Decomposing the rTMS-related brain functional heterogeneity of Alzheimer's disease
Poster No:
126
Submission Type:
Abstract Submission
Authors:
Yilu Li1, Hao Wu1, Lin Liu1, Pan Wang1, Bharat Biswal2
Institutions:
1University of Electronic Science and Technology of China, Chengdu, Sichuan, 2New Jersey Institute of Technology, Newark, NJ
First Author:
Co-Author(s):
Introduction:
Repetitive transcranial magnetic stimulation (rTMS) has emerged as a promising non-invasive intervention for Alzheimer's disease (AD). However, the effects of rTMS in AD patients exhibit significant heterogeneity. Understanding the heterogeneity is crucial for optimizing rTMS protocols and tailoring interventions to maximize therapeutic efficacy in AD patients.
Methods:
Current study included 92 individuals with mild-to-moderate probable AD. Participants were classified into real group (56 individuals) or sham group (36 individuals) randomly. The therapy was performed based on an accelerated high-dose (iTBS) protocol, lasted 14 days for 50400 pulses. All MRI images were collected on a clinical Siemens 3.0 T scanner. The T1 images were collected with TR = 2.53 s, TE = 2.98 ms, slice thickness = 1 mm, slice number = 192, voxel size = 1 × 1 × 1 mm3. The resting-state fMRI images were acquired with TR = 3 or 2 s, TE = 30 ms, time points = 160 or 200, slice thickness = 3 or 3.5 mm, slice number = 47 or 32, voxel size = 3.5 × 3.5 × 3 or 3.5 × 3.5 × 4.375 mm3, flip angle = 90°. The MRI data were preprocessed with DPARSF 5.4 and SPM 12. For each participant, the first 10 or 50 time points of functional data were removed, and the remaining 150 volumes were used for subsequent preprocessing including slice-timing correction and realignment, nuisance regressing, normalization and smooth. The T1 images were co-registered to the functional one and then segmented. All participants had max head motion less than 3 mm or 3°.
To obtain the mask of gray matter, we took the union with Harvard-Oxford cortical and subcortical structural atlases to make the template. The template was then taken intersection with all ALFF maps to get the final group mask (40725 voxels). We calculated the percentage change in ALFF before and after rTMS for each voxel in the group mask. The percentage change is defined as follows:
PC=(A-B)/B,B≠0
Where PC is the percentage change value, A is the ALFF value after rTMS, B is the ALFF value before rTMS.
The PC matrix (56 participants × 40725 voxels) was obtained and then Z-scored as inputs for NMF analysis. Back Reconstruction was performed on the sham group data (36 participants × 40725 voxels) by using factors from the real group. K-means clustering was performed on the weight matrix of participants in real and sham group respectively.
To obtain the mask of gray matter, we took the union with Harvard-Oxford cortical and subcortical structural atlases to make the template. The template was then taken intersection with all ALFF maps to get the final group mask (40725 voxels). We calculated the percentage change in ALFF before and after rTMS for each voxel in the group mask. The percentage change is defined as follows:
PC=(A-B)/B,B≠0
Where PC is the percentage change value, A is the ALFF value after rTMS, B is the ALFF value before rTMS.
The PC matrix (56 participants × 40725 voxels) was obtained and then Z-scored as inputs for NMF analysis. Back Reconstruction was performed on the sham group data (36 participants × 40725 voxels) by using factors from the real group. K-means clustering was performed on the weight matrix of participants in real and sham group respectively.
Results:
We identified three subtypes among participants in the real group. The stable Subtype 1, comprising 43.86% of the participants, exhibited a dispersed distribution with no significant bias toward any factor, indicating minimal brain changes before and after rTMS stimulation. In contrast, Subtype 2, a responsive subtype, showed a greater weighting toward positive factor 1 and negative factor 2, suggesting enhanced activity in brain regions related to higher-order cognitive functions and reduced activity in primary functional areas post-stimulation. Subtype 3, another responsive subtype, representing 17.54% of the group, displayed strong weights on positive factor 2 and negative factor 1, indicating increased activity in primary functional regions but decreased activity in higher-order cognitive areas following stimulation. Moreover, the real group had significantly more participants in responsive subtype compared to the sham group (χ2 = 4.27, p = 0.039).
·Fig.1 The identified four latent disease factors. PF1, positive factor 1; PF2, positive factor 2; NF1, negative factor 1; NF2, negative factor 2.
·Fig.2 Three subtypes. PW1, weight of positive factor 1; PF2, weight of positive factor 2; NW1, weight of negative factor 1; NW2, weight of negative factor 2.
Conclusions:
Current study identifies three subtypes of brain functional changes in response to rTMS treatment in AD, highlight the complexity of neural responses to rTMS, underscores the importance of personalized treatment protocol. These findings could lay the groundwork for future studies exploring the mechanisms underlying the three subtypes and their implications for optimizing rTMS treatment protocols.
Brain Stimulation:
Non-invasive Magnetic/TMS
TMS 2
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Modeling and Analysis Methods:
Task-Independent and Resting-State Analysis
Keywords:
Transcranial Magnetic Stimulation (TMS)
Other - Alzheimer's disease; Heterogeneity
1|2Indicates the priority used for review