Project Description & Aims

Overview

Inspired by the groundbreaking impact of the Human Genome Project, the Human Radiome Project (THRP) aims to revolutionize medical imaging by developing an advanced AI-powered foundation model capable of deciphering the intricate complexities of 3D radiological data. Funded by the Helmholtz Association's Foundation Model Initiative (HFMI), THRP seeks to create a comprehensive digital framework that seamlessly integrates structural, functional, and pathological information.

Overview Image

Approach

At its core, THRP focuses on developing a foundation model using self-supervised learning on an unparalleled dataset of over 4.8 million 3D radiological image volumes. This dataset, sourced from leading clinical data repositories at the German Cancer Research Center (DKFZ) and partner hospitals in Bonn, Heidelberg, and Basel, combines extensive population studies with numerous public datasets. The project also engages with other Helmholtz centers for cross-disciplinary applications, such as oncologic risk prediction at DKFZ, neuroradiological disease analysis at DZNE, and preclinical studies at MDC.

Impact

In addressing the challenges of modern medical imaging—an ever-increasing demand driven by aging populations, lower imaging costs, and a shortage of clinical experts—THRP offers innovative solutions to manage the growing complexity of 3D imaging data. Long term, the project envisions an interactive and generalist medical AI that seamlessly integrates diverse data modalities, paving the way for breakthroughs in genomics, phenotyping, and personalized healthcare.

Network

The project engages multiple Helmholtz centers for cross-disciplinary applications:

  • Oncologic risk prediction at DKFZ
  • Neuroradiological disease analysis at DZNE
  • Preclinical studies at MDC

If you want to become part of THRP by providing datasets or collaborating, please reach out to us.