Methodological Research on Radiogenomics: Combining Radiomics and Genomics for Classification

Turkiye Klinikleri Journal of Biostatistics | , Vol 16(1)

DOI

Objective: Radiogenomics investigates the use of radiomics, and genomics features in clinical decision-making. The purpose of this study is to classify a clinical outcome by using radiomics and genomics features. The performances of different classification methods are compared and the effect of feature selection on classification performance is investigated.

Material and Methods: Non-small cell lung cancer dataset from The Cancer Imaging Archive was used. The type of histology was selected as binary clinical outcome for classification. This dataset contains computed tomography images and RNA-sequence gene expressions. To standardize features, z scaling was applied to radiomics features and logarithmic transformation was applied to genomics features. Data was divided into 70% train set and 30% test set. Classification was carried out by modeling only radiomics features, only genomics features, and radiomics and genomics features together. Elastic net, random forest, support vector machines, and XGBoost algorithms were used for classification. Different feature selection approaches were explored to see the effect of feature selection on classification. Performance measures were calculated by using the test set.

Results: The use of radiomics and genomics features improved the classification performance of random forest and XGBoost when feature selection was either not applied or when AUC was used for the feature selection method and elastic net when Recursive Feature Elimination was used for feature selection.

Conclusion: Feature selection-based classification approach has a limited impact on model performance. Also, integration of two different data sources does not result in higher performance for every classification method.