ImagingGenesis

Short description of the innovation study

ImagingGenesis aims to develop a generative AI model for medical image analysis, specifically utilising latent diffusion models (LDMs) to address challenges in organ-specific CT image segmentation and reconstruction. This software leverages HPC resources to manage the computational intensity required for training advanced AI models.

The goal is to demonstrate that LDMs can effectively mitigate the challenges posed by limited or incomplete datasets. By generating high-quality synthetic medical images, the diffusion model will complement existing datasets, enhancing the accuracy of future AI models for specific organs. This will lay the groundwork for improving CT imaging AI capabilities across medical applications, proving that with adequate computational resources, insufficient data can be overcome.

The success of this project will significantly reduce dependency on large datasets, making AI-driven medical innovations more accessible and efficient.

Organisations involved: 

End User: Better Medicine

HPC provider: HPC at the University of Tartu, part of LUMI

Medical expert: Pärnu Hospital

Technology expert: University of Tartu