Automated PET tracer classification in multi-site, multi-cohort studies using deep learning
Deep learning classifies brain PET radiotracers directly from scans across sites and cohorts, improving metadata quality control.
robel-k.-gebre
RGDr. Robel Gebre
Assistant Professor of Radiology
I develop PET and MRI biomarkers that reveal how neurodegenerative disease varies across people and over time.
Research Focus
I use deep learning, explainable AI, and disease progression models to turn medical images into useful measures of disease. My current work focuses on tau PET quantification, MRI harmonization across sites, and imaging differences between multiple system atrophy and Parkinson’s disease.
I created the THETA score, a tau PET measure that captures the spatial pattern of tau across the brain.
My publications also appear under Robel K. Gebre and Robel Kebede Gebre.
Featured Publications
•
•
1 min read
Precise disease heterogeneity and progression quantification in MSA and Parkinson's disease using machine learning
We developed a machine learning framework to quantify disease heterogeneity and progression in multiple system atrophy and Parkinson's disease using structural and diffusion MRI.
robel-k.-gebre
Precise Estimation Of Heterogenous Global Tau Burden In The Brain Sheds Light On Cognitive Resilience Mechanisms
We applied the THETA score to investigate cognitive resilience mechanisms in individuals with heterogeneous global tau burden.
robel-k.-gebre
Can integration of Alzheimer's plasma biomarkers with MRI, cardiovascular, genetics, and lifestyle measures improve cognition prediction?
MRI, plasma biomarkers, genetics, cardiovascular risk, and lifestyle measures were compared for predicting cognitive decline.
robel-k.-gebre
Advancing Tau PET Quantification in Alzheimer Disease with Machine Learning: Introducing THETA, a Novel Tau Summary Measure
THETA is a machine-learning tau PET measure that captures the spatial pattern of tau deposition across the brain in Alzheimer's disease.
robel-k.-gebre
Identifying transition points in AD biomarkers using machine learning: A supplement to the reliable worsening method
We developed a machine learning method to identify transition points in Alzheimer's disease biomarkers as a supplement to the reliable worsening method.
robel-k.-gebre
Cross-scanner harmonization methods for structural MRI may need further work: A comparison study
A systematic comparison questions the reliability of widely used cross-scanner harmonization methods for structural MRI.
robel-k.-gebre
Detecting hip osteoarthritis on clinical CT: A deep learning application based on 2-D summation images derived from CT
We developed the first deep learning application for detecting hip osteoarthritis from 2-D summation images derived from routine clinical CT scans.
robel-k.-gebre
Latest Publications
(2026).
Regional diffusion imaging measures to disentangle SVD-related hypertensive arteriopathy versus cerebral amyloid angiopathy.
Molecular Degeneration.
(2026).
Chapter 8 - Artificial intelligence: Relevance and applications in dementia.
Artificial Intelligence and Brain-Computer Interfaces in Healthcare, Academic Press.
(2025).
Automatic Classification of Head MR Series.
Alzheimer’s & Dementia.
(2025).
Developing scanner change invariant brain age models for aging and dementia studies.
Alzheimer’s & Dementia.
(2025).
Diffusion Imaging Measures to Disentangle SVD-related Hypertensive Arteriopathy versus Cerebral Amyloid Angiopathy.
Cerebral Circulation - Cognition and Behavior.
Recent & Upcoming Talks
Alzheimer's Association International Conference (AAIC) 2026
I will remotely attend AAIC 2026 in London to present recent work in neuroimaging and machine learning for Alzheimer's disease.
PET Tracer Classification and the Tau Therapeutic Window
A podium talk on a 2.5D deep learning classifier for PET tracer identification, and a poster introducing the tau therapeutic window as an eligibility criterion for anti-amyloid …
Recent Blogs
From noise to PET: diffusion is winning, but is it really?
Diffusion methods have overtaken GANs for generating images. Turning MRI into PET is within reach, yet a realistic picture is not a trustworthy one.
TPE: finding the turning point in disease progression
Disease trajectories are not lines. They bend, and the bend is where the clinical meaning hides.
Where a model looks is not why it decides
Explainability tools tell you which features a model used. They rarely tell you which ones mattered.
Cerebral Microbleeds
Why cerebral microbleeds are difficult to detect and segment on MRI, and what they mean for small vessel disease and anti-amyloid therapy safety.
•
1 min read