Healthcare is now evolving towards a computer system that learns from extensive medical data and offers personalized advice for patients. This could involve, for instance, comparing a patient's MRI scan with a database of scans and comprehensive medical histories from similar cases. The complexity of this system lies in handling various data types, including textual information, blood test results, medical imagery, and genetic data.

An international team of researchers, which includes investigators from the Radboud University Medical Center in Nijmegen, Netherlands, and backed by a EURO 11 million grant from the European Commission, is in the process of creating an artificial intelligence (AI) system. This AI is designed to provide insights into several brain and intestinal disorders such as depression, anxiety, and obesity, and to explore the interrelations between these conditions. The computer system, named Ciompi, will be capable of connecting and analyzing diverse types of medical data. The focus is on disorders related to the brain and intestines due to the significant interplay between these two organs, known as the "gut-brain axis." The system will look for patterns in this multimodal data, like the simultaneous presence of specific conditions or states.

A considerable amount of data is already available from earlier studies, including 20,000 digitized images of intestinal polyps and biopsies, data on intestinal bacteria, genetic information, and numerous MRI brain scans. The researchers plan to interlink these data sets, and the broader scope of the EU project includes examining factors like air pollution. The computer system will employ algorithms that learn from this pool of data. These algorithms will be housed on the Grand Challenge platform, renowned for hosting global competitions to develop superior algorithms for medical image analysis, like CT or MRI scans. This platform also supports hosting various algorithms and data types, accessible in different formats. Presently, the platform accommodates medical images and digital pathology slides, but the project aims to incorporate additional data types such as genetic information. The new algorithms will be integrated into this platform. However, not all the data used for training the system will be stored online.

Increasingly, 'federated' methods are being employed to both train AI algorithms and access data. For example, in federated learning, the algorithms virtually visit different hospitals via the platform, learning directly from the medical data on-site, without the need to transfer the data out of the hospital. Once the algorithms have sufficiently learned from these virtual visits, they can then aid doctors in the future. For instance, Ciompi will be able to compare a patient's diverse data from the gut and brain, such as fMRI scans, intestinal biopsies, and metabolome sequences from fecal samples, with scans and medical records from similar cases. This system can then assist healthcare providers in diagnosing, predicting outcomes, identifying potential connections with other conditions, and recommending treatment strategies. 

http://www.gzjiayumed.com/en/index.asp