Related Experiment Video
Updated: Sep 19, 2025
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
Radiomics-Based Differentiation of Primary Central Nervous System Lymphoma and Solitary Brain Metastasis Using
Xueming Xia1, Jiajun Qiu2, Qiaoyue Tan3
1Division of Head & Neck Tumor Multimodality Treatment, Cancer Center, West China Hospital, Sichuan University, Chengdu, China (X.X., Q.G.).
Radiomics models using contrast-enhanced T1-weighted imaging can accurately differentiate primary central nervous system lymphoma (PCNSL) from solitary brain metastasis (SBM). The support vector machine with radial basis function (SVMRBF) classifier demonstrated the highest diagnostic performance in this non-invasive approach.
Area of Science:
- Neuroradiology
- Medical Imaging Analysis
- Machine Learning in Medicine
Background:
- Distinguishing primary central nervous system lymphoma (PCNSL) from solitary brain metastasis (SBM) is crucial for effective treatment planning.
- Accurate non-invasive differentiation remains a challenge in clinical practice.
- Radiomics offers a quantitative approach to extract imaging features for improved diagnostic capabilities.
Purpose of the Study:
- To develop and validate radiomics-based models for the non-invasive differentiation of PCNSL and SBM using contrast-enhanced T1-weighted imaging (CE-T1WI).
- To enhance diagnostic accuracy and aid clinical decision-making in neuro-oncology.
Main Methods:
- Retrospective analysis of 324 patients (115 PCNSL, 209 SBM) with pathological diagnosis.
- Manual segmentation of tumor regions on CE-T1WI and extraction of 1561 radiomic features.
- Feature selection using least absolute shrinkage and selection operator (LASSO) regression, followed by training and validation of machine learning classifiers.
Main Results:
- 23 significant radiomic features were identified via LASSO regression.
- Multiple classifiers achieved high performance, with 15 out of 20 exceeding an area under the curve (AUC) of 0.9.
- The support vector machine with radial basis function (SVMRBF) classifier achieved the highest AUC (0.9310) and accuracy (0.8780) on independent testing.
- Selected models demonstrated significant clinical utility, with standardized net benefits surpassing 0.6.
Conclusions:
- Radiomics models derived from CE-T1WI show significant potential for accurate non-invasive differentiation of PCNSL and SBM.
- The SVMRBF classifier exhibited superior diagnostic efficacy, highlighting its clinical utility in differential diagnosis.
