Paper Proffered Program Diagnostic and Interventional Radiology Physics

BEST IN PHYSICS (IMAGING): Machine Learning-Based Rapid Prediction of Patient-Specific Organ Doses In Fluoroscopically Guided Procedures

Abstract
Purpose

To develop a machine learning–based method for rapid estimation of patient-specific organ doses in fluoroscopically guided procedures.

Methods

A total of 286 RDSRs (>40,000 exposure events) from three medical centers were processed using the National Cancer Institute dosimetry system for Radiography and Fluoroscopy (NCIRF), a Monte Carlo-based fluoroscopic patient dose calculation system. Organ doses normalized by dose–area product (DAP) were calculated using Monte Carlo simulations and served as reference outputs for supervised model training. Two categories of input features were derived for each exposure event: RDSR-based irradiation parameters and ray-tracing-based geometric metrics characterizing beam-organ relationships. Tree-based XGBoost regression models were trained separately for 31 organs/tissues using a total of 75 features.

Results

The XGBoost models achieved high predictive performance, with coefficients of determination (R²) generally exceeding 0.95 for most organs near or within the x-ray field. Those organs with lower values (>0.83) were typically further from the field and received lower doses. For the five highest-dose organs, the 5th-95th percentile relative error with respect to Monte Carlo–derived doses were −9.1% to 8.8%, and −3.7% to 3.9% for detriment-weighed dose. The dose estimation time of the proposed method was ~1.2 s, representing approximately 50-fold reduction in computation time compared with Monte Carlo simulations.

Conclusion

The proposed methodology preserves the source and patient modeling fidelity of NCIRF while substantially reducing computation time, enabling practical estimation of cumulative organ doses for complex fluoroscopically guided procedures. Occasional large prediction errors under specific source conditions were primarily attributable to limited training data and are expected to be mitigated through dataset expansion. The proposed approach will be integrated into NCIRF, significantly improving the accessibility and computational efficiency of patient dosimetry in interventional radiology.

People

Related

Similar sessions

Poster Poster Program
Jul 19 · 07:00
B-Trac – Breast Tissue Rotation and Compression Apparatus for Calibration

Mammography (compressed 2D) and MRI (uncompressed 3D) capture breast tissue under different conditions, complicating tumor localization across modalities. To bridge this gap, we developed a customizable physical platform to simul...

Dayadna Hernandez Perez
Diagnostic and Interventional Radiology Physics 0 people interested
Poster Poster Program
Jul 19 · 07:00
Comprehensive Medical Physics Assessment of Digital Mammography Equipment: A Three-Year Multi-Site Evaluation of Technical Performance and Radiation Safety at 24 Saudi Arabian Healthcare Institutions (2022–2024)

To conduct a comprehensive multi-center audit evaluating the technical performance, image quality, and radiation safety of digital mammography systems across 24 unique healthcare facilities in Saudi Arabia. This study aims to est...

Sami Alshaikh, PhD
Diagnostic and Interventional Radiology Physics 0 people interested
Poster Poster Program
Jul 19 · 07:00
Starting Small: Implementing a CT Protocol Optimization Program

This talk describes our organization’s CT optimization program, and how we implemented it to make efficient use of limited physicist time.

Robert J. Cropp, PhD
Diagnostic and Interventional Radiology Physics 0 people interested