Fully Automated Dual-Projection Topogram-Based Region-Specific Size-Specific Dose Estimate Calculation for Personalized CT Dose Management
Abstract
Purpose
Size-specific dose estimates (SSDE) are essential for personalized CT dose management; however, implementation in routine clinical workflows remains challenging. Existing automated approaches based on axial CT images are often constrained by field-of-view (FOV) limitations in large patients and substantial computational requirements. Although AAPM Report 204 recommends dual-projection topograms for patient size estimation, fully automated region-specific implementations suitable for routine clinical use have not been clearly described. This study aimed to develop and validate a fully automated dual-projection topogram-based framework for region-specific SSDE calculation.
Methods
An automated pipeline was developed to detect anatomical landmarks on paired anteroposterior and lateral topograms for region localization and patient size estimation. Anatomical regions were defined based on pelvis and lung boundaries to ensure clinically relevant dose assessment. Anterior-posterior (AP) and lateral (LAT) dimensions were measured and converted to effective diameters using AAPM Report 204 conversion factors. The framework was retrospectively validated using 90 adult CT examinations from two vendors (Canon, Philips), yielding 121 evaluable anatomical regions. Technical success was defined as successful algorithm execution with anatomically plausible landmark detection and valid measurements.
Results
Technical success rates were 98% (50/51) for chest, 88% (30/34) for abdomen, and 97% (35/36) for pelvis. Overall, 115 of 121 regions (95%) were successfully processed, producing valid AP and LAT measurements. Mean effective diameters were 27.7 ± 3.8 cm for the chest, 27.3 ± 3.8 cm for the abdomen, and 28.3 ± 3.7 cm for the pelvis, demonstrating clinically meaningful region-dependent size variation. The method generalized across heterogeneous patient anatomies.
Conclusion
Fully automated dual-projection topogram-based SSDE calculation is clinically feasible, achieving a 95% overall success rate. By enabling region-specific dose assessment without the computational burden or FOV limitations of axial images, this approach provides a scalable solution for personalized CT dose management and supports integration into routine clinical workflows.