ESTRO 2025 - Abstract Book

S3723

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

Purpose/Objective: Hypoxia is a well-known factor of radioresistance and poor prognosis in cancer. Measurements with Eppendorf pO2 microelectrodes are considered the gold standard to assess hypoxic regions. However, this technique is highly invasive and constrained to accessible tumors [1] . Non-invasive imaging methods provide an indirect view of tumor cells and vascular architecture. Hypoxia mapping from MRI IntraVoxel Incoherent Motion (IVIM) models [2] have shown good results in detecting and diagnosing prostate cancer (PCa) from benign tissues [3] . Other studies revealed a significant correlation with PCa aggressiveness and a promising risk stratification [4] . This study aims at exploring how hypoxia mapping could provide relevant features to differentiate biochemical recurrent (BCR) from non-BCR peripheral zone PCa, the most frequent zone of prostate carcinomas. Material/Methods: A cohort of 37 patients with peripheral zone PCa having undergone radiotherapy (RT) was used. Ten patients suffered BCR within 5 years of irradiation . T2-w, ADC maps and multi b-values DWI sequences were acquired before RT using two different 3.0 T MR machines (Achieva TX, Philips Healthcare; Magnetom Verio, Siemens Healthcare) . An IVIM model [2] , to distinguish tissue perfusion and diffusion effects, was used ( Equation (1) ).

Apparent Diffusion Coefficient ( ADC ) correlated with cell density, fractional Blood Volume ( fBV ) correlated with vessel density and pseudo-diffusion coefficient ( D* ) correlated with perfusion, were computed by pixel-wise fitting the IVIM model with signal s( b ) from multi b - values DW MR images ( Figure 1, step 1 and 2). Then, hypoxia mapping was performed from ADC and fBV maps, in accordance with [5], in 5 segmented areas ( Figure 1, step 3). Mean, median, standard deviation (std) and entropy from hypoxia maps and raw MR images were computed for the 5 segmented areas. A statistical t-test or a Mann-Whitney test was chosen for group comparison between BCR and non-BCR patients according to the result of normality Shapiro-Wilk test.

Results: Figure 2 presents the top 20 discriminating features according to p-value. Furthermore, distributions for the two groups of patients are presented for the two most discriminating features. The test associated to the first feature from hypoxia maps produced a p-value two times lower than the second most significant one from raw ADC image (ADCReg).