3365 - Automated Mapping of Bladder Subregions to Assess Localized Dose-Response Patterns in Trimodality Therapy for Muscle Invasive Bladder Cancer

12:45pm - 02:00pm PT

Hall F

Screen: 2

POSTER

Presenter(s)

Matthew Whitmill, MD - UNC Health, Chapel Hill, NC

M. A. Whitmill¹, K. H. Gessner², Z. Feuer², B. M. Anderson¹, D. Melwani³, M. C. Repka¹, S. Sud¹, and A. Wijetunga⁴; ¹Department of Radiation Oncology, University of North Carolina, Chapel Hill, NC, ²Department of Urology, University of North Carolina, Chapel Hill, NC, ³Chicago College of Osteopathic Medicine, Chicago, IL, ⁴Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY

Purpose/Objective(s):

As a hollow, thin-walled organ, the bladder poses challenges for dosimetric analysis, particularly in assessing regional dose-response relationships. Most dosimetric studies treat the bladder as a single structure, despite known anatomic and functional differences across subregions. Bladder subregion dosimetry remains limited due to the lack of automated, reproducible segmentation methods. This study developed and validated a deformable-registration-based approach for bladder subregion mapping and dosimetric evaluation in trimodality therapy (TMT) for muscle-invasive bladder cancer (MIBC).

Materials/Methods:

A template DICOM with 0.25 mm edge-length voxels was fused to a sample pelvis CT. The bladder was contoured into six anatomic subregions (anterior wall [AW], posterior wall [PW], left/right lateral walls [LW/RW], trigone [TG] and dome [DM]) per EAU cystoscopy and VI-RADS guidelines then further sub-divided into 32 approximately equal-area subregions. This template was deformably registered to simulation scans of 24 MIBC patients and contours were propagated. Dose-volume histograms (DVHs) were generated for each subregion. Gross tumors on simulation scans were mapped to the template and assigned to specific subregions if overlapping by >1 mm³. Two urologic oncologists independently mapped tumor location to six anatomic subregions based on chart review. Cohen’s kappa was used to assess inter-rater agreement between 1) the two urologists and 2) the template-based mapping vs the urologist consensus mapping.

Results:

The template successfully fused to all simulation scans, and segmented subregions were successfully transferred in 99.9% of cases. DVHs were generated and exported for all 911 mapped regions.

Pre-treatment tumor localization agreement between urologists varied: fair for AW (?=0.26) and PW (?=0.36), moderate for TG (?=0.55), and substantial for LW (?=0.92), RW (?=0.85), and DM (?=0.91).

For the subset of patients with gross disease at simulation (n=7), the maximum concordance between the template-based and urologist mapping was 69%, whereas concordance between urologists was 50%. Agreement between the template and union of urologist mappings was substantial for TG (?=0.72), LW (?=0.72), and RW (?=0.70), but was not better than chance for DM, AW, or PW.

Conclusion:

This study establishes a method for automated, standardized bladder subregion dosimetry. Successful template fusion and DVH generation demonstrate feasibility, while high concordance with urologist mapping supports accuracy and reliability. Discrepancies may reflect inconsistent subregion definitions, which could be mitigated through higher-resolution mapping to reduce reliance on boundary definition, allowing same- or adjacent-region classification to better represent tumors near boundaries. This technique lays the foundation for subregion-based dose-response studies, potentially improving understanding of toxicity and recurrence patterns in MIBC treatment.