Qualitative and Quantitative MRI Analysis in IDH1 Genotype Prediction of Lower-Grade Gliomas: A Machine Learning...

Area of Science:

• Neuro-oncology
• Radiology
• Machine Learning in Medicine

Background:

• Accurate preoperative prediction of isocitrate dehydrogenase 1 (IDH1) mutations is vital for managing lower-grade gliomas (LGGs).
• IDH1 mutation status influences treatment strategies and patient prognosis in LGGs.
• Current diagnostic methods may require invasive procedures.

Purpose of the Study:

• To evaluate a machine learning approach for predicting IDH1 mutations in LGGs.
• To assess the predictive value of qualitative (VASARI) and quantitative (radiomics) MRI features.
• To determine if combining feature sets enhances prediction accuracy.

Main Methods:

• A cohort of 102 LGG patients was divided into training and validation sets.
• Visually Accessible Rembrandt Images (VASARI) and radiomics features were extracted from multimodal MRI and ADC maps.
• Feature selection was performed, and a random forest classifier was trained to predict IDH1 mutation status.

Main Results:

• Optimal VASARI and radiomics features were identified, including enhancement quality, tumor invasion, location, necrosis, T1/FLAIR ratio, and various radiomic descriptors.
• The VASARI-only model achieved an AUC of 0.779, and the radiomics-only model achieved an AUC of 0.849 on the validation cohort.
• A fusion model integrating both VASARI and radiomics features yielded the highest AUC of 0.879, demonstrating superior predictive performance.

Conclusions:

• Machine learning models utilizing MRI-derived VASARI and radiomics features can effectively predict IDH1 mutations in LGGs.
• The combined approach offers a non-invasive method for preoperative IDH1 mutation status determination.
• This predictive capability can aid in clinical decision-making for LGG patients.