Development of an algorithm for automated estimation of fetal head dose in pelvimetric CT scans based on geometrical considerations

This study addresses radiation safety for fetal exposure during pelvimetry Computed Tomography (CT) scans in pregnant women, specifically in the Moroccan context. While previous studies in Morocco have utilized Monte Carlo simulations or established formalisms for effective dose estimations, none have computed a size-adjusted dose estimates for the fetal head using an automated method. The goal of this research was to enhance fetal radiation safety by developing a Python-based script for automating fetal head contouring and estimating a Size-Adjusted Fetal Head Dose (SAFHD). The automation process addresses challenges posed by the presence of other bone structures in the pelvis, therefore, a custom Python script was developed for image preprocessing, including normalization, region of interest (ROI) selection, binary image creation, morphological processing, mask creation, and region labeling. These steps ensured detection and isolation of the fetal head region. The SAFHD method adjusts the dose estimation for the fetal head based on its unique anatomical characteristics, adapting the conventional SSDE framework. This retrospective study used datasets comprising axial CT images from 136 pregnant women, all scanned using a standardized 80-detector CT scanner. The process of estimating SAFHD included several key steps: (1) calculating the pelvis SSDE using the AAPM Report 220 method for adult body CT scans, (2) introducing a correction factor (k) to adjust the pelvis SSDE to fetal head dose from previous findings, (3) performing automated fetal head contouring to extract the lateral (LAT) and anteroposterior (AP) dimensions of the fetal head, (4) calculating the water-equivalent diameter (Dw) from LAT and AP values and (5) using a size-specific conversion factor derived from interpolated pediatric head data in AAPM reports 220 and 293 to complete the SAFHD estimation. The automated method showed strong correlation with manual measurements taken by ten observers, with correlation coefficients of 0.96 for LAT and 0.94 for AP dimensions. The intraobserver agreement for Dw measurements was highly consistent, with ICC values ranging from 0.960 to 0.999. The final SAFHD values, derived using the correction factor and size conversion based on fetal head dimensions, correlated strongly (0.98) with manual estimations. By incorporating both maternal pelvis dose and specific fetal head dimensions, the SAFHD estimation offers a promising tool for clinicians to accurately estimate fetal radiation doses and verify dose limits, minimizing radiation-related risks to the fetus.