Ability of Radiomics in Differentiation of Anaplastic Oligodendroglioma From Atypical Low-Grade Oligodendroglioma...

Area of Science:

• Neuro-oncology
• Medical imaging analysis
• Artificial intelligence in medicine

Background:

• Distinguishing anaplastic oligodendroglioma (AO) from atypical low-grade oligodendroglioma is clinically significant for treatment planning.
• Accurate differentiation is challenging based on conventional imaging alone.

Purpose of the Study:

• To evaluate the efficacy of radiomics features extracted from MRI in differentiating AO from atypical low-grade oligodendroglioma.
• To develop and assess machine-learning models for this diagnostic task.

Main Methods:

• Retrospective analysis of 101 patients with AO or atypical low-grade oligodendroglioma.
• Extraction of 40 radiomics features from contrast-enhanced T1-weighted (T1C) and FLAIR MRI sequences.
• Application of feature selection methods (distance correlation, LASSO, GBDT) and machine-learning classifiers (LDA, SVM, RF).

Main Results:

• Nine predictive models were developed, all demonstrating good differentiation ability (AUC > 0.840).
• The combination of LASSO and Random Forest (RF) on T1C images yielded the highest AUC of 0.904.
• The combination of GBDT and RF on FLAIR images achieved an AUC of 0.861.

Conclusions:

• Radiomics combined with machine learning shows promise as a non-invasive method for distinguishing AO from atypical low-grade oligodendroglioma.
• This approach may aid in improving diagnostic accuracy and guiding clinical management decisions.