Alberto de Luca

UMC Utrecht

In short, Alberto de Luca's research entails the following:

Fast and accurate intraoperative reconstruction of brain tracts with AI to optimize efficacy of pediatric posterior fossa tumor surgery

Tumors of the posterior fossa account for nearly half of pediatric brain cancers. While less common, they also occur in adults. These tumors are associated with devastating outcomes, contributing to a 25% mortality rate among affected children. Brain surgery is the primary treatment for posterior fossa tumors, regardless of their aggressiveness. However, even when surgery successfully removes the tumor, it can lead to short-term adverse effects in approximately one-third of cases. One particularly impactful complication is cerebellar mutism syndrome (CMS), a condition where children lose the ability to speak for weeks to months, placing a significant emotional and practical burden on both patients and their families. 

CMS is believed to result from damage to specific brain connections. Intra-operative diffusion-MRI holds the potential to visualize these critical connections and guide surgeons in minimizing harm during the procedure. However, this application remains limited by the lack of fast and reliable methodologies compatible with the stringent time constraints of surgery and the suboptimal imaging quality of intra-operative MRI systems.

This project aims to address these challenges by developing innovative methods, including artificial intelligence techniques, to enable high-quality reconstructions of brain connections using intra-operative diffusion-MRI. We will evaluate the potential of these methods to support surgeries for children with posterior fossa tumors.

Publications

• Structural Fiber Tract Alterations in Relation to Surgery in Children With a Posterior Fossa Tumor
  • NMR in Biomedicine (20-02-2026)
  • Click here to view the publication
• The tractographer’s dilemma: understanding sources of variability in tractography
  • Brain Structure and Function (16-06-2025) 
  • Click here to view the publications

Short summary of progress/results February 2026

Over the past months, we have made concrete progress toward improving diffusion MRI — a technique used to map the brain’s wiring with AI methods. Because MRI scan time is limited in clinical practice, we are developing computational methods to enhance the quality of images within routine scanning time. In the future, these methods will be applied to support better mapping of brain connections during surgery.

During the first eight months of the project, we have carried out an in-depth review of existing approaches to enhance diffusion MRI resolution using artificial intelligence. Based on this analysis, we have implemented two promising AI models and trained them on high-quality brain imaging data from the Human Connectome Project. To support this work, we have also built a complete data-processing pipeline, including realistic simulations of lower-resolution scans. The first training results are encouraging and demonstrate that the proposed methods can meaningfully improve the level of detail in reconstructed brain fibre pathways. We have also established a new international collaboration with Professor Daniel Alexander’s group at University College London (UCL).

In parallel, we have focused on complementary strategies to enhance diffusion MRI encoding without increasing scan time. After reviewing the latest scientific developments, a comprehensive experimental plan was established and will be implemented in the next months.