracy of 73.8% (45.0% improvement over baseline; F1-score: 0.74), and poor sleep quality with an accuracy of 72.0% (28.1% improvement over baseline; F1-score: 0.70). Further, sensor data were largely sufficient for predicting symptom severity, while the prediction of depressive symptoms benefited from minimal active patient input in the form of response to two brief questions on the day before the prediction point.

Conclusions

Our digital phenotyping approach using passive sensors on smartphones and fitness trackers may help patients with real-world, continuous, self-monitoring of common symptoms in their own environment and assist clinicians with better triage of patient needs for timely interventions in MS (and potentially other chronic neurological disorders).

Accelerated spine MRI with deep learning based image reconstruction: a prospective comparison with standard MRI

RATIONALE AND OBJECTIVES

To evaluate the performance of deep learning (DL) reconstructed MRI in terms of image acquisition time, overall image quality and diagnostic interchangeability compared to standard-of-care (SOC) MRI.

MATERIALS AND METHODS

This prospective study recruited participants between July 2023 and August 2023 who had spinal discomfort. All participants underwent two separate MRI examinations (Standard and accelerated scanning). Signal-to-noise ratios (SNR), contrast-to-noise ratios (CNR) and similarity metrics were calculated for quantitative evaluation. Four radiologists performed subjective quality and lesion characteristic assessment. Wilcoxon test was used to assess the differences of SNR, CNR and subjective image quality between DL and SOC. Various lesions of spine were also tested for interchangeability using individual equivalence index. Interreader and intrareader agreement and concordance (κ and Kendall τ and W statistics) were computed and McNemar tests were performed for comprehensive evaluation.

RESULTS

200 participants (107 male patients, mean age 46.56 ± 17.07 years) were included. Compared with SOC, DL enabled scan time reduced by approximately 40%. The SNR and CNR of DL were significantly higher than those of SOC (P < 0.001). DL showed varying degrees of improvement (0-0.35) in each of similarity metrics. All absolute individual equivalence indexes were less than 4%, indicating interchangeability between SOC and DL. Kappa and Kendall showed a good to near-perfect agreement in range of 0.72-0.98. There is no difference between SOC and DL regarding subjective scoring and frequency of lesion detection.

CONCLUSION

Compared to SOC, DL provided high-quality image for diagnosis and reduced examination time for patients. DL was found to be interchangeable with SOC in detecting various spinal abnormalities.

Ultrafast Brain MRI at 3 T for MS: Evaluation of a 51-Second Deep Learning-Enhanced T2-EPI-FLAIR Sequence

In neuroimaging, there is no equivalent alternative to magnetic resonance imaging (MRI). However, image acquisitions are generally time-consuming, which may limit utilization in some cases, e.g., in patients who cannot remain motionless for long or suffer from claustrophobia, or in the event of extensive waiting times. For multiple sclerosis (MS) patients, MRI plays a major role in drug therapy decision-making. The purpose of this study was to evaluate whether an ultrafast, T2-weighted (T2w), deep learning-enhanced (DL), echo-planar-imaging-based (EPI) fluid-attenuated inversion recovery (FLAIR) sequence (FLAIRUF) that has targeted neurological emergencies so far might even be an option to detect MS lesions of the brain compared to conventional FLAIR sequences. Therefore, 17 MS patients were enrolled prospectively in this exploratory study. Standard MRI protocols and ultrafast acquisitions were conducted at 3 tesla (T), including three-dimensional (3D)-FLAIR, turbo/fast spin-echo (TSE)-FLAIR, and FLAIRUF. Inflammatory lesions were grouped by size and location. Lesion conspicuity and image quality were rated on an ordinal five-point Likert scale, and lesion detection rates were calculated. Statistical analyses were performed to compare results. Altogether, 568 different lesions were found. Data indicated no significant differences in lesion detection (sensitivity and positive predictive value [PPV]) between FLAIRUF and axially reconstructed 3D-FLAIR (lesion size ≥3 mm × ≥2 mm) and no differences in sensitivity between FLAIRUF and TSE-FLAIR (lesion size ≥3 mm total). Lesion conspicuity in FLAIRUF was similar in all brain regions except for superior conspicuity in the occipital lobe and inferior conspicuity in the central brain regions. Further findings include location-dependent limitations of signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) as well as artifacts such as spatial distortions in FLAIRUF. In conclusion, FLAIRUF could potentially be an expedient alternative to conventional methods for brain imaging in MS patients since the acquisition can be performed in a fraction of time while maintaining good image quality.

