ESTRO 2024 - Abstract Book

S5106

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

tumour volume (GTV) using both a mono-exponential model to derive the standard apparent diffusion coefficient (ADC), and a decomposition method recently proposed by Rahbek et al. [2]. Parameter values at fraction one and time-trends of the parameters were included in the analysis, inspired by a recent framework proposed by Rahbek et al. [3]. Clinical parameters included age, sex, performance status, GTV volume at baseline, the time between diagnosis and SBRT, and primary tumour/local recurrence. A Cox proportional hazards model with OS as an endpoint was made based on best-subset selection using cross validation based on Bootstrap. OS was defined as the time between the first SBRT fraction and death for any cause. The model’s discriminating power was evaluated using the C-Harrell index. Further, to validate the model, the survival probability based on the Cox model was compared with the Kaplan Meier estimator within low-, medium-, and high risk groups (calibration plot).

Table 1. Patient demographics for 45 patients receiving SBRT for localized pancreatic cancer. The reported T-stage is at the time of diagnosis. No re-staging was performed at the time of recurrence.

Overall (N=45)

Age (years)

Mean (SD)

67.5 (10.1)

Median [Min, Max]

69.5 [45.2, 85.3]

Sex

F

26 (57.8%)

M

19 (42.2%)

Performance status

0

17 (37.8%)

1+

28 (62.2%)

Clinical presentation

Primary tumour (LAPC)

37 (82.2%)

Local recurrence

8 (17.8%)

Time between diagnosis and RT (months)

Mean (SD)

7.40 (5.89)

Median [Min, Max]

6.48 [1.15, 35.4]

GTV volume (cm3)

Mean (SD)

21.6 (15.8)

Median [Min, Max]

19.1 [2.68, 77.0]

T-stage