Introduction: Determining the optimal course of treatment for low grade glioma (LGG) patients is challenging and frequently reliant on subjective judgment and limited scientific evidence. In the past, radiomic models capable of predicting overall survival or specific genetic mutations in LGG have been established. However, the majority of these models were constructed under constrained conditions and were sensitive to subjective rater bias. The radiomics technique utilized in this study is the first to accurately predict natural LGG behavior in a clinically relevant environment. Moreover, to the best of our knowledge, there has never been a radiomics investigation of natural LGG growth dynamics.
Methods: We retrospectively included 349 LGG patients to develop a prediction model using clinical, anatomical, and preoperative MRI data. To avoid bias, we trained a deep learning U2-model for tumor segmentation. Following feature extraction, Cox proportional hazard models were fitted for 10-year overall survival and time to malignancy estimation models. To investigate the factors that influence the rate of tumor growth, a basic linear model was applied.
Results: In a postoperative model combining radiomic features, we derived a c-index of 0.82 (CI: 0.79-0.86) for the training cohort over 10 years and 0.74 (Cl: 0.64-0.84) for the test cohort. Preoperative models showed a c-index of 0.77 (Cl: 0.73-0.82) for training and 0.67 (Cl: 0.57-0.80) test sets. We achieved a mean whole tumor Dice score of 0.837 for tumor delineation.
Conclusion : Our findings indicate that we can accurately predict the survival of a heterogeneous population of glioma patients in a pre- and postoperative setting. We further show that radiomic feature analysis may be beneficial in determining biological tumor activity, such as time to malignancy and LGG growth rate.