Hippocampal subfields and thalamic nuclei associations with clinical outcomes in multiple sclerosis: An ultrahigh field MRI study

BACKGROUND

Previous studies have shown that thalamic and hippocampal neurodegeneration is associated with clinical decline in Multiple Sclerosis (MS). However, contributions of the specific thalamic nuclei and hippocampal subfields require further examination.

OBJECTIVE

Using 7 Tesla (7T) magnetic resonance imaging (MRI), we investigated the cross-sectional associations between functionally grouped thalamic nuclei and hippocampal subfields volumes and T1 relaxation times (T1-RT) and subsequent clinical outcomes in MS.

METHODS

High-resolution T1-weighted and T2-weighted images were acquired at 7T (n=31), preprocessed, and segmented using the Thalamus Optimized Multi Atlas Segmentation (THOMAS, for thalamic nuclei) and the Automatic Segmentation of Hippocampal Subfields (ASHS, for hippocampal subfields) packages. We calculated Pearson correlations between hippocampal subfields and thalamic nuclei volumes and T1-RT and subsequent multi-modal rater-determined and patient-reported clinical outcomes (∼2.5 years after imaging acquisition), correcting for confounders and multiple tests.

RESULTS

Smaller volume bilaterally in the anterior thalamus region correlated with worse performance in gait function, as measured by the Patient Determined Disease Steps (PDDS). Additionally, larger volume in most functional groups of thalamic nuclei correlated with better visual information processing and cognitive function, as measured by the Symbol Digit Modalities Test (SDMT). In bilateral medial and left posterior thalamic regions, there was an inverse association between volumes and T1-RT, potentially indicating higher tissue degeneration in these regions. We also observed marginal associations between the right hippocampal subfields (both volumes and T1-RT) and subsequent clinical outcomes, though they did not survive correction for multiple testing.

CONCLUSION

Ultrahigh field MRI identified markers of structural damage in the thalamic nuclei associated with subsequently worse clinical outcomes in individuals with MS. Longitudinal studies will enable better understanding of the role of microstructural integrity in these brain regions in influencing MS outcomes.

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.

Impact of Deep Learning Image Reconstruction Methods on MRI Throughput

Purpose To evaluate the effect of implementing two distinct commercially available deep learning reconstruction (DLR) algorithms on the efficiency of MRI examinations conducted in real clinical practice within an outpatient setting at a large, multicenter institution. Materials and Methods This retrospective study included 7346 examinations from 10 clinical MRI scanners analyzed during the pre- and postimplementation periods of DLR methods. Two different types of DLR methods, namely Digital Imaging and Communications in Medicine (DICOM)-based and k-space-based methods, were implemented in half of the scanners (three DICOM-based and two k-space-based), while the remaining five scanners had no DLR method implemented. Scan and room times of each examination type during the pre- and postimplementation periods were compared among the different DLR methods using the Wilcoxon test. Results The application of deep learning methods resulted in significant reductions in scan and room times for certain examination types. The DICOM-based method demonstrated up to a 53% reduction in scan times and a 41% reduction in room times for various study types. The k-space-based method demonstrated up to a 27% reduction in scan times but did not significantly reduce room times. Conclusion DLR methods were associated with reductions in scan and room times in a clinical setting, though the effects were heterogeneous depending on examination type. Thus, potential adopters should carefully evaluate their case mix to determine the impact of integrating these tools. Keywords: Deep Learning MRI Reconstruction, Reconstruction Algorithms, DICOM-based Reconstruction, k-Space-based Reconstruction © RSNA, 2024 See also the commentary by GharehMohammadi in this issue.

