Anthropometric approaches and their uncertainties to assigning computational phantoms to individual patients in pediatric dosimetry studies.

Current efforts to reconstruct organ doses in children undergoing diagnostic imaging or therapeutic interventions using ionizing radiation typically rely upon the use of reference anthropomorphic computational phantoms coupled to Monte Carlo radiation transport codes. These phantoms are generally matched to individual patients based upon nearest age or sometimes total body mass. In this study, we explore alternative methods of phantom-to-patient matching with the goal of identifying those methods which yield the lowest residual errors in internal organ volumes. Various thoracic and abdominal organs were segmented and organ volumes obtained from chest-abdominal-pelvic (CAP) computed tomography (CT) image sets from 38 pediatric patients ranging in age from 2 months to 15 years. The organs segmented included the skeleton, heart, kidneys, liver, lungs and spleen. For each organ, least-squared regression lines, 95th percentile confidence intervals and 95th percentile prediction intervals were established as a function of patient age, trunk volume, estimated trunk mass, trunk height, and three estimates of the ventral body cavity volume based on trunk height alone, or in combination with circumferential, width and/or breadth measurements in the mid-chest of the patient. When matching phantom to patient based upon age, residual uncertainties in organ volumes ranged from 53% (lungs) to 33% (kidneys), and when trunk mass was used (surrogate for total body mass as we did not have images of patient head, arms or legs), these uncertainties ranged from 56% (spleen) to 32% (liver). When trunk height is used as the matching parameter, residual uncertainties in organ volumes were reduced to between 21 and 29% for all organs except the spleen (40%). In the case of the lungs and skeleton, the two-fold reduction in organ volume uncertainties was seen in moving from patient age to trunk height-a parameter easily measured in the clinic. When ventral body cavity volumes were used, residual uncertainties were lowered even further to a range of between 14 and 20% for all organs except the spleen, which continued to remain at around 40%. The results of this study suggest that a more anthropometric pairing of computational phantom to individual patient based on simple measurements of trunk height and possibly mid-chest circumference or thickness (where influences of subcutaneous fat are minimized) can lead to significant reductions in organ volume uncertainties: ranges of 40-50% (based on patient age) to between 15 and 20% (based on body cavity volumes tied to trunk height). An expanded series of non-uniform rational B-spine (NURBS) pediatric phantoms are being created at the University of Florida to allow the full application of this new approach in pediatric medical imaging studies.