Supervisor: Alessandra Bertoldo
Co-supervisor: Prof. Alessandro Salvalaggio, Prof. Renzo Manara
Co-supervisor Department/Company: Dipartimento di Neuroscience - DNS
Creation Date: 24/08/2026 16:38
Background and preliminary results
A previous study investigated the potential of Quantitative Susceptibility Mapping (QSM) to identify glioblastoma infiltration beyond the conventional MRI-defined tumor boundaries, particularly within peritumoral edema. Using pre- and post-operative MRI data from 20 glioblastoma patients, susceptibility alterations within the edematous region were analyzed to identify spatial patterns potentially associated with tumor infiltration. The predicted infiltration patterns were subsequently monitored longitudinally and supported by an independent voxel-wise analysis. In addition, their spatial relationship with white-matter pathways and tissue microstructural integrity was explored using standard tractography and Fractional Anisotropy atlases. Overall, these preliminary results suggest that QSM may provide a quantitative imaging marker of otherwise occult tumor infiltration, supporting its potential use for improved characterization of glioblastoma extent and longitudinal disease monitoring.
Proposed project
The main objective of the project is to develop a voxel-wise predictive model that uses pre-surgical QSM information to generate patient-specific maps of the probability of future tumor recurrence. These predictions will be validated against the spatial location of recurrence observed during longitudinal follow-up at 3, 6, and 9 months after surgery.
Dataset type: Already acquired data
Dataset description: Data for this study were retrieved from the clinical database of the Neuroradiology Unit at the University Hospital of Padova (AOPD). From this database, patients diagnosed with glioblastoma will be selected, including only those with complete available All patients underwent brain MRI scans at the Neuroradiology Unit between 2023 and 2026, following imaging protocols approved by the institutional ethical committee. Imaging was/are performed on a 3 T Philips Ingenia scanner equipped with a 32-channel headneck coil. Each of the 20 subjects in the longitudinal cohort have multiple timepoints (TP) with fully complete datasets, corresponding to routine MRI scans performed at specific stages of their glioblastoma treatment. Within the scope of this thesis, these timepoints were categorized according to the following chronological scheme: • TP0: Baseline scan (pre-surgery). • TP1: 1–2 months after TP0 (always not considered due to post-surgical artifacts). • TP2: 3–4 months after TP0. • TP3: 5–6 months after TP0. • TP4: 9 months after TP0.
List of Methods: Initial analyses will focus on voxel-wise and region-based characterization of pre-surgical QSM abnormalities associated with subsequent tumor recurrence. QSM-derived intensity, spatial, and radiomic features will then be used to develop predictive models generating patient-specific maps of recurrence probability. Model performance will be evaluated through longitudinal spatial validation against recurrence observed at follow-up MRI.
Imaging for Neuroscience, Metodi Statistici per la Bioingegneria