Related Experiment Video
Updated: Aug 26, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.9K
Differentiating Glioblastoma Multiforme from Brain Metastases Using Multidimensional Radiomics Features Derived from
Salar Bijari1, Amin Jahanbakhshi2, Parham Hajishafiezahramini3
1Department of Medical Physics, Faculty of Medical Sciences, Tarbiat Modares University, Tehran, Iran.
Biomed Research International
|October 10, 2022
Summary
Differentiating glioblastoma (GBM) from brain metastases (MET) is crucial for treatment. This study introduces a radiomics analysis using routine MRI sequences and wavelet transform, achieving high accuracy in distinguishing between GBM and MET.
Area of Science:
- Neuroradiology
- Medical Imaging Analysis
- Machine Learning in Oncology
Background:
- Distinguishing glioblastoma multiforme (GBM) from brain metastases (MET) is critical for patient treatment due to differing therapeutic strategies.
- Magnetic resonance imaging (MRI) is vital for brain tumor detection, characterization, and monitoring, but differentiating GBM and MET based solely on imaging can be challenging.
- Definitive diagnosis often requires invasive surgical methods, highlighting the need for non-invasive, accurate diagnostic tools.
Purpose of the Study:
- To develop and validate an accurate, convenient, and user-friendly method for differentiating between GBM and MET using routine MRI sequences and radiomics analysis.
- To evaluate the efficacy of machine learning (ML) models combined with wavelet transform for improved diagnostic performance.
Main Methods:
- A retrospective study included 91 patients (50 GBM, 41 MET) with pathologically confirmed diagnoses.
- Tumors were segmented across T1-weighted imaging (T1WI), contrast-enhanced T1WI (T1C), T2WI, and FLAIR sequences to create volumes of interest (VOI).
- Eight ML models were evaluated with and without multidimensional discrete wavelet transform using radiomics features extracted from routine MRI sequences, with model selection based on accuracy, AUC-roc, and F1-score.
Main Results:
- The optimal model, utilizing radiomics with wavelet transform, achieved an accuracy of 0.98, AUC-roc of 0.99, and F1-score of 0.98.
- Significant improvements in performance were observed in models incorporating multidimensional wavelet transform compared to those without.
- Multidimensional discrete wavelet transform effectively analyzed subtle MRI features, generating accurate predictors for model performance.
Conclusions:
- Routine MRI sequences combined with radiomics and multidimensional discrete wavelet transform provide a highly accurate method for differentiating GBM from MET.
- This non-invasive approach offers a valuable tool to aid clinical decision-making, potentially reducing the need for invasive diagnostic procedures.
- The integration of advanced image analysis techniques like wavelet transform enhances the diagnostic power of conventional MRI in neuro-oncology.
