ESTRO 2024 - Abstract Book

S5089

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

Conclusion:

Even though the model coefficients are not fully validated on a new group of patients treated in the same Institute after the previous training/validation of the model; training a new model with the same RFs shows that Morphology_COMshift is still predictive. The relatively low number of patients of the new cohort may have influenced the results. A new model trained on both cohorts (36 patients + 176 patients from the referenced article) could increase its predictive potential. The size of the combined dataset could make a more robust cross-validation possible.

Keywords: radiomics, pancreatic cancer, radiotherapy

References: