Clinical Scorecard: Artificial Intelligence-Driven Synthetic Brain MRI for Evaluating Regional Atrophy in Neurodegenerative Disorders
At a Glance
Category
Detail
Condition
Neurodegenerative disorders including Alzheimer's dementia (AD) and Frontotemporal dementia (FTD)
Key Mechanisms
Use of AI-based deep learning to generate synthetic high-resolution 3D T1-weighted brain MRIs from various input MRI sequences for volumetric analysis of regional brain atrophy
Target Population
Adult patients (>18 years) with clinically confirmed AD or FTD and healthy controls
Care Setting
Clinical routine MRI imaging and diagnostic evaluation in neuroimaging and neurology departments
Key Highlights
SynthSR deep learning tool generates synthetic 3D T1-weighted MRIs from multimodal clinical MRI inputs (e.g., FLAIR, T2) with high correlation to acquired 3D T1 sequences.
Synthetic sequences enable volumetric analysis of disease-specific regional atrophy patterns in AD and FTD patients.
Proposed standardized nomenclature differentiates real (r) vs synthetic (s) images and specifies input sequence type and orientation for clarity.
Guideline-Based Recommendations
Diagnosis
Use 1-mm isotropic 3D T1 sequences as reference standard for volumetric brain atrophy analysis in neurodegenerative disorders.
Consider AI-generated synthetic 3D T1 sequences from available clinical MRI inputs when 3D T1 is not acquired.
Management
Incorporate synthetic 3D T1-weighted images into automated volumetric and classification workflows to improve diagnostic decision-making.
Maintain interdisciplinary clinical confirmation including psychiatry and nuclear medicine for diagnosis.
Monitoring & Follow-up
Perform visual quality inspection of MRI sequences prior to volumetric analysis.
Use age and sex as covariates in volumetric assessments to account for demographic influences.
Risks
Potential variability in synthetic image quality depending on input sequence contrast and resolution.
Need for anonymization (defacing) when uploading images externally, which has negligible impact on atrophy estimation.
Patient & Prescribing Data
Adults with Alzheimer’s dementia, Frontotemporal dementia, and healthy controls undergoing routine clinical MRI
Synthetic MRI sequences generated via deep learning can supplement missing 3D T1 acquisitions to enable volumetric brain atrophy evaluation and support automated classification algorithms.
Clinical Best Practices
Ensure acquisition of at least 1 mm isovoxel 3D T1, 3D FLAIR, and coronal T2 sequences during MRI for comprehensive analysis.
Apply SynthSR tool within FreeSurfer for generating synthetic 3D T1 images from diverse input sequences.
Use standardized nomenclature (rT1, sT13DFLAIR, sT1CorT2) to clearly distinguish real and synthetic images and their input origins.
Conduct retrospective image quality checks and use appropriate anonymization methods before external data sharing.
Incorporate demographic covariates such as age and sex in volumetric and classification analyses.
