Data-driven staging of genetic frontotemporal dementia using multi-modal MRI

Area of Science:

• Neuroscience
• Machine Learning
• Radiology

Background:

• Genetic frontotemporal dementia (FTD) is highly heterogeneous, presenting diagnostic and therapeutic challenges.
• Developing unified methods to stage FTD during presymptomatic and symptomatic phases is crucial for clinical trial advancement.

Purpose of the Study:

• To apply contrastive trajectory inference (cTI), an unsupervised machine learning algorithm, for staging genetic FTD.
• To validate cTI's ability to identify individual disease stages using MRI metrics in a large cohort.

Main Methods:

• Utilized cross-sectional MRI data (gray matter density, T1/T2 ratio, functional amplitude, FA, MD) from 383 gene carriers and 253 controls.
• Employed cTI to analyze temporal patterns and generate individual disease stage scores.
• Correlated cTI scores with estimated years to onset, clinical assessments, and neuropsychological tests.

Main Results:

• cTI-derived disease scores significantly correlated with clinical and neuropsychological measures across behavioral, cognitive, and executive domains.
• Mean diffusivity showed the highest contribution to cTI disease staging.
• Presymptomatic carriers exhibited higher mean cTI scores than controls, suggesting detection of early cerebral changes.

Conclusions:

• cTI serves as a proof-of-concept for data-driven disease staging in heterogeneous genetic FTD.
• The method successfully integrates various MRI metrics to capture disease progression.
• cTI holds potential for improving clinical trial design and outcome assessment in FTD research.