Abstract
2243
Objectives SPECT can provide functional images, but has limited success in providing quantitative images, largely attributed to a lack of standardized procedures in vendor software, causing inconsistent image contrast among institutions. This study was intended to develop a novel technique to minimize the institutional inconsistency of SPECT images.
Methods The method includes 4 processes; (a) Acquired SPECT images were spatially aligned to the digital data of recently-developed a 3-dimensional brain phantom recently in our laboratory. (b) The spatial resolution was evaluated by means of inverse Fourier transformation of SPECT images and a digital data. (c) SPECT images were further smoothed to become consistent among the data acquired at different institutions. (d) Accuracy defined as difference between the smoothed-digital and SPECT images, and inter-institutional deviation were assessed. Feasibility was tested on the data acquired at 5 institutions. A recently developed quantitative SPECT reconstruction software, QSPECT, was applied to reconstruct the projection data acquired using different SPECT devices.
Results Spatial resolution obtained by this technique agreed with values assessed using a multiple rod phantom within an accuracy of +/- 0.2mm, suggesting validity of the inverse Fourier transformation technique. The pixel variability was 10-30 %, if not applying the present technique. With the present technique, pixel variability was dramatically reduced to 3-10 %. The present technique provided the high reproducibility of the images obtained by different cameras, which showed that the inter-institutional difference of SPECT images is intrinsically a little.
Conclusions The method developed in this study provided simultaneous evaluation of the spatial resolution and image accuracy using the 3-D brain phantom, and might be effective in minimizing effects of the inter-institutional difference of SPECT systems.
Research Support Spatial resolution was accurately estimated by use of this method, and image accuracy could be evaluated independent to the differences of the spatial resolution