(P21-009-24) Predicting Glioma Risk: A Population-Based Case-Control Study Using Machine Learning and Multifactorial Data Analysis

Objectives: Gliomas are diffusely infiltrative brain tumors with multiple subtypes. Due to the highly malignant nature of tumor and poor prognosis, prevention is of high priority. The relationship between glioma risk and factors such as diet, lifestyle, and demographics remains poorly understood. We used various machine learning techniques to develop predictive models for glioma risk, utilizing data on diet, lifestyle, and demographic factors.

Methods: This population-based case-control study was done 128 newly diagnosed glioma cases and 256 age- and sex-matched controls. Data on life style, demographic characteristics and dietary intakes were obtained. In this study, the analyzed dataset comprised 384 entries and 269 features. The Light Gradient Boosting Machine (LGBM) framework was employed, and optimized hyperparameter sets were obtained through Bayesian Optimization. Consequently, five sets of hyperparameters were utilized to create five distinct LGBM models, implemented in combination with Cross-Validation - RepeatedKFold with five splits and 10 repeats, forming the so-called Megamodel.

he log loss served as the evaluation metric, and KNN and XGB models were also explored. While each model has its own strengths and weaknesses, the LGBM framework is considered more promising for handling features, feature importance, and computation. For the test dataset, the log loss of the LGBM framework was 0.2788, with an AUC of 0.9599.

Two distinct methods were employed to calculate feature importance within the LGBM framework: LGBM Feature Importance and SHAP. Feature Importance quantifies the contribution of a specific feature to the model's predictions.

Results: According to LGBM Feature Importance, the top five features contributing to the glioma prediction model for the absence of glioma, in descending order, were the consumption of bananas, pomegranate, soy, strawberries, and cabbage.

Meanwhile, when assessing feature importance through SHAP, the top five features for the absence of glioma, in descending order, were associated with the consumption of cabbage, soy, strawberries, pomegranate, and refined grain.

Conclusions: We demonstrated in principle the potential use of machine learning to identify items for potential evaluation for the prevention of glioma.

Funding Sources: No funding