Impact of the COVID-19 Pandemic on the Personal Networks and Neurological Outcomes of People With Multiple Sclerosis: Cross-Sectional and Longitudinal Case-Control Study

BACKGROUND

The coronavirus disease 2019 (COVID-19) pandemic has negatively affected the social fabric.

OBJECTIVE

We evaluated the associations between personal social networks and neurological function in people with multiple sclerosis (pwMS) and controls in the prepandemic and pandemic periods.

METHODS

During the early pandemic (March-December 2020), 8 cohorts of pwMS and controls completed a questionnaire quantifying the structure and composition of their personal social networks, including the health behaviors of network members. Participants from 3 of the 8 cohorts had additionally completed the questionnaire before the pandemic (2017-2019). We assessed neurological function using 3 interrelated patient-reported outcomes: Patient Determined Disease Steps (PDDS), Multiple Sclerosis Rating Scale-Revised (MSRS-R), and Patient-Reported Outcomes Measurement Information System (PROMIS) Physical Function. We identified the network features associated with neurological function using paired 2-tailed t tests and covariate-adjusted regressions.

RESULTS

In the cross-sectional analysis of the pandemic data from 1130 pwMS and 1250 controls during the pandemic, having a higher percentage of network members with a perceived negative health influence was associated with worse disability in pwMS (MSRS-R: β=2.181, 95% CI 1.082-3.279; P<.001) and poor physical function in controls (PROMIS Physical Function: β=-5.707, 95% CI -7.405 to -4.010; P<.001). In the longitudinal analysis of 230 pwMS and 136 controls, the networks of all participants contracted, given an increase in constraint (pwMS-prepandemic: mean 52.24, SD 15.81; pwMS-pandemic: mean 56.77, SD 18.91; P=.006. Controls-prepandemic: mean 48.07, SD 13.36; controls-pandemic: mean 53.99, SD 16.31; P=.001) and a decrease in network size (pwMS-prepandemic: mean 8.02, SD 5.70; pwMS-pandemic: mean 6.63, SD 4.16; P=.003. Controls-prepandemic: mean 8.18, SD 4.05; controls-pandemic: mean 6.44, SD 3.92; P<.001), effective size (pwMS-prepandemic: mean 3.30, SD 1.59; pwMS-pandemic: mean 2.90, SD 1.50; P=.007. Controls-prepandemic: mean 3.85, SD 1.56; controls-pandemic: mean 3.40, SD 1.55; P=.01), and maximum degree (pwMS-prepandemic: mean 4.78, SD 1.86; pwMS-pandemic: mean 4.32, SD 1.92; P=.01. Controls-prepandemic: mean 5.38, SD 1.94; controls-pandemic: mean 4.55, SD 2.06; P<.001). These network changes were not associated with worsening function. The percentage of kin in the networks of pwMS increased (mean 46.06%, SD 29.34% to mean 54.36%, SD 30.16%; P=.003) during the pandemic, a change that was not seen in controls.

CONCLUSIONS

Our findings suggest that high perceived negative health influence in the network was associated with worse function in all participants during the pandemic. The networks of all participants became tighter knit, and the percentage of kin in the networks of pwMS increased during the pandemic. Despite these perturbations in social connections, network changes from the prepandemic to the pandemic period were not associated with worsening function in all participants, suggesting possible resilience.

DeepSWI: Using Deep Learning to Enhance Susceptibility Contrast on T2*-Weighted MRI

BACKGROUND

Although susceptibility-weighted imaging (SWI) is the gold standard for visualizing cerebral microbleeds (CMBs) in the brain, the required phase data are not always available clinically. Having a postprocessing tool for generating SWI contrast from T2*-weighted magnitude images is therefore advantageous.

