TY - JOUR T1 - Direct VOI-dedicated voxelwise Patlak estimation for quantitative dynamic imaging JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 1929 LP - 1929 VL - 57 IS - supplement 2 AU - Wentao Zhu AU - Zhifeng Yao AU - Yun Dong AU - Defu Yang AU - Mu Chen AU - Hongdi Li AU - Jun Bao AU - Zhongwei Lv Y1 - 2016/05/01 UR - http://jnm.snmjournals.org/content/57/supplement_2/1929.abstract N2 - 1929Objectives Direct Patlak analysis may outperform indirect ones with higher sensitivity and specificity in lesion detection. However, conventional direct Patlak analysis is applied to the entire image, which may violate the model assumption as some tissue types have not reached the steady state. This study developed a VOI-dedicated direct Patlak analysis method, which restrains direct Patlak analysis to the VOI only, and does not require the rest of the image reaching steady state.Methods Several patients underwent 1h PET scanning for the torso FOV (field of view) in this study. The entire acquisition was partitioned to 60 1-min frames and reconstructed individually. An image-derived input function was obtained after segmenting the aorta in the reconstructed frames. The direct Patlak parametric estimation method was applied to the heart VOI with 40~60 min data. Two other methods were performed as comparison: (a) image-based Patlak analysis based on the 60 static reconstructed images; (b) routine direct Patlak estimation on the entire image space. The mean and variance of Patlak parameters in the heart ROI were computed for all methods.Results The Patlak images computed by the proposed method differed less than 3% in the heart VOI mean from the ones obtained with (a) and (b) after sufficient number of iterations. However, the noise for the proposed method and (b) was more than 75% lower than (a). The Patlak slope image provided a ratio of 5 or higher contrast of heart muscle against left and right ventricles than SUV image using the same acquired data. Besides, the Patlak parameters obtained with the proposed method and (b) differed less than 3%. This consistent quantitative result indicated that 40 min was sufficient for Patlak model to be valid for most tissue types. Experiments using 20-40 min data revealed more than 7% difference between the proposed method and (b) in some ROIs in the heart, possibly because of model violation for tissues out of the heart VOI during this early stage.Conclusions The proposed Patlak analysis algorithm provides quantitative heart Patlak parametric images with higher image contrast than SUV. Unlike conventional direct Patlak analysis, it does not demand Patlak model validation for the entire image. On the other hand, comparing with image-based Patlak analysis, the advantage of the proposed approach is that the noise can be significantly reduced. ER -