A multiple regression model for peak skin dose using principal component analysis in interventional radiology.

Abstract

This study addresses the growing concerns of increased radiation doses to patients resulting from the increased complexity of interventional radiology procedures. Despite the importance of dose management, few facilities use dosimetry systems to measure and control patient radiation doses. To aid in patient exposure control, this research aimed to predict the peak skin dose (PSD) using dose parameters from digital imaging and communication in medicine-radiation dose structured reports. The study focused on air kerma (K_a,r) and air kerma area product (KAP) values categorized into fixed dose (radiography and fluoroscopy) and motion dose (rotational digital subtraction angiography) for frontal and lateral biplane devices. Using single and multiple regression analysis, model equations for PSD were developed based on data from a radio-photoluminescence glass dosimeter and five dose parameters. Principal component analysis (PCA) was applied to consolidate the data, and multiple regression models were created using principal component scores. The results showed that rotational digital subtraction angiography had a minimal impact on PSD, whereas the K_a,r value demonstrated higher accuracy in predicting PSD than KAP. The inclusion of PCA in the multiple regression model further improved accuracy, with a root mean squared error of 226, confirming that PCA-enhanced models are more effective in predicting PSD.