GLIOMAID is a scientific research project focused on improving how brain tumors, specifically gliomas, are diagnosed and managed. It uses Artificial Intelligence (AI) to analyze MRI brain scans and patient data. The project collects existing clinical information and imaging from glioma patients to build AI models that support doctors in making better and faster treatment decisions.Gliomas, especially high-grade ones, are among the most common and challenging brain tumors. Many patients have poor survival chances, and diagnosis often requires invasive procedures like biopsies.

Despite medical advances, current treatments have limited effectiveness. Better non-invasive diagnostic tools are urgently needed to:

• Detect tumors earlier.
• Predict how aggressive they are.
• Help doctors plan the most effective treatments. The GLIOMAID study aims to reduce the need for invasive diagnostics by creating AI tools that interpret brain scans with high accuracy.

Primary Objectives

• Create Italy's First Glioma Imaging Database This database will store anonymized MRI scans and clinical records from around 700 patients.
• Improve Early Detection Develop AI systems to identify brain tumors earlier from MRI scans.
• Automate Tumor Mapping Use AI to outline tumors on MRI images to assist with surgical planning and treatment follow-up.
• Non-Invasive Tumor Characterization Train AI models to predict tumor type and severity without needing a biopsy.

Secondary Objectives

• Study how well AI tools fit into research and future clinical workflows.
• Test how well AI can predict changes in tumors over time.

Lead Institution: University of Trento and Santa Chiara Hospital, Trento (Prof. Silvio Sarubbo, Principal Investigator).

Partner Hospitals: 7 neurosurgery and neuro-oncology centers across Italy.

Inclusion Criteria

• Adults aged 18-60 with a confirmed glioma diagnosis (from 2019 to 2024).
• Patients who had surgical tumor removal, with or without further treatment (e.g., chemotherapy, radiotherapy).
• MRI scans and basic clinical data must be available.

Exclusion Criteria

• Poor quality or incomplete MRI scans.
• Missing essential clinical information.
• If consent is explicitly refused (when it can be obtained).

Clinical Data

• Age, sex, diagnosis date.
• Tumor type and genetic information.
• Treatments received (surgery, chemo, radiation).
• Patient outcomes (e.g., survival, tumor progression).

Imaging Data

• Pre- and post-operative MRI scans (T1, T2, FLAIR).
• Segmented images highlighting tumor areas and post-surgery cavities.
• Time points: before surgery, up to 6 months post-op, and during follow-up.

All data is pseudonymized (no personal identifiers) and securely stored.

Expected Results

• Faster, more accurate diagnosis.
• More personalized treatment planning.
• Reduced need for invasive biopsies.

Benefits for Patients and Doctors Patients: Earlier diagnosis, less invasive procedures, better treatment outcomes.

Doctors: Improved decision-making tools, automated image analysis, consistent data for treatment planning.

Inclusion Criteria:

• Imaging (MRI) of confirmed glioma diagnosis in the period 2019-2024, for whom cancer types and stages, from diagnosis to post-treatment are available
• Having undergone a full brain tumor resection operation, followed or not by treatment with RT or CHT
• Adults aged 18 to 60 years
• Informed consent available, when possible and applicable

Exclusion Criteria:

• Poor quality or artifact-laden MRI images
• Lack of a minimal set of clinical information
• Explicit refusal of consent (if possible to obtain)
• Age under 18 years or over 60 years

The study population consists of approximately 700 adult patients (100 per center) aged 18 to 60 years, diagnosed with brain glioma between 2019 and 2024 across seven specialized Italian neurosurgical centers. All participants underwent surgical tumor resection, with or without subsequent radiotherapy or chemotherapy. Only patients with high-quality MRI scans and essential clinical information are included. The study uses retrospective data, and where possible, informed consent is obtained. If consent cannot be collected due to patient death or unreachability, inclusion may still occur under ethically approved conditions. Data are pseudonymized and used to train and validate AI models.