Optimization of activity level in rCBF SPECT using the observer study Visual Grading Regression

Abstract: The purpose of this work is to assess the activity level needed to achieve satisfactory diagnostic information in regional cerebral blood flow single photon emission computed tomography (rCBF SPECT) by introducing a new evaluation method to be used for and hopefully facilitate optimization studies in nuclear medicine in the future. The purpose is further to perform a visual grading study and investigate the use of this new evaluation method, Visual Grading Regression (VGR). Image quality criteria applicable to rCBF SPECT images will be defined in this work and their relevance for evaluating rCBF SPECT image quality is investigated. The study comprises a material of 21 consecutive patients with dementia issue that have undergone an rCBF SPECT examination. An administered activity of 1000 MBq 99mTc labelled HMPAO was injected to all patients in the study. From one single examination, five studies corresponding to different activity levels (500, 625, 750, 875 and 1000 MBq) were generated by using a gated acquisition. Iterative image reconstruction, OS-EM, including corrections for attenuation, scatter and distance dependent resolution was used. Three experienced observers, i.e. specialists in nuclear medicine, evaluated the images by rating their confidence about the fulfilment of specific image quality criteria. Seven criteria were defined in this study, developed in collaboration with experienced specialists in nuclear medicine with comprehensive knowledge on how to evaluate rCBF SPECT images. The result of the observers assessment were analysed using Visual Grading Regression, a method based on ordinal logistic regression with the aim to analyse data from visual grading experiments. The result shows that there is a significant difference in perceived image quality between 500 MBq and the reference activity, 1000 MBq, in five of the seven image quality criteria. No statistical significant degradation was found between any other activity level than 500 MBq and the reference activity (1000 MBq). This study doesn’t prove that any other activity level provides the same image quality as 1000 MBq, only because no difference was seen, but it gives an indication that the activity level could be reduced without losing too much diagnostic information. The analysis method used, Visual Grading Regression, has proven to be convenient and easy to use for this kind of optimisation studies in nuclear medicine. The defined criteria cover the areas of the brain that are of interest in blood flow examinations and the results of this study showed that the observers used the whole confidence rating scale for each criterion, which is desirable. Some of the criteria had a very low proportion of rating scores corresponding to a fulfilment of the criterion, meaning that the satisfaction of the observers is low. A reversion or adjustment of these criteria might be needed to investigate whether the low satisfaction level is due to the formulation of the criteria or if only so the particular area is difficult to assess.