Advancing Understanding of ALS Motor Dysfunction: 7T fMRI Explores Neural Mechanisms

306 

Abstract Submission 

Avinash Kalyani¹, Christoph Reichert², Alicia Northall³, Esther Kühn⁴

¹Otto-von-Guericke-University Magdeburg, Magdeburg, Saxony Anhalt, ²Leibniz Institute for Neurobiology, Magdeburg, Saxony Anhalt, ³Nuffield Department of Clinical Neurosciences, Oxford, United Kingdom, ⁴German Center for Neurodegenerative Diseases (DZNE), Tübingen, Baden-Württemberg

Avinash Kalyani  
Otto-von-Guericke-University Magdeburg
Magdeburg, Saxony Anhalt

Christoph Reichert  
Leibniz Institute for Neurobiology
Magdeburg, Saxony Anhalt

Alicia Northall  
Nuffield Department of Clinical Neurosciences
Oxford, United Kingdom

Esther Kühn  
German Center for Neurodegenerative Diseases (DZNE)
Tübingen, Baden-Württemberg

Amyotrophic Lateral Sclerosis (ALS) poses challenges in understanding motor learning deficits. Leveraging the heightened sensitivity of 7 T fMRI, we explored the neural substrates of motor learning in ALS. Employing Shared Response Modeling (SRM) for inter-subject normalization, we identified robust neural patterns. Partial Least Squares (PLS) analysis revealed associations between 7T fMRI data and behavioral outcomes, shedding light on the intricate neural mechanisms underlying motor dysfunction in ALS. Our study highlights the pivotal role of 7T fMRI and advanced analytical techniques in enhancing our understanding of ALS-related motor impairments which can help to develop potential therapeutic strategies.

In the study conducted between June 2018 and December 2022, 12 individuals diagnosed with Amyotrophic Lateral Sclerosis (ALS) (6 females, age: M = 60.5, SD = 12.7) were compared to an equal number of age, handedness, gender, and education-matched healthy controls (6 females, age: M = 61.1, SD = 11.9). Disease onset varied among patients, with 7 having upper limb-onset, 2 with lower limb-onset, and 3 with bulbar-onset (Alicia, 2023).

Data collection utilized a 7T-MRI scanner in Magdeburg, Germany, a 1.5 mm isotropic resolution fMRI was obtained using an EPI GE-BOLD sequence. The fMRI employed a blocked-design paradigm with 12-sec movement periods (left/right foot, left/right hand, tongue) alternated with 15-sec rest intervals, with 20 trials (75 min). A GLM analysis was employed to localize the voxels specific to the body part moved. In the SRM (Chen, 2015) analysis, the data underwent reordering, and subsequently, a linear SVM was trained for both classification purposes and in the context of Partial Least Squares (PLS) analysis. The ALS Functional Rating Scale-Revised (ALSFRS-R) data served as the behavioral dataset to train and fit the PLS regression model. This approach allowed for the exploration and integration of shared neural response patterns and behavioral outcomes, providing a comprehensive understanding of the relationships between the observed brain activity and motor function in the context of ALS.

The study identified significant group differences in different ROIs localized using GLM for various body part movements, evidenced by a permutation test score showing above-chance accuracy in distinguishing healthy controls from ALS patients (Figure 1, A). Tongue region classification accuracy was relatively lower, potentially due to the limited number of Bulbar onset ALS subjects. Predictability for upper limb movements in ALS was notably reduced, in line with the prevalence of upper limb onset in patients (Figure 1, B). Controls exhibited higher movement-based task stimulus classification, with mean values of 0.53 (SD = 0.16) compared to 0.45 (SD = 0.11) for ALS patients. PLS regression analysis connecting ALSFRS-R scores and fMRI data revealed higher mean squared error (MSE) for foot region and hand task behavioral scores, indicating lower predictability for hand-based measurements from foot movement-invoked fMRI data (Figure 1, C). In an exploratory analysis, scatterplots of latent variables highlighted distinct clustering for Bulbar onset ALS and other onset types, underscoring unique patterns in the association between neural and behavioral measures (Figure 1, D).

In summary, our 7T fMRI study revealed group-specific differences particularly in the tongue and upper limb regions, underscored the complexity of ALS subtypes. The overall movement-based task stimulus classification favored controls, suggesting distinctive neural responses. The higher MSE for foot-related fMRI data in relation to hand-based behavioral measurements warrants careful interpretation. Additionally, our exploratory analysis highlighted separate clustering for Bulbar onset ALS, revealing potential subtype-specific patterns.These findings enhance our understanding of ALS-related motor impairments for targeted interventions.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ¹

Activation (eg. BOLD task-fMRI)

Classification and Predictive Modeling ²

Motor Behavior Other

Affective Disorders

Data analysis

Design and Analysis

DISORDERS

Machine Learning

Modeling

Motor

MRI

Multivariate