DosimetrEYE: A GenAI-Based Tool for 3D Preclinical Dosimetry from 2D Dynamic Imaging

Organisations involved

Main Participant: BIOEMTECH is an SME in the field of biomedical engineering, offering hardware and software solutions for medical imaging and dosimetry applications.

Technology / XAI expert: Alethia is a newly established start-up with expertise in AI monitoring, explainability, and governance solutions.

HPC Expert: GRNET provides networking and cloud computing services to a broad range of public and private sectors.

The challenge

Biomedical research still relies heavily on animal studies, with more than 100 million animals used worldwide each year. Drug-delivery and toxicology studies require repeated imaging or sacrifice of animals to measure how radiopharmaceuticals behave inside the body. Existing methods for calculating absorbed dose are slow, imprecise, or depend on post-mortem analysis. This lack of accuracy increases costs, limits scientific insights, and prevents researchers from reducing the number of animals used.

BIOEMTECH aimed to change this by offering real-time, image-based dose estimation within its preclinical imaging systems. Achieving this required a major leap forward: the objective was to generate high-quality training data from complex Monte Carlo simulations and developing a generative artificial intelligence model able to infer 3D dose maps directly from 2D scans. These tasks require levels of computing power far beyond what the business can host or afford. Consequently, access to large-scale high-performance computing resources through FFplus was essential to produce the necessary data, train the model, and validate the new product.

The Solution

BIOEMTECH is a biomedical engineering SME that develops imaging systems and dosimetry software for pre-clinical research. ALETHIA is a start-up specialising in trustworthy artificial intelligence, monitoring, and explainability. GRNET provides high-performance computing services and coordinates the National Competence Centre Greece. Together, the partners combined domain expertise, artificial intelligence skills, and computing capability to build a new real-time dosimetry tool.

BIOEMTECH, ALETHIA, and GRNET created the first AI-driven tool that generates real-time three-dimensional dosimetry from routine pre-clinical imaging. The partnership produced a large dataset of ground-truth dose maps by running extensive Monte Carlo simulations on high-performance computing systems, covering around one thousand imaging cases. Using this dataset, they trained a generative model to estimate dose directly from two-dimensional scans in seconds rather than hours. The workflow also includes automated detection of unusual or low-quality inputs, using explainable-AI methods from ALETHIA to ensure reliability and support industrial use. Running computational workloads on high-performance computing was essential for both the simulations and training of the generative model.

Impact 

DosimetrEYE adds a new real-time dosimetry feature to BIOEMTECH’s imaging systems, reducing analysis time from hours to seconds and lowering per-study costs. The enhancement strengthens the product portfolio, enables new service offerings, and supports collaborations with scanner vendors. It is expected to increase annual revenues and improve the competitiveness of European medical-software SMEs. ALETHIA validated its artificial-intelligence monitoring platform in a real industrial setting, while GRNET expanded its role as a trusted computing partner.

The tool supports safer and more personalised medicine by enabling precise organ-level dose assessments that can be translated from preclinical to clinical workflows. Finally, more accurate image-based dose estimates significantly reduce the number of animals required for pre-clinical studies, helping research organisations to meet ethical and sustainability goals.

Benefits

• Real-time 3D dose estimation is now achieved in under 20 seconds, replacing lengthy workflows.
• Up to 60% reduction in analysis costs and major acceleration of drug-delivery studies.
• Expected 60% increase in BIOEMTECH’s dosimetry service volume, generating around €800,000 in new revenue.
• <80% reduction in animal use by replacing post-mortem analysis with accurate imaging-based dose prediction.
• ALETHIA validated its explainable-AI platform in a demanding industrial use case.