Abstract
Purpose: Voxel-based dose-response maps (DRM) can identify high-risk sub-regions in organs-at-risk (OARs) with potential dose minimization. Different methods of obtaining dose-surface maps (DSM) can affect the resulting DRM. We compared two DSM methods (independently developed at two different centres)of hollow OARs to obtain DRM in prostate cancer (PCa) patients undergoing radical radiotherapy. Methods: The rectum and bladder of 700 PCa patients from REQUITE trial were auto-segmented using ADMIRE (ELEKTA, Sweden) on planning computed tomography. Centre A unwrapped rectum and bladder DSMs using cylindrical coordinates and spatially normalized using 91x90 voxel grid. Centre B spatially normalized rectum with 3cm fixed-length (most caudal corresponding to the anal canal) resulting in 97x100 grid (median height of DSMs=97) and without 3cm fixed resampled to obtain 91x90 grid. Bladder DSMs were spatially normalized with 1.5cm fixed-length (corresponding to bladder neck) resulting in 50x100 grids. At both centres, doses were converted to equivalent doses in 2Gy fractions with α/β=1.5Gy. The clinically reported endpoints included late rectal (bleeding) and bladder (obstruction, frequency, urgency, retention, incontinence) toxicities 2-years post-radiotherapy. To investigate whether planned dose was associated with each toxicity, a voxel-wise Welch's t-test was performed between the event (≥grade 1)and the non-event groups; the maximum t-value over the image was obtained for1000 permutations, the distribution threshold at 95th percentiles to identify voxels where dose differences in patients with/without events were statistically significant. Results: For bladder, tract obstruction (rate:7.4%) and urinary incontinence(rate:28.9%) were associated with posterior wall and trigone area (corresponding to the high dose area), respectively. No sub-regions were identified for the rectum. Visual comparison showed good agreement between the DRM obtained at both centres. Conclusion: Comparison of various DSM methods, utilizing identical segmentations and clinical outcomes, revealed minimal differences between the resulting DRM, instilling confidence in the robustness of such voxel-based techniques.
Original language | English |
---|---|
Publication status | Accepted/In press - Jul 2024 |
Event | AAPM 2024 - , United States Duration: 21 Jul 2024 → 25 Jul 2024 https://w4.aapm.org/meetings/2024AM/ |
Conference
Conference | AAPM 2024 |
---|---|
Country/Territory | United States |
Period | 21/07/24 → 25/07/24 |
Internet address |