PURPOSE

To create synthetic SWI images from clinical T2*-weighted magnitude images using deep learning and evaluate the resulting images in terms of similarity to conventional SWI images and ability to detect radiation-associated CMBs.

STUDY TYPE

Retrospective.

POPULATION

A total of 145 adults (87 males/58 females; 43.9 years old) with radiation-associated CMBs were used to train (16,093 patches/121 patients), validate (484 patches/4 patients), and test (2420 patches/20 patients) our networks.

FIELD STRENGTH/SEQUENCE

3D T2*-weighted, gradient-echo acquired at 3 T.

ASSESSMENT

Structural similarity index (SSIM), peak signal-to-noise-ratio (PSNR), normalized mean-squared-error (nMSE), CMB counts, and line profiles were compared among magnitude, original SWI, and synthetic SWI images. Three blinded raters (J.E.V.M., M.A.M., B.B. with 8-, 6-, and 4-years of experience, respectively) independently rated and classified test-set images.

STATISTICAL TESTS

Kruskall-Wallis and Wilcoxon signed-rank tests were used to compare SSIM, PSNR, nMSE, and CMB counts among magnitude, original SWI, and predicted synthetic SWI images. Intraclass correlation assessed interrater variability. P values <0.005 were considered statistically significant.

RESULTS

SSIM values of the predicted vs. original SWI (0.972, 0.995, 0.9864) were statistically significantly higher than that of the magnitude vs. original SWI (0.970, 0.994, 0.9861) for whole brain, vascular structures, and brain tissue regions, respectively; 67% (19/28) CMBs detected on original SWI images were also detected on the predicted SWI, whereas only 10 (36%) were detected on magnitude images. Overall image quality was similar between the synthetic and original SWI images, with less artifacts on the former.

CONCLUSIONS

This study demonstrated that deep learning can increase the susceptibility contrast present in neurovasculature and CMBs on T2*-weighted magnitude images, without residual susceptibility-induced artifacts. This may be useful for more accurately estimating CMB burden from magnitude images alone.

EVIDENCE LEVEL

3.

TECHNICAL EFFICACY

Stage 2.

Applications of Artificial Intelligence in the Radiology Roundtrip: Process Streamlining, Workflow Optimization, and Beyond

There are many impactful applications of artificial intelligence (AI) in the electronic radiology roundtrip and the patient's journey through the healthcare system that go beyond diagnostic applications. These tools have the potential to improve quality and safety, optimize workflow, increase efficiency, and increase patient satisfaction. In this article, we review the role of AI for process improvement and workflow enhancement which includes applications beginning from the time of order entry, scan acquisition, applications supporting the image interpretation task, and applications supporting tasks after image interpretation such as result communication. These non-diagnostic workflow and process optimization tasks are an important part of the arsenal of potential AI tools that can streamline day to day clinical practice and patient care.

Impact of the COVID-19 Pandemic on Personal Networks and Neurological Outcomes of People with Multiple Sclerosis: A Case-Control Cross-sectional and Longitudinal Analysis

Background

The COVID-19 pandemic has negatively impacted the social fabric of people with multiple sclerosis (pwMS).

Objective

To evaluate the associations between personal social network environment and neurological function in pwMS and controls during the COVID-19 pandemic and compare with the pre-pandemic baseline.

Methods

We first analyzed data collected from 8 cohorts of pwMS and control participants during the COVID-19 pandemic (March-December 2020). We then leveraged data collected between 2017-2019 in 3 of the 8 cohorts for longitudinal comparison. Participants completed a questionnaire that quantified the structure and composition of their personal social network, including the health behaviors of network members. We assessed neurological disability using three interrelated patient-reported outcomes: Patient Determined Disease Steps (PDDS), Multiple Sclerosis Rating Scale â€" Revised (MSRS-R), and Patient Reported Outcomes Measurement Information System (PROMIS)-Physical Function. We identified the network features associated with neurologic disability using paired t-tests and covariate-adjusted regressions.

