Matched subspace detection for dynamic PET: an ROC phantom study for MAP reconstruction

We describe a modified matched subspace detection algorithm to assist in the detection of small tumors in dynamic PET images. The algorithm is designed to differentiate tumors from background using the time activity curves (TACs) that characterize the uptake of PET tracers. Our detector assumes additive Gaussian noise with known covariance. The covariance is computed using a plug estimator applied to the observed data. To evaluate the method, an ROC study for dynamic PET tumor detection was designed. The detector uses a dynamic sequence of frame-by-frame 2D MAP reconstructions as input. We compare the performance of this subspace detector with that of a nonprewhitened linear observer applied to a single frame image.