End-to-end imaging pipelines
Designing and building the full path from raw scan to analysable measure — multimodal MRI, fMRI, PET and DTI. Neuroimaging remains the core; the same approach now extends to retinal imaging and abdominal MR.
Boehringer Ingelheim · Germany
Amir Dehsarvi
Principal Scientist, Imaging Biomarkers at Boehringer Ingelheim
I build end-to-end pipelines for processing and analysing medical imaging data, turning MRI, fMRI, PET and DTI into quantitative biomarkers for diagnosis, target identification and drug development. Neuroimaging is the core, increasingly alongside retinal imaging and abdominal MR.
0 yrs
Working with medical imaging data, from PhD through academia to industry
0 yrs
As Chief Technology Officer of an AI medical-device company
0
Clinical ML device programmes contributed to — PD-, LID- and MCI-Monitor
0
Multi-site Parkinson's trials supported as technical and data coordinator
What I do
Research rigour, built to ship
My work sits between the lab and the clinic: methods that hold up to peer review, delivered as systems that teams can run, trust and act on.
Reproducible analysis at scale
Automated, reproducible processing and data engineering on HPC infrastructure, so multi-site imaging studies run consistently and other teams can pick the work up and run it themselves.
Evidence teams can act on
Analyses validated against clinical endpoints and delivered as measures that support diagnosis, target identification and drug development — informed by earlier work building machine learning for regulated medical devices and clinical trials.
Selected work
Projects with a measurable footprint
A few pieces of work that other people now depend on — pipelines, devices and trial infrastructure rather than papers alone.
Imaging biomarkers across organ systems
Building and running the pipelines and analyses that turn imaging data into quantitative biomarkers for drug development. Neuroimaging is the core of the work, with the remit extending to retinal imaging and abdominal MR.
ADPrep — end-to-end multimodal imaging pipeline
An automated pipeline standardising multimodal imaging data (MRI, fMRI, PET, DTI) for large multi-site studies, from raw acquisition through quality control to analysable measures. Set for integration into the GRIP platform, a Gates Ventures initiative.
ClearSky diagnostic device portfolio
Seven years leading technology for ML-based medical devices for diagnosing and monitoring neurodegenerative conditions — PD-Monitor, LID-Monitor and MCI-Monitor — working alongside clinicians and engineers to move AI into clinical use.
Device and digital-marker data across multi-site trials
Technical and data coordination for the devices and digital markers used in Parkinson's disease trials — PD-STAT, spanning 235 participants at 23 NHS hospitals, the Phase III Exenatide-PD3 study and the Phase II AZA-PD study — keeping instrumentation and data consistent across sites.
Career
From evolutionary algorithms to imaging biomarkers in drug development
A decade spanning academic research, a medical-device company and now pharmaceutical R&D — across the UK, Ireland and Germany.
2025 — present
Principal Scientist, Imaging Biomarkers
Boehringer Ingelheim, Germany
Building end-to-end pipelines for processing and analysing imaging data, and the quantitative biomarkers derived from them — neuroimaging at the core, extending across organ systems including the retina and abdominal MR.
2022 — 2025
Postdoctoral Researcher
Institute for Stroke and Dementia Research (ISD), LMU Munich
Built ADPrep and the tau-PET A/T/N model; managed HPC resources and supported labs across the institute.
2018 — 2025
Chief Technology Officer
ClearSky Medical Diagnostics Ltd., York
Led development of ML-based medical devices for neurodegenerative diagnosis and monitoring.
2021 — 2022
Postdoctoral Research Associate
University of York
White-box machine learning on resting-state fMRI to separate depression from healthy controls in UK Biobank data.
2021
ML & Image Processing Engineer
smartR.ai, Edinburgh
Deep learning pipelines to normalise and match colour profiles between FIBI and H&E histological images.
2020 — 2021
Research Fellow
University of Aberdeen
Defined a fatigue-related brain network in rheumatoid arthritis and explored how therapies affect it.
2019
Postdoctoral Researcher in Neuroimaging
Trinity College Dublin
Linked speech patterns to brain volume change in MCI and Alzheimer's disease as an early marker of decline.
2014 — 2018
PhD, Electronic Engineering
University of York
Cartesian Genetic Programming classification of resting-state fMRI: towards a brain imaging biomarker for Parkinson's disease.
Recent publications
Latest from the literature
A selection of recent work, largely on imaging biomarkers of neurodegeneration. The complete record is on Google Scholar and ORCID.
2026
ApoE4 lowers the ptau217 threshold for tau aggregation and spread in an allele dose-dependent manner
Brain
This study demonstrates that the ApoE4 risk allele lowers the threshold of plasma phosphorylated tau 217 (ptau217) required to drive fibrillar tau aggregation and spread in Alzheimer’s disease in a dose-dependent manner.
2026
Defining patient-centered amyloid PET thresholds for the onset of tauopathy in Alzheimer's disease
Alzheimer's & Dementia, 2026 (Online Early View)
This study establishes patient-centered amyloid PET thresholds for the onset of tauopathy, demonstrating that age and sex significantly influence the amyloid-to-tauopathy transition in Alzheimer’s disease.
2025
Connector hubs accelerate the spread of tau pathology in Alzheimer's disease
Alzheimer's & Dementia, vol 21 (Suppl 8), e109933
AAIC 2025 conference abstract, published in the Alzheimer's & Dementia supplement.
2025
Amyloid-induced neuronal hyperactivity and metabolism are associated with faster tau accumulation in Alzheimer's disease
Alzheimer's & Dementia, vol 21, e099685
AAIC 2025 conference abstract, published in the Alzheimer's & Dementia supplement.
Get in touch
Open to collaboration
I work across imaging analysis, research infrastructure and clinical science. If you are working on imaging biomarkers, trial imaging or getting an analysis pipeline into practice, I would be glad to hear from you.