Researchers in Spain have developed an AI-driven approach that could dramatically reduce the time required for advanced brain MRI scans while maintaining high diagnostic accuracy. The technique, which combines AI with physics-based computer simulations, may help make sophisticated neuroimaging faster, more accessible and more practical for routine clinical use.
The research was carried out by scientists at the Institute for Neurosciences, a joint centre of the Spanish National Research Council and Miguel Hernández University of Elche. The findings were published in Communications Medicine. According to the researchers, the method can reduce data acquisition requirements by up to 90 percent, potentially shortening certain advanced MRI procedures from around 40 minutes to approximately eight minutes.
AI training in medical imaging
Most AI systems used in medical imaging are trained on large datasets collected from real patients. While effective, this approach often depends on extensive clinical data collection and raises issues related to privacy, availability and dataset bias. The Spanish research team adopted a different strategy. Instead of relying primarily on patient data, they generated synthetic training datasets using computer simulations based on the physical principles of water diffusion in brain tissue.
These simulations were then used to train neural networks capable of reconstructing detailed information about brain microstructure from only a small fraction of the MRI measurements normally required. Professor Silvia De Santis, who leads the Translational Imaging Biomarkers Laboratory at the institute, says reducing scan acquisition time could enable hospitals to incorporate more advanced imaging techniques into everyday clinical workflows.
Researcher Maximilian Eggl adds that simulation-based training also allows scientists to generate virtually unlimited datasets without depending on patient recruitment or sensitive personal data.
Faster MRI with minimal data
The methodology focuses on diffusion-weighted MRI, an advanced imaging technique that analyses the movement of water molecules through brain tissue. These signals provide valuable insights into tissue microstructure and neurological integrity. Traditionally, diffusion MRI requires a large number of measurements and relatively long scanning times to achieve sufficient accuracy. The newly developed AI system, however, was able to reconstruct highly detailed tissue information using only 10 percent of the usual data volume.
According to the researchers, this reduction could significantly improve hospital efficiency, particularly in healthcare systems facing growing imaging demand and long waiting lists. Shorter scans may also benefit patients directly. MRI procedures can be physically uncomfortable for some individuals, particularly older patients, people with cognitive disorders or children who struggle to remain still during lengthy imaging sessions.
Neurodegenerative disease detection
Beyond improving workflow efficiency, the researchers believe the technology could support earlier detection of neurodegenerative diseases such as Alzheimer's disease. Many neurological disorders develop gradually over decades before symptoms become clinically apparent. According to De Santis, current diagnostic approaches still rely heavily on imaging methods developed several decades ago, while translating newer research advances into clinical practice remains difficult.
By enabling more detailed imaging within shorter examination times, the AI-driven method may help clinicians identify subtle structural brain changes earlier in the disease process. The researchers also note that the technology creates new opportunities for retrospective analysis of older MRI datasets. Brain scans acquired years or even decades ago using older imaging protocols could potentially be reprocessed with the new AI models to extract additional information that was previously inaccessible.
More accessible neuroimaging
The study highlights a broader shift in medical imaging toward AI-assisted, data-efficient diagnostics. Rather than simply accelerating image interpretation, the new approach changes how imaging data itself is acquired and reconstructed. If validated further in clinical settings, the method could help hospitals increase imaging capacity, reduce costs and expand access to advanced neuroimaging techniques.
The researchers believe simulation-driven AI may ultimately become an important tool in precision neurology, supporting faster diagnosis, longitudinal monitoring and more personalised treatment strategies for patients with neurological disorders.
Cutting MRI waiting list
Two years ago, th Dutch hospital ‘Zaans Medisch Centrum’ (ZMC) succeeded in eliminating MRI waiting lists by installing a second MRI scanner and introducing AI-powered image reconstruction software. The extra scanner doubled their scanning capacity and with the introduction of Deep Resolve AI-software they also were able to reduce the average MRI scan time by up to 33%.
The AI -oftware also allows radiologists to choose between faster scans or higher image resolution. In brain imaging, ZMC opted for improved resolution, enabling detection of smaller abnormalities that may previously have gone unnoticed. The hospital says the combination of increased capacity and AI support has improved workflow efficiency and patient access to MRI diagnostics.
