3193 - A Feasibility Study Assessed the Predictive Ability of MRI Delta-Radiomics for GU Toxicity and IPSS Changes in PCa Hypofractionated RT

H. Inokuchi, N. Mukumoto, N. Mukumoto, M. Sakagami, M. Yamagishi, N. Hamaura, K. Hayashi, R. Ogino, and K. Shibuya; Osaka Metropolitan university, Osaka, Japan

Purpose/Objective(s): This pilot study aimed to assess the feasibility of using radiomic models to predict acute and sub-acute genitourinary (GU) toxicity and changes in International Prostate Symptom Score (?IPSS) based on quantitative radiomic features extracted from daily and follow-up MR images during hypofractionated radiation therapy (RT) for prostate cancer (PCa). Materials/

Methods: Hypofractionated RT for localized PCa was initiated using a 1.5T MR-Linac from May 2022. Delta radiomics, derived from daily MR images radiomic features (RFs) at three timepoints averaged per accumulated biological effective dose (BED) bins and follow-up period after irradiation (median interval = 6 months), were analyzed to assess its capability to predict acute and subacute treatment-related toxicities. Seventy-eight RFs were extracted from daily and follow-up MR images. Clinical endpoints of acute and sub-acute GU adverse events and changes in IPSS were considered. Logistic regression models were built based on CTCAE v5.0 grade for GU toxicities and ?IPSS. The Random Forest approach was utilized for feature model building and data analysis. Cross-validation with bootstrapping iterations was employed, and the Area under the Curve (AUC) and ROC curves were calculated to evaluate model performance.
Results: The study included 33 cases: 3 cases of Damico low-risk group, 18 cases of intermediate-risk group, and 12 cases of high-risk group, with a median age of 68 years and a median PSA of 8.9 ng/mL. Prescription doses were 26Gy/2fx (2 cases), 40Gy/5fx (15 cases), 54Gy/12fx (5 cases), and 60Gy/20fx (11 cases) to the prostate plus seminal vesicle base (2/3 in some cases) as clinical target volume. For acute GU toxicity models, both daily-based and follow-up based models showed varying AUC values, with median AUCs ranging from 0.55 to 0.84. Notably, regarding sub-acute GU toxicity models, the choice of daily-based RFs influenced the AUC, with noticeable drops observed when comparing methods with both RFs cohorts. The ?IPSS models showed less variation between daily and follow-up based models compared to GU toxicity models. Overall, both daily and follow-up delta-radiomics models for ?IPSS demonstrated AUC values greater than 0.75. The daily model showed strong performance with an AUC of 0.83 after 30 Gy accumulated BED but decreased thereafter for acute GU toxicity models. The follow-up models AUC fluctuated between 0.65 and 0.76 across different timepoints.
Conclusion: This study suggests that delta radiomics derived from daily and follow-up MR images could serve as predictive biomarkers for treatment-related toxicities and patient-reported outcomes in PCa hypofractionated RT. Further validation and exploration of their clinical implications are needed to enhance PCa patient management and treatment outcomes.