Background Longitudinal measurement of medial temporal lobe (MTL) atrophy is shown to be effective in measuring Alzheimer’s disease (AD) progression and has been extensively studied on T1-weighted MRI. However, there...
Background Longitudinal measurement of medial temporal lobe (MTL) atrophy is shown to be effective in measuring Alzheimer’s disease (AD) progression and has been extensively studied on T1-weighted MRI. However, there has been little work analyzing atrophy on high-resolution T2-weighted MRI of the hippocampal region, and it is unknown which MRI modality would yield a more effective biomarker of disease progression. We compare atrophy measurements derived from T1 and T2-weighted MRI using a deep learning framework DeepAtrophy and a deformation-based morphometry method ALOHA. Method T2-weighted MRI of 481 participants from ADNI2/3 are included in the study. DeepAtrophy was trained on 74 participants and evaluated on 407. During training, DeepAtrophy explicitly infers temporal information, such as the scan temporal order for a single scan pair and which image pair has a longer scan time interval for multiple scan pairs. During inference, DeepAtrophy predicts a “total progression score” for each scan pair. This score is expected to be higher (more progression) in individuals with more advanced AD. A similar experiment was conducted using T1-weighted MRI from 481 ADNI2/GO participants (partially overlapping with T2-weighted experiment). ALOHA pipelines tailored for T1-weighted and hippocampal T2-weighted MRI were evaluated on the same subjects and scan pairs as DeepAtrophy. Result For both T1 and T2-weighted MRI, DeepAtrophy is more accurate, compared with ALOHA, in inferring the temporal order of a pair of longitudinal scans. In hypothetical one-year and two-years clinical trials scenarios simulated using ADNI data, DeepAtrophy has an improvement over ALOHA in detecting differences in progression between disease groups in both imaging modalities. The strongest effect for differentiating between Amyloid-positive and Amyloid-negative cognitively unimpaired group was obtained using DeepAtrophy progression measures on T2-weighted MRI. Conclusion This study supports the poten
Background Genetic forms of frontotemporal degeneration (FTD) and amyotrophic lateral sclerosis (ALS) provide an important target for early therapeutic interventions. However, pre-symptomatic biomarkers to monitor dis...
Background Genetic forms of frontotemporal degeneration (FTD) and amyotrophic lateral sclerosis (ALS) provide an important target for early therapeutic interventions. However, pre-symptomatic biomarkers to monitor disease progression are lacking. Prior neuroimaging studies suggest disruption of functional and structural networks are an early feature of neurodegenerative disease. Network modularity quantifies the degree of segregation between sub-networks. We hypothesize that structural network modularity is a marker of neurodegeneration in C9orf72 carriers. Method We evaluated diffusion magnetic resonance imaging (dMRI) in two independent cohorts of C9orf72 carriers and controls: University of pennsylvania Prodromal Study (penn-C9) including symptomatic C9orf72 carriers (SC9; N=27; FTD=20; ALS=7), asymptomatic C9orf72 carriers (AC9; N=24), and healthy controls (HC; N=29); and Human Connectome Project-FTD (HCP-FTD) also including SC9 (N=12), AC9 (N=12), and HC (N=20) individuals. We quantified modularity in 7 intrinsic networks of 400 nodes from the Schaefer Atlas using participation coefficient (PC), where lower PC corresponds to higher modularity, and also investigated its constituent measures: within-network connectivity for each network and pairwise between-network connectivity. Modularity metrics were residualized in the penn-C9 dataset using linear regression to adjust for age. Statistical analyses on residualized scores were first performed in penn-C9 and cross-validated in the HCP-FTD dataset. Result ANCOVAs on residualized scores in penn-C9 revealed group differences in PC for limbic network with SC9 lower than HC (partial-η 2 =0.12, p=0.011). We also observed group differences in within-module connectivity for limbic (partial-η 2 =0.16, p=0.0016) and dorsal-attention (partial-η 2 =0.1, p=0.019) networks with SC9 lower than HC. Cross-validation in HCP-FTD revealed group differences in modularity metrics for limbic network (PC: partial-η 2 =0.29, p=0.0012 and
Background The relationship between tau neurofibrillary tangles (T) and neurodegeneration (N) may offer clues to potential presence of comorbidities, as well as relative resilience and vulnerability to Alzheimer’s di...
Background The relationship between tau neurofibrillary tangles (T) and neurodegeneration (N) may offer clues to potential presence of comorbidities, as well as relative resilience and vulnerability to Alzheimer’s disease (AD) pathology. We have previously developed measures of tau PET (T) and cortical thickness (N) mismatch based on linear model residuals. However, the underlying relationships between T and N are likely complex and non-linear. Moreover, predicting cortical thickness based on local tau level may miss non-local contributions of tauopathy. Here we investigate T-N mismatch using a deep-learning 3D image translation neural network to estimate synthetic maps of cortical thickness based on tau PET images. Deviation between the synthetic and actual cortical thickness serves as a metric of T-N mismatch. Method We derived 3D cortical thickness from T1-MRI and SUVR from 18F-flortaucipir tracer uptake to represent N and T maps respectively. We predicted cortical thickness maps from tau SUVR by training a 3D U-Net (Ronneberger et al. 2015) to learn the relationship between paired cortical tau SUVR and thickness maps from 70 symptomatic patients from ADNI. Approximately 70% of the training set was A+. We used regional standardized mean absolute error between predicted and actual thickness across 104 bilateral gray matter regions of interest as T-N mismatch for clustering in a separate independent sample of 194 A+ symptomatic patients. Result The voxel-wise mean absolute error of thickness translation on 194 tested patients was 0.54 mm. We obtained six T-N data-driven clusters (Figure 2). The group with lowest reconstruction error was defined as canonical – meaning that the degree of neurodegeneration was accurately predicted by tau PET. There were three groups with lower thickness than predicted, which were denoted as temporal-limbic (TL) vulnerable, posterior vulnerable and diffuse vulnerable based on their spatial patterns. Additionally, two groups with greate
暂无评论