Related Experiment Video
Updated: Jan 10, 2026
09:53
Quantifying the Brain Metastatic Tumor Micro-Environment using an Organ-On-A Chip 3D Model, Machine Learning, and Confocal Tomography
Published on: August 16, 2020
7.6K
Electron Density and Effective Atomic Number as Quantitative Biomarkers for Differentiating Malignant Brain Tumors:
Tsubasa Nakano1, Daisuke Hirahara1,2, Tomohito Hasegawa1
1Department of Radiology, Kagoshima University Graduate School of Medical and Dental Sciences, 8-35-1 Sakuragaoka, Kagoshima 890-8544, Japan.
Tomography (Ann Arbor, Mich.)
|November 26, 2025
Summary
Dual-energy CT derived electron density and effective atomic number show promise for differentiating malignant brain tumors. These quantitative imaging biomarkers, combined with machine learning, can improve diagnostic accuracy for brain metastasis, glioblastoma, and CNS lymphoma.
Area of Science:
- Radiology
- Medical Imaging
- Oncology
Background:
- Accurate differentiation of malignant brain tumors is crucial for effective treatment planning.
- Conventional imaging techniques may have limitations in distinguishing between different types of primary and metastatic brain lesions.
- Quantitative imaging biomarkers offer potential for improved diagnostic performance.
Purpose of the Study:
- To investigate the utility of electron density (ED) and effective atomic number (Zeff) derived from dual-energy computed tomography (DECT) as quantitative imaging biomarkers.
- To evaluate the diagnostic performance of DECT parameters and relative apparent diffusion coefficient (rADC) in differentiating malignant brain tumors including brain metastasis (BM), glioblastoma, and primary central nervous system lymphoma (PCNSL).
- To develop and assess machine learning (ML)-based diagnostic models incorporating DECT and rADC for tumor classification.
Main Methods:
- Retrospective analysis of 136 patients with pathologically confirmed BM, glioblastoma, or PCNSL.
- Comparison of conventional CT values (CTconv), ED, Zeff, and rADC (from MRI) using 10th percentile, mean, and 90th percentile values within contrast-enhanced tumor regions.
- Development of ML diagnostic models using the AutoGluon-Tabular framework with patient-level data splitting into training, validation, and independent test sets.
Main Results:
- The 10th percentile of Zeff was significantly higher in glioblastomas compared to BMs (p=0.02).
- All indices (CTconv, Zeff, rADC) showed significant differences when PCNSLs were included (p<0.001-0.02).
- DECT-based ML models achieved high AUCs for pairwise tumor differentiation (0.82-0.91), with combined DECT and rADC models showing excellent performance (AUC=1.0 for BM vs. PCNSL, AUC=0.93 for Glioblastoma vs. PCNSL).
Conclusions:
- DECT-derived ED and Zeff show potential as novel quantitative imaging biomarkers for differentiating malignant brain tumors.
- ML models incorporating DECT parameters and rADC demonstrate high diagnostic accuracy for classifying brain tumors.
- These quantitative imaging approaches may enhance the non-invasive diagnosis of brain malignancies.
