Extraction of vessel-related information from PET images without continuous blood sampling using modified independent component analysis

Quantitative positron emission tomography (PET) analysis with a compartment model requires a plasma time-activity curve (pTAC). In the analysis, continuous blood sampling via an inserted catheter in the brachial artery is widely used to obtain pTAC. However, this continuous blood sampling encounters some difficulties. Therefore, it is of clinical, interest to develop a method that does not involve continuous blood sampling. We proposed a method for extracting pTAC from dynamic PET image sequences using a modified independent component analysis (ICA). To improve the mismatch between properties of the dynamic PET image structure and assumptions of ICA, two preprocessings were proposed: (1) appending negative images and (2) standardization. This proposed method was applied to simulated data and clinical data for evaluation. The extracted pTAC obtained using the proposed method was nonnegative and the distribution of the pTAC image was similar to that of blood volume data. In contrast, the extracted pTAC obtained by conventional ICA included negative values. It was concluded that the proposed method is useful for extracting pTAC.