Abstract
Purpose: We aimed to analyze $^{18}$F-fluorodeoxyglucose positron emission tomography ($^{18}$F-FDG PET) images via the radiomic method to develop a model and validate the potential value of features reflecting glioma metabolism for predicting isocitrate dehydrogenase (IDH) genotype and prognosis. Methods: PET images of 127 patients were retrospectively analyzed. A series of quantitative features reflecting the metabolic heterogeneity of the tumors were extracted, and a radiomic signature was generated using the support vector machine method. A combined model that included clinical characteristics and the radiomic signature was then constructed by multivariate logistic regression to predict the IDH genotype status, and the model was evaluated and verified by receiver operating characteristic (ROC) curves and calibration curves. Finally, Kaplan-Meier curves and log-rank tests were used to analyze overall survival (OS) according to the predicted result. Results: The generated radiomic signature was significantly associated with IDH genotype ($p < 0.05$) and could achieve large areas under the ROC curve of 0.911 and 0.900 on the training and validation cohorts, respectively, with the incorporation of age and type of tumor metabolism. The good agreement of the calibration curves in the validation cohort further validated the efficacy of the constructed model. Moreover, the predicted results showed a significant difference in OS between high- and low-risk groups ($p < 0.001$). Conclusions: Our results indicate that the $^{18}$F-FDG metabolism-related features could effectively predict the IDH genotype of gliomas and stratify the OS of patients with different prognoses.
Introduction
Gliomas represent the most prevalent primary malignant brain tumors. A critical factor in determining patient prognosis is the Isocitrate Dehydrogenase (IDH) genotype. Patients harboring IDH-mutant tumors typically experience significantly better survival outcomes compared to those with IDH wild-type tumors. Currently, identifying this genotype requires invasive neurosurgery or biopsy for tissue analysis. This study explores a non-invasive "virtual biopsy" alternative using radiomics to predict genetic makeup and prognosis from standard medical imaging.
Methods and Workflow
The researchers utilized a radiomics pipeline to transform $^{18}$F-FDG PET scans into high-dimensional quantitative data. The process involved three primary stages: data preparation (converting PET to SUV images and segmenting the tumor), radiomic analysis (feature extraction, selection via Elastic Net, and SVM classification), and results display (validation via ROC and survival curves).
From manually segmented 3D tumor regions, 1,561 features were extracted, including shape, intensity statistics, and complex textures. These were eventually narrowed down to the 11 most predictive features using the Elastic Net method to prevent overfitting.
Datasets
The study retrospectively analyzed 127 consecutive patients diagnosed with primary glioma at the Peking Union Medical College Hospital:
- Training Cohort: 84 patients (38 IDH-mutant, 46 wild-type).
- Validation Cohort: 43 patients (13 IDH-mutant, 30 wild-type).
Key Insights and Ideas
- Metabolic Heterogeneity: Simple metrics like SUVmax were insufficient for genotype prediction. Instead, complex texture features describing metabolic heterogeneity proved to be the strongest predictors.
- The Sphericity Factor: A notable finding was that IDH-mutant gliomas often exhibit higher metabolic sphericity, meaning their metabolic activity is more regularly shaped and less fragmented than wild-type tumors.
- Model Synergy: A combined model integrating the radiomic signature with clinical factors (age and tumor metabolism type) significantly outperformed models using either data source in isolation.
Key Results
- High Predictive Accuracy: The combined model achieved an Area Under the Curve (AUC) of 0.911 in the training set and 0.900 in the independent validation set.
- Sensitivity: In the validation cohort, the model demonstrated a high sensitivity of 92.3% for identifying IDH mutants.
- Prognostic Stratification: Kaplan-Meier analysis confirmed a statistically significant difference ($p < 0.001$) in overall survival between high-risk and low-risk groups as categorized by the model's predictions.
Limitations and Future Work
While promising, the study was limited by its single-center, retrospective design. Future research should focus on multi-center validation across different PET scanner protocols and the potential integration of MRI data to enhance anatomical detail. Additionally, further work is needed to map mathematical texture features back to specific biological processes within the tumor cells.
Summary updated