No abstract

he prospect of improved clinical outcomes and more efficient health systems has fueled a rapid rise in the development and evaluation of AI systems over the last decade. Because most AI systems within healthcare are complex interventions designed as clinical decision support systems, rather than autonomous agents, the interactions among the AI systems, their users and the implementation environments are defining components of the AI interventions' overall potential effectiveness. Therefore, bringing AI systems from mathematical performance to clinical utility needs an adapted, stepwise implementation and evaluation pathway, addressing the complexity of this collaboration between two independent forms of intelligence, beyond measures of effectiveness alone 1 . Despite indications that some AI-based algorithms now match the accuracy of human experts within preclinical in silico studies 2 , there

Interventional radiologists receive nonuniform occupational radiation doses, with relatively high doses to the head and extremities and low doses to the trunk, which is protected by a lead apron. Twenty-eight interventional radiologists from 17 institutions wore thermoluminescent dosimeters over their collars and under their aprons for a 2-month period. The estimated annual radiation dose was converted to effective dose as suggested by the International Commission on Radiological Protection. Effective dose is used to relate the risk associated with nonuniform dose to that associated with an equivalent uniform whole-body dose. The mean annual effective dose was 3.16 mSv (316 mrem), with a range of 0.37-10.1 mSv. The mean annual effective dose is approximately equal to the mean natural background dose of 3 mSv per year from radon and other natural sources and is only 6% of the National Council on Radiation Protection and Measurements' recommended effective dose equivalent limit of 50 mSv per year. The annual radiation risk of fatal cancer would be less than one per 10,000 for almost the entire career of an interventional radiologist.

Computer programs for calculation of the backscatter factor and absorbed dose in breast phantoms are developed based on Monte Carlo simulation. Rad/roentgen conversion factors are calculated for water, fat, a mixture of 50% water and 50% fat, and lucite irradiated with monoenergetic and polyenergetic x rays from a tungsten or molybdenum anode x-ray tube. These factors can be used to estimate skin dose, average depth dose, or average integral dose in a breast when the beam quality, exposure in air at the skin position, and composition of the breast are known. Calculated backscatter factors are considerably greater than measured values reported previously.