Abstract
Single-element bolometers have been widely used for medical thermometry to predict a core body temperature based on the measured surface temperature and a pre-determined clinical correlation. The size-of-source effect (SSE) in bolometers is recognized as an important source of error and has been extensively studied such that SSE can be controlled sufficiently to obtain the required accuracy. Thermal imaging cameras relying on much of the same principles have also been widely used for medical thermometry, but recent work has shown that SSE in thermal imaging differs from SSE in single-element bolometers. An unappreciated aspect of this artifact in thermal imaging has been its outsize impact on accuracy, producing more than a degree of inaccuracy in typical scenarios. However, because the artifact has so far avoided a satisfactory characterization, this impact has not been incorporated into the standards and expert recommendations for body thermometry with thermal imaging. In this work, we characterize SSE-like artifact as proportional to the difference between source and non-source temperatures as a function of source size, showing that the same percent deviation artifact is obtained at different source temperatures. We then derive an objective method to obtain convolution kernel parameters and apply these to several thermal imaging sensors and optics configurations. Finally, we show that these methods are sufficient to achieve the required accuracy for body thermometry with thermal imaging.