Two researchers at the Institute for Neurosciences (IN), a joint centre of the Spanish National Research Council (CSIC) and the Miguel Hernández University of Elche (UMH), have developed a new strategy based on artificial intelligence and computer simulations that makes it possible to obtain detailed brain information more quickly from MRI scans using far less data than usual. The method, published in the journal Communications Medicine , can reduce the time required for certain advanced MRI scans by up to 90% while maintaining a high level of accuracy, paving the way for more efficient and accessible neuroimaging in clinical settings.
The study proposes a shift in the way artificial intelligence is applied to neuroimaging. Instead of training models using real patient data, as is common in many current applications, the team used a model based on the physics of the diffusion process in brain tissue to generate simulations. These data are then used to train neural networks to estimate model parameters that serve as biomarkers indicating the state of the tissue using a very small number of resonance images.
“Reducing the acquisition time required makes it possible to incorporate much more advanced MRI techniques, resulting in a greater amount of clinical information available to medical staff”, explains researcher Silvia De Santis , who leads the Translational Imaging Biomarkers Laboratory at the IN CSIC-UMH .
This approach also reduces the biases associated with traditional clinical datasets. “Using simulations allows us to generate as much data as we need, without depending on patient availability and while avoiding privacy issues,” adds researcher Maximilian Eggl , who leads the AI-inspired Biomarkers of Brain Structure and Function research line at the IN CSIC-UMH .
Less scanning time, more information
The methodology relies on advanced MRI techniques such as diffusion-weighted MRI, which makes it possible to non-invasively study the movement of water in brain tissue and thereby obtain information about its microstructure. From these signals, artificial intelligence efficiently reconstructs detailed features of the brain tissue.
One of the study’s most significant findings is the drastic reduction in the number of measurements required. “We have shown that our network, trained entirely on simulations, can achieve a very high level of accuracy using only 10% of the data”, says Eggl. “This could have a direct impact in clinical settings, especially in hospitals with very long waiting lists”, adds the researcher.
In practice, this breakthrough could translate into a significant reduction in scan time: “Imagine going from about 40 minutes to roughly 8 to obtain the same information. This procedure would make it possible to increase the number of patients treated in the same amount of time and make the system much more efficient”, both researchers agree.
Towards early diagnosis in neurodegenerative diseases
This approach also opens up new possibilities in the study of neurodegenerative diseases such as Alzheimer’s, which present a very long preclinical phase of up to two decades, during which no visible symptoms appear. “The clinical diagnosis of degenerative diseases is still based on techniques developed more than 30 years ago, while incorporating advances generated in the laboratory remains a major challenge. This new approach would make it possible to obtain more detailed information and, in turn, improve the diagnosis of these diseases”, explains De Santis.
In addition, the system allows for the reanalysis of magnetic resonance imaging data acquired decades ago, which had previously been limited by the technologies available at the time. Thanks to this new simulation-based approach, these data can be reinterpreted to extract new relevant information about neurological diseases.
This work has been possible thanks to funding from the “laCaixa” Foundation; the Spanish State Research Agency (AEI) – Spanish Ministry of Science, Innovation and Universities; the Severo Ochoa Programme for Centres of Excellence; the grant from the Generalitat Valenciana for the recruitment of outstanding doctoral researchers (CIDEGENT 2021); and the Pasqual Maragall Researchers Programme (PMRP) of the Pasqual Maragall Foundation.
Communications Medicine
Computational simulation/modeling
Not applicable
). Simulation-based inference at the theoretical limit for fast, robust microstructural MRI with minimal diffusion data
1-May-2026