3280 - Morphological-Dosimetric Model for Adaptive Radiotherapy Triggering in Bladder Cancer Radiotherapy

Purpose/Objective(s): To analyze the morphological and dosimetric variations in bladder patients during radiotherapy, and establish a dose-anatomy adaptive radiotherapy(ART) triggering model, providing a basis for individualized ART in clinical practice.

Materials/Methods: A retrospective study of 29 bladder patients receiving radiotherapy was prescribed with 12-25 Gy /7-12 fractions. The target volume (GTV) encompasses the primary tumor lesion and is extended outward by 5 mm to form the planned target volume (PTV). The organs at risk (OARs) include the rectum, bladder, small intestine, and colon. Image guidance was performed prior to treatment using Fan Beam CT(FBCT) after bladder filling. Morphological and dosimetric parameters of the initial plan (Plan0), plans calculated on FBCTs (PlanF), and Adaptive planning for the modified targets and OARs (PlanA) were collected for each fraction.GTV D_98% and D_95% were used as ART trigger. Logistic regression was employed to establish the relationship between the triggers and the morphological characteristics (the centroid shift, Dice, HD95) of the targets and OARs.

Results: A total of 226 treatment sessions. In terms of morphological changes, the morphological changes of GTV(the centroid shift, Dice,HD95) and bladder(Dice,HD95) were significantly associated with off-target events (off-target defined as GTV exceeding the irradiation field) (P < 0.05). The Off-targets incidence reached 50% or higher when bladder volume changes were = -20% or within the 50% –70% range. The PlanA showed significantly higher GTV dose (D_98%, D_95%) than the Plan0 and PlanF (P < 0.05), and delivered lower dose to the rectum V_10Gy, bladder(V_15Gy and V_10Gy), and colons D_max than PlanF (P < 0.05). However, PlanF demonstrated better efficacy in the local high-dose areas D_2% and D_2cc, with no significant difference observed between Plan0 and PlanA. Fractions were divided into four groups ( non-trigger ART, D_98%-triggered only, D_95%-triggered only, and dual-triggered) using ART triggers of D_98% = 95% and D_95% = 98%. Among the 226 fractions, 47.36% (107/226) met this criterion. A logistic regression-derived predictive model combined morphological and dosimetric indicators achieved favorable balance in overall predictive accuracy with an AUC of 0.869.

Conclusion: The morphological changes of GTV and bladder are key anatomical factors influencing off-target events. ART demonstrated significant improvements in target coverage and OARs protection, but increased control of local high-dose areas. The logistic regression model based GTV D_98% and D_95% thresholds can accurately predicts ART triggering.