HEREDITARY project lays foundation for European AI research platform

Monday, June 17, 2024

The European HEREDITARY project, which started in January 2024, has great ambitions - such as building a federated learning platform to support medical-scientific research. However, the project firsy needs a good foundation. This is what Francesco Ciompi (Associate Professor of Pathology), Nils Kohn (Associate Professor of Human Cognitive Neuroimaging) and Annemarie Boleij (Associate Professor, Department of Pathology) state in an interview with ICT&health. They are involved in HEREDITARY on behalf of Radboudumc. Their contribution to HEREDITARY must help build the aforementioned foundation.

The various HEREDITARY studies examine various brain diseases, including ALS, MS, Parkinson's, depression, anxiety, ADHD and Alzheimer's. But also, intestinal diseases such as inflammatory diseases and intestinal infections. The researchers hope that HEREDITARY will open new avenues for the treatment and diagnosis of these conditions. AI technology should support this to better understand and treat the complex relationship between brain and gut health.

Strong influence

The Radboud project focuses on connecting and analyzing the enormous data streams that come with such studies. It focuses on disorders of gut health and the brain, because these organs strongly influence each other. The analyses will look for patterns in the data, such as the simultaneous occurrence of certain disorders. This is done with algorithms that can learn from large amounts of data.

According to Ciompi, an important ambition is to facilitate multimodal research through AI applications - with multiple modalities of data sources. This also includes the study of the interaction between the brain, the intestinal system and disorders such as inflammatory bowel diseases and intestinal inflammation.

“The hospital has pathology and radiology images, text-based reports, genetic sequences, MRI scans from neurosurgery. A wide variety of data sources that, when combined, provide a more complete picture of a patient and their condition. This often starts with a few modalities, such as a scan and a biopsy of a breast lesion. This way, you can determine the type of breast cancer and what the best treatment is. In this project, the focus is on bundling information based on data about the brain and the intestinal system. AI serves as an overarching technology to link these modalities and the questions we have.”

Role of AI

“In neuroscience, the problem is that the collected data is very complex,” adds Nils Kohn. “Various parts of the brain have different roles and functions. The same applies to the microbiome in the intestinal system. We have known for some time that the brain and intestinal system influence each other, but we struggle with the additional complexity that bundling this multimodal data entails. AI applications can help us to visualize this multimodality more easily and to provide more insight into the mutual influence. AI will not be the only technology that can help with this, but it is a very promising one.”

The primary goal of the Radboudumc researchers is to bring together two modalities, or data sources, and to gain new insights from this bundling, says Annemarie Boleij.

"Then we gradually expand the complexity. Otherwise, the complexity becomes too great, also because we have little previous research to base ourselves on. We started with pathology images and data on bacterial infections. Then we add genetic data about the bacteria to see how this affects the AI model developed up to that point. As a third step, we will add data from FMRI images."

The entire interview was published in ICT&health 3 on June 14.