Development of a Novel Automatically Quantitated Software for Custom-Designed Thoracic Bone Phantom

Objectives: We have designed a thoracic bone phantom (we have named "Sim2 BONE phantom") to evaluate acquisition and reconstruction parameter. However, since it requires expert knowledge of image analysis, it is difficult to evaluate image quality. Furthermore, a repeatability and reproducibility of evaluation for an image quality is very important. Therefore, a novel automatically quantitative software for Sim2 BONE phantom (we have named "Hone graph"), in this study, was developed and evaluated accuracy of the software. Methods: We acquired an excellent and ten routine single-photon emission computed tomography (SPECT) images using a dual-head gamma camera with a low-energy high-resolution collimator . Contrast and full width at half maximum (FWHM) for the one excellent SPECT image were measured using Hone graph automated method and manual method by two board certified nuclear medicine technologists (BCNMTs) each ten times. Coefficient of variations (CVs) for contrast and FWHM were calculated to evaluate repeatability. Furthermore, contrast, FWHM and recovery coefficient (RC) for another ten routine SPECT images were measured to evaluate inter-operator variations using both automated and manual method by two BCNMTs. And CVs for contrast, FWHM and RC were calculated to evaluate reproducibility for two methods. Results: Even the manual method, the repeatability with the one excellent SPECT image was good (CV for contrast : 2.7%, CV for FWHM : 1.4%). Although, the automated method gave the same results for an excellent SPECT image as all each ten times. On other hands, with ten routine SPECT images, inter-operator variations were significantly larger with the manually method (contrast : 12%, FWHM : 0.2%, RC : 13%) than with the automatically method (contrast : 0.2%, FWHM : 0.7%, RC : 0.4%). The mean CVs for contrast, FWHM and RC with the automatically method were 3.1%, 1.4% and 2.1%, and those with manually method were 7.9%, 5.3% and 9.2%, respectively. Conclusion: The repeatability and reproducibility for Hone graph automated method was significantly superior as compared with those for manual methods. The developed software should be easy to use for everyone.