Radiomics Features Predict Telomerase Reverse Transcriptase Promoter Mutations in World Health Organization Grade II...

Area of Science:

• Neuro-oncology
• Radiology
• Machine Learning

Background:

• Mutations in the telomerase reverse transcriptase promoter (pTERT) are crucial for glioma prognosis and surgical planning.
• Accurate preoperative diagnosis of pTERT status is essential for effective glioma management.

Purpose of the Study:

• To develop and evaluate a radiomics-based machine learning algorithm for predicting pTERT mutations in World Health Organization (WHO) grade II gliomas.
• To assess the algorithm's performance in distinguishing between wild-type and mutant pTERT statuses using multi-parametric MRI data.

Main Methods:

• Retrospective analysis of 164 patients with WHO grade II gliomas.
• Extraction of 1,293 radiomics features from multi-parametric MRI scans.
• Application of elastic net for feature selection and support vector machine with a linear kernel for prediction, evaluated using nested 10-fold cross-validation.

Main Results:

• A model utilizing 12 radiomics features was developed, achieving an Area Under the Curve (AUC) of 0.8446.
• The model demonstrated high sensitivity (0.9355) and an overall accuracy of 0.7988.
• Significant differences in posterior predictive probabilities of pTERT mutations were observed between wild-type and mutant groups.

Conclusions:

• Radiomics analysis combined with machine learning offers a valuable, non-invasive tool for predicting pTERT status in WHO grade II gliomas.
• This predictive capability can potentially assist clinicians in preoperative evaluation, prognosis assessment, and surgical strategy determination for glioma patients.