Our calculation of dose to the brain was based on estimating the amount of absorbed energy from X-rays within a mathematical model of the brain. The brain model, developed by the Medical Internal Radiation Dose Committee, is a mathematical description of the size, composition, and density of the cranium and brain. For each age (0, 1, 5, 10 and 15 years, and adult), Bouchet et al.  and Bouchet  provide the dimensions and characteristics of the cranium and the brain, which is considered as two half-ellipsoids. The caudal layers of the brain model are truncated to simulate the brain stem and cerebellum. For the purpose of our calculations, the brain was divided into 1-cm thick layers, each composed of 1 cm3 cubic volumes. The radiation dose absorbed within each volume of tissue was estimated based on the attenuation of X-radiation from the surface of the head to the depth of each volume. From those calculations, the approximate spatial pattern of absorbed dose was determined, from which the average dose to the entire brain was estimated (see Fig. 1). Additional details on the dose calculation model will be described in a subsequent publication.
Information necessary for these calculations included the age of the patients (because cranium thickness and density as well as brain size are a function of age), entrance dose (or peak skin dose) and characteristics of the radiation fields (energy and geometry of irradiation). Some of this information was available from a review of the literature. Patient-specific data from the RAD-IR database were also used; these data were stripped of protected health information and the research was conducted with the approval of our Institutional Review Board.
Typical devices used for interventional fluoroscopy procedures have two C-shape arms, each with an X-ray tube and image receptor, that can move and rotate about the patient. The proportion of exposure coming from each X-ray tube has been investigated; in our dataset, we estimate that about 60% of the exposure is from the tube below the patient and 40% from the lateral tube. The distribution of X-ray energies generated by the fluoroscopy machines is determined by settings of potential (voltage) and filtration. Peak potential is set automatically by modern fluoroscopic machines based on patient thickness and is not recorded. From special measurements conducted on a RANDO phantom, it was assumed that for pediatric examinations, peak potentials would rarely be greater than 90 kVp. Based on our review of the literature [6, 7], we found that in the range of peak potentials usually involved, the attenuation of the broad X-ray spectrum can be reasonably simulated by a single energy at 30 keV.
Information on field size and location in the brain was not available for individual cases. We simulated a variety of fields of a size and orientation characteristic of typical pediatric neuroradiologic examinations. In one scenario we considered the head irradiated by two uniform fields as large as the entire brain, with the radiation directed from the X-ray tube below the table (PA geometry) and from the lateral plane. This extreme case is unlikely. In most clinical procedures, the radiation field is focused on the diseased part of the brain, and the entire brain is not irradiated uniformly throughout the procedure. Examination fields are usually not static; they are moved in real-time to track the movement of the catheter within blood vessels. Taking into account these possibilities, a variety of possible field sizes and orientations were investigated, including large and small static and moving fields as well as combinations.
The absorption of energy decreased with depth in the tissue, resulting in the highest dose either immediately below the point of entrance or where fields overlapped. The average brain dose was determined from the dose received within all of the small volumes defined in the age-dependent brain phantom (Fig. 1). Based on the average dose to the brain, the risk of developing brain cancer during the remainder of a normal lifespan was estimated for each case using the Interactive RadioEpidemiological Program  developed by the US National Cancer Institute for estimating the probability of cancer causation following radiation exposure. IREP calculates Assigned Share, defined as AS=ERR/(1+ERR), for a specific age at cancer diagnosis. For the present paper, age-specific estimates of AS were obtained from IREP for each year of age after exposure and converted to ERR=AS/(1−AS), and a summary ERR value was computed as a life-table-weighted average of the age-specific values.