ESTRO 2024 - Abstract Book

S5075

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

Results:

As expected, in all five models, the pre-selected candidate dosimetric predictor and abdominal surgery were predictive. The EUD (n= 0.1) –based final model (Figure 1) performed best based on discriminative ability and calibration in the internal validation, with a mean AUC of 0.64, a mean calibration slope of 1.06 and a mean ICI of 0.047. To corroborate that the EUD-based model is applicable to both fractionation schedules, the combined model was compared to separate models for CF and HF. Figure 2 shows that fitting CF and HF data separately with EUD as only predictor, resulted in relations similar to the when their combined data was fitted.