Analysis of the minimum detectable activity of a small animal scanner

Studies in nuclear medicine frequently require the measurement of very low activity lesions in a background region. The detectability depends on a number of properties of the lesion (size, shape, signal amplitude), the background (quantum noise from the imaging process and anatomical variability) and the observer performance. We investigated the differences between the human and mathematical model observers. We used simulated images in this work. The simulated images were investigated by human observers (HuO) using an own software tool, testing whether these signals (lesions) are visible or not. For model observers the images were compared on the basis of the channelized Hotelling observer (CHO) using predefined channels.