Results

In the cross-sectional analysis of the pandemic data from 1130 pwMS and 1250 control participants, higher percent of network members with a perceived negative health influence was associated with greater neurological symptom burden in pwMS (MSRS-R: Beta[95% CI]=2.181[1.082, 3.279], p<.001) and worse physical function in controls (PROMIS-Physical Function: Beta[95% CI]=-5.707[-7.405, -4.010], p<.001). In the longitudinal analysis of 230 pwMS and 136 control participants, the networks of both pwMS and controls experienced an increase in constraint (pwMS p=.006, control p=.001) as well as a decrease in network size (pwMS p=.003, control p<.001), effective size (pwMS p=.007, control p=.013), maximum degree (pwMS p=.01, control p<.001), and percent contacted weekly or less (pwMS p<.001, control p<.001), suggesting overall network contraction during the COVID-19 pandemic. There was also an increase in percentage of kin (p=.003) in the networks of pwMS but not controls during the COVID-19 pandemic when compared to the pre-pandemic baseline. These changes in personal social network due to the pandemic were not associated with worsening neurological disability during the pandemic.

Conclusions

Our findings suggest that perceived negative health influences in personal social networks are associated with worse disability in all participants during the COVID-19 pandemic. Despite the perturbation in social environment and connections during the pandemic, the stability in neurological function among pwMS suggests potential resilience.

Predicting Multiple Sclerosis Outcomes During the COVID-19 Stay-at-home Period: Observational Study Using Passively Sensed Behaviors and Digital Phenotyping

BACKGROUND

The COVID-19 pandemic has broad negative impact on the physical and mental health of people with chronic neurological disorders such as multiple sclerosis (MS).

OBJECTIVE

We presented a machine learning approach leveraging passive sensor data from smartphones and fitness trackers of people with MS to predict their health outcomes in a natural experiment during a state-mandated stay-at-home period due to a global pandemic.

METHODS

First, we extracted features that capture behavior changes due to the stay-at-home order. Then, we adapted and applied an existing algorithm to these behavior-change features to predict the presence of depression, high global MS symptom burden, severe fatigue, and poor sleep quality during the stay-at-home period.

RESULTS

Using data collected between November 2019 and May 2020, the algorithm detected depression with an accuracy of 82.5% (65% improvement over baseline; F₁-score: 0.84), high global MS symptom burden with an accuracy of 90% (39% improvement over baseline; F₁-score: 0.93), severe fatigue with an accuracy of 75.5% (22% improvement over baseline; F₁-score: 0.80), and poor sleep quality with an accuracy of 84% (28% improvement over baseline; F₁-score: 0.84).

CONCLUSIONS

Our approach could help clinicians better triage patients with MS and potentially other chronic neurological disorders for interventions and aid patient self-monitoring in their own environment, particularly during extraordinarily stressful circumstances such as pandemics, which would cause drastic behavior changes.

Role of artificial intelligence in MS clinical practice

Machine learning (ML) and its subset, deep learning (DL), are branches of artificial intelligence (AI) showing promising findings in the medical field, especially when applied to imaging data. Given the substantial role of MRI in the diagnosis and management of patients with multiple sclerosis (MS), this disease is an ideal candidate for the application of AI techniques. In this narrative review, we are going to discuss the potential applications of AI for MS clinical practice, together with their limitations. Among their several advantages, ML algorithms are able to automate repetitive tasks, to analyze more data in less time and to achieve higher accuracy and reproducibility than the human counterpart. To date, these algorithms have been applied to MS diagnosis, prognosis, disease and treatment monitoring. Other fields of application have been improvement of MRI protocols as well as automated lesion and tissue segmentation. However, several challenges remain, including a better understanding of the information selected by AI algorithms, appropriate multicenter and longitudinal validations of results and practical aspects regarding hardware and software integration. Finally, one cannot overemphasize the paramount importance of human supervision, in order to optimize the use and take full advantage of the potential of AI approaches.