Dr. Amir Dehsarvi

Dr. Amir Dehsarvi

I am the Principal Scientist of Neuroimaging at Boehringer Ingelheim. My research focuses on the application of machine learning and deep learning to medical data, with a strong emphasis on leveraging neuroimaging to accelerate disease diagnosis and therapeutic target identification.

ADPREP – A Fully-Automated Software for Large-scale Multimodal MRI and PET Imaging Workflows

Date:

For my talk at the Alzheimer’s Association International Conference (AAIC) 2025, I had the honor of chairing the session on “Imaging in neurodegenerative diseases.” During this session, I presented our work on a groundbreaking neuroimaging toolbox. I also presented this work as a poster at the conference.

My talk was all about ADPREP, a fully automated and containerized neuroimaging toolbox we developed to solve a major problem in Alzheimer’s disease research: the lack of standardized data. Processing large-scale multimodal MRI and PET data is a huge hurdle. It requires extensive expertise in multiple software packages and programming languages, and different labs use different methods, which leads to inconsistent results.

I presented how ADPREP simplifies this process, making it accessible to researchers without deep programming knowledge. We designed it to be user-friendly while maintaining state-of-the-art robustness. It can handle all the key data types—structural and functional MRI, along with multi-tracer PET (amyloid, tau, FDG, and TSPO)—and produces standardized outputs in a fraction of the time.

To show how effective ADPREP is, I shared some impressive results. We successfully tested it on thousands of scans from major datasets like ADNI and A4, achieving a processing failure rate of less than 4%. The best part? When we benchmarked our results against existing pipelines, we found that ADPREP’s outputs for amyloid-PET and tau-PET were nearly identical, with correlation coefficients of 0.99 and 0.98, respectively. This means our tool is not only reliable but also comparable to the gold standard.

Ultimately, my presentation highlighted how ADPREP can foster harmonization and data sharing across the AD neuroimaging community. It’s an essential step toward a future where our collective data can be used more effectively to drive discovery and innovation. I also announced its full integration into the cloud-based GRIP platform, making it even more accessible for large-scale data processing.