Exploring the Potential of Radiomic Features in Predictive Murine Tumor Growth Modeling

A growing number of mathematical models have been developed, providing a powerful means to forecast tumor growth. These models hold great promise as tools in the advancement of personalized medical care. Yet, their clinical relevance often remains restricted. Adequate parameter estimation and integration of existing clinical data from the patient is crucial for individualized predictions. Since CT scans form an integral part of radiotherapy, a widely used cancer treatment, we propose the use of CT-derived radiomic features to attempt to predict the course of tumor growth curves. Therefore, in this exploratory study, we examine the association between radiomic features and the three parameters of the Gompertz tumor growth model. As a proof of concept, murine tumor growth (CT26) is monitored and the Gompertz model is fitted to each animal's data to obtain the parameters. Radiomic features of the tumors were then obtained via CT, which were subsequently analyzed using exploratory factor analysis. Finally, correlations between the two were computed. We were able to discover latent factors within the radiomic features that revealed significant correlations with the parameters of the Gompertz model. Additionally, distinct features exhibiting strong correlations with the factors were identified. These may be instrumental in predicting future tumor growth. These findings offer encouraging preliminary evidence for supporting the notion that CT scans contain information that can estimate the future course of tumor growth curves. This approach opens up promising possibilities for personalized medicine in oncology, by facilitating the development of treatment plans tailored to individual patients' tumor growth patterns, right from the time of diagnosis.