Machine-learning-based Classification of Lower-grade gliomas and High-grade gliomas using Radiomic Features in Multi-parametric MRI (1911.10145v1)

Abstract: Objectives: Glioblastomas are the most aggressive brain and central nervous system (CNS) tumors with poor prognosis in adults. The purpose of this study is to develop a machine-learning based classification method using radio-mic features of multi-parametric MRI to classify high-grade gliomas (HGG) and low-grade gliomas (LGG). Methods: Multi-parametric MRI of 80 patients, 40 HGG and 40 LGG, with gliomas from the MICCAI BRATs 2015 training database were used in this study. Each patient's T1, contrast-enhanced T1, T2, and Fluid Attenuated Inversion Recovery (FLAIR) MRIs as well as the tumor contours were provided in the database. Using the given contours, radiomic features from all four multi-parametric MRIs were extracted. Of these features, a feature selection process using two-sample T-test and least absolute shrinkage, selection operator (LASSO), and a feature correlation threshold was applied to various combinations of T1, contrast-enhanced T1, T2, and FLAIR MRIs separately. These selected features were then used to train, test, and cross-validate a random forest to differentiate HGG and LGG. Finally, the classification accuracy and area under the curve (AUC) were used to evaluate the classification method. Results: Optimized parameters showed that on average, the overall accuracy of our classification method was 0.913 or 73 out of 80 correct classifications, 36/40 for HGG and 37/40 for LGG, with an AUC of 0.956 based on the combination with FLAIR, T1, T1c and T2 MRIs. Conclusion: This study shows that radio-mic features derived from multi-parametric MRI could be used to accurately classify high and lower grade gliomas. The radio-mic features from multi-parametric MRI in combination with even more advanced machine learning methods may further elucidate the underlying tumor biology and response to therapy.

• The paper presents a machine-learning approach leveraging radiomic features from multi-parametric MRI to distinguish between lower-grade and high-grade gliomas.
• It employs statistical tests and LASSO regression for feature selection, achieving an accuracy of 91.3% and an AUC of 0.956.
• T1c MRI sequences emerge as particularly informative, suggesting potential for streamlined diagnostic protocols.

Machine Learning Classification of Glioma Grades using Radiomic Features in Multi-parametric MRI

Introduction

Gliomas are malignant brain tumors representing a significant portion of CNS tumors. Classified into lower-grade gliomas (LGG, grades 1-3) and high-grade gliomas (HGG, grade 4), they differ markedly in terms of patient prognosis. Whereas LGG presents a more favorable survival rate, HGGs are characterized by rapid progression and poor outcomes. The heterogeneity of gliomas complicates clinical decision-making and necessitates advanced characterization strategies. This paper presents a machine-learning approach leveraging radiomic features from multi-parametric MRIs to enhance glioma classification efficacy. By exploiting the complementary contrast capabilities of T1, T1c, T2, and FLAIR MRIs, the paper aims to improve diagnostic precision beyond that achievable with single-sequence imaging.

Methods

The paper utilizes data from the MICCAI BRATs 2015 training dataset comprising 40 HGG and 40 LGG patient images. Multi-parametric MRIs are processed to extract radiomic features, totaling 6,920 per patient. Feature selection is conducted using statistical tests and LASSO regression to isolate significant predictors and eliminate redundancy across MRI modalities. Selected features train a random forest model optimized for the classification task. Evaluations based on classification accuracy and AUC metrics validate the method's performance.

Results

The optimized classification model achieved an accuracy of 91.3% and an AUC of 0.956, demonstrating robust differentiation between HGG and LGG. Notably, T1c MRI proved especially informative, yielding high accuracy in classifications that included this sequence. The findings indicate that T1c intensity features play a critical role in the successful categorization of gliomas, offering valuable insights into tumor biology.

Discussion

The inclusion of multi-parametric MRIs enhances classification accuracy but poses challenges such as computational complexity. T1c MRI emerges as a particularly valuable modality, suggesting possibilities for reduced reliance on other sequences if computational constraints exist. Radiomic features' ability to capture tumor heterogeneity can potentially inform individualized treatment strategies, promising better patient outcomes. Future research directions may explore the integration of additional imaging modalities and the influence of patient demographics like age on feature selection and disease characterization.

Conclusion

The paper establishes the efficacy of machine learning and radiomic analysis in distinguishing glioma grades. It highlights the potential of T1c sequences as a key component in radiomic profiling, offering gains in predictive accuracy. This approach underscores the importance of multi-parametric MRI in advancing the non-invasive characterization of brain tumors, laying the groundwork for enhanced diagnostic frameworks and personalized medicine. Continued research in this domain promises improvements in glioma treatment efficacy and patient stratification.