An innovative AI-powered tool designed to predict regions of tumor infiltration and recurrence in glioblastoma using radiomic features extracted from multiparametric MRI.
🔍 Supports non-invasive tumor characterization
📊 Aims to guide personalized follow-up and treatment strategies
👉 Learn more and explore the tool: https://geibac.uva.es/ 🔗
🔐 License
GlioMap v2.0 is distributed under the Academic Non-Commercial Source Code License Agreement
© 2024 UiT – The Arctic University of Norway (Licensor)
This software is intended exclusively for research purposes and not for clinical decision-making.
🧠 Developed by the [GEIBAC Research Group] – Biomedical Imaging and Computational Analysis
🏥 In collaboration with the Neurosurgery Department of Río Hortega University Hospital and UiT – The Arctic University of Norway
💬 We welcome collaboration!
If you're a researcher, clinician, or student interested in tumor modeling, radiomics, or glioblastoma research, join us!
📧 Get in touch: [email protected]
Predicting regions of local recurrence in glioblastomas using voxel-based radiomic features of multiparametric MRI
This repository contains the Python implementation of the papers:
- Cepeda S, Luppino LT, Pérez-Núñez A, Solheim O, García-García S, Velasco-Casares M, Karlberg A, Eikenes L, Sarabia R, Arrese I, Zamora T, Gonzalez P, Jiménez-Roldán L, Kuttner S. Predicting Regions of Local Recurrence in Glioblastomas Using Voxel-Based Radiomic Features of Multiparametric Postoperative MRI. Cancers. 2023; 15(6):1894. https://doi.org/10.3390/cancers15061894 (https://www.mdpi.com/2072-6694/15/6/1894)
- Cepeda S, Luppino L, Wodsinki M, Solheim O, Pérez-Núñez A, García-García S, Karlberg A, Eikenes L, Zamora T, Sarabia R, Arrese I, Kuttner S. NIMG-45. EXTERNAL EVALUATION OF A MACHINE LEARNING MODEL EMPLOYING RADIOMICS TO IDENTIFY REGIONS OF LOCAL RECURRENCE IN GLIOBLASTOMA FROM POSTOPERATIVE MRI. Neuro-Oncology, Volume 25, Issue Supplement_5, November 2023, Pages v195–v196, https://doi.org/10.1093/neuonc/noad179.0741
- Cepeda S, Luppino L, Solheim O, Pérez-Núñez A, García-García S, Karlberg A, Eikenes L, Zamora T, Sarabia R, Arrese I, Kuttner S. Machine Learning-based Identification of Local Recurrence Regions in Glioblastoma using Postoperative MRI: Implications for Survival Prognostication. Brain and Spine Volume 3, Supplement 1, 2023, 101960. https://doi.org/10.1016/j.bas.2023.101960
This model uses as input the voxelwise radiomic features of the non-enhancing peritumoral region of glioblastomas extracted from multiparametric structural MRI. As output, the probability for each voxel of becoming a site of future tumor recurrence is obtained. The probabilities are represented through color-coded maps. In addition, a segmentation of the regions identified as high-risk by the model is generated.
Raw MRI sequences need to be pre-processed according to the following pipeline: https://github.com/smcch/Postoperative-Glioblastoma-Segmentation
After preprocessing, segmentation of the following structures is mandatory: a) peritumoral region, b) tumor core (enhancing volume + necrosis) or surgical cavity depending on whether it is a preoperative or postoperative study.
Below is an example of the volumes and segmentation that will be used as input, and the model's output (recurrence probability map).
In order to run this code, Python 3.9.15 or above is required. The Python packages listed in requirements.txt are also necessary. One can install them by running
pip install -r requirements.txt
A folder named Patients containing the patients' data to be analysed is expected to be placed in the same folder as the code. Patients should contain a subfolder for each of the patients. Each patient's folder must contain:
- t1ce.nii.gz
- t1.nii.gz
- t2.nii.gz
- flair.nii.gz
- adc.nii.gz
- peritumor.nii.gz
- tumor.nii.gz OR cavity.nii.gz
The main function is in main.py, and to run the code over all the patients, one must run
python main.py
from within the code folder. Once the script is done, each patient will have their results stored within their respective subfolder.