.jpg)
Six research institutions, including the University of Valencia (UV), have developed a new procedure that will enable brain magnetic resonance imaging (MRI) scans to be carried out more easily, more rapidly and with higher quality, thereby facilitating the diagnosis of neurological disorders. The study, published in Imaging Neuroscience, will allow lesions associated with conditions such as Alzheimer’s disease, multiple sclerosis or brain tumours to be highlighted with great clarity.
This new method is based on a 3D deep neural network that generates T2-weighted images — which are highly sensitive to the presence of water and therefore useful for detecting oedema, inflammation or ischaemia — from T1-weighted images, which provide a detailed anatomical representation of the brain and allow clear differentiation between white and grey matter. In this way, T1 images contribute the structural information, while T2 and FLAIR images reveal potential pathological alterations.
The research was coordinated by José V. Manjón, from the Medical Image Analysis (MIA-LAB) group at the ITACA Institute of the Polytechnic University of Valencia (UPV), with the participation of the Department of Psychobiology (UV) and the Department of Applied Mathematics (UPV), as well as the La Fe University and Polytechnic Hospital, the Foundation for the Promotion of Health and Biomedical Research of the Valencian Community (FISABIO-CIPF), the French National Centre for Scientific Research, and the University of Bordeaux.
The Department of Psychobiology at the UV worked jointly with medical informatics specialists to provide insight into the neuropathological substrate, ensuring correct labelling of the images to be processed by the algorithms. Marien Gadea, a researcher in the department specialising in neuropsychology and neuroimaging, explains: “This method uses semi-supervised learning techniques in which the pre-labelling of the images used by the model is subjected to expert evaluation”.
The system integrates prior anatomical information and employs supervised learning techniques — an artificial intelligence approach that combines a small number of medically annotated images with a large volume of unlabelled images. This enables powerful models to be trained without the need for fully annotated databases.
Medical staff, neurobiologists, biologists and anatomists play a vital role in decisions regarding image labelling through the algorithms because they “help clarify reasonable doubts based on real anatomical and clinical practice”, adds Marien Gadea, an expert in brain–behaviour relationships. This gives the system a solid grounding in the study and investigation of those aspects of the central nervous system (CNS) that have the greatest functional relevance.
The method combines real anatomical knowledge, specific training strategies and a semi-supervised approach that enhances its ability to generalise across different patients and scanners. In brain segmentation tests, it outperformed the most advanced techniques available, even in complex cases such as brains with lesions or high anatomical variability. Moreover, it generates results within seconds, which facilitates its application in hospital settings.
“In an MRI scan, each type of image provides different information about the brain, but obtaining them all prolongs the test, increases the costs and may be uncomfortable for patients. Our system can generate the missing images from those already acquired, reducing both time and resources”, explains Sergio Morell, lead author of the study.
The next challenge for the team is to extend the technique to other sequences such as FLAIR (fluid-attenuated inversion recovery), a variant of T2-weighted imaging that suppresses the signal from cerebrospinal fluid and allows lesions associated with conditions such as Alzheimer’s disease, multiple sclerosis or brain tumours to be highlighted with great clarity.
This study was funded by the Spanish Ministry of Science, Innovation and Universities and by the French National Research Agency. It forms part of the project Development of Ultra-High-Resolution Multimodal Brain MRI Image Analysis Software for Clinical Application, co-led by Marien Gadea in collaboration with José V. Manjón and funded by the Ministry of Science and Innovation and the Spanish State Research Agency (PID2023-152127OB-I00).
Reference: Sergio Morell-Ortega, Marina Ruiz-Perez, Marien Gadea, Roberto Vivo-Hernando, Gregorio Rubio, Fernando Aparici, Mariam de la Iglesia-Vaya, Thomas Tourdias, Boris Mansencal, Pierrick Coupé, José V. Manjón; Robust deep MRI contrast synthesis using a prior-based and task-oriented 3D network. Imaging Neuroscience 2025; 3 IMAG.a.116. doi: https://doi.org/10.1162/IMAG.a.116
