Single input multiple output (SIMO) optimization for input function estimation: a simulation study

Abstract

Blood input function (BIF) estimation is required in many applications. A minimally invasive approach based on single input multiple output (SIMO) system optimization was investigated for extraction of BIF from small animal dynamic PET images. In the SIMO based approaches, ROI curves from different regions were considered the output of the SIMO system and the BIF was the input. Parametric estimation was performed to estimate the BIF and the impulse responses. Feng et al. has previously reported the application of this approach on human brain images. The focus of this work was to examine its feasibility for small animal PET images, and in particular, with two regions of interest (ROI) defined in the heart area. Monte Carlo simulations were performed to generate multiple sets of the ROI curves. A weighted least square (WLS) approach was applied for estimation of the parametric input function, the kinetic parameters and the mixing coefficients for the TACs, with 0, 1 and 2 blood samples as constraints to the BIF. It was observed that the SIMO based approach was feasible for BIF estimation even with only one blood sample as constraint. The application of the blood sample constraints further improved the BIF estimation accuracy.