Automatic Identification of Disruptive Events in Imaging Scans

摘要

An incident or behaviour in a medical image scan, e.g. PET scan, of a subject (202) is detected. A list of image data points (204) captured during the scan is obtained, and a scan period of the scan selected. A model, e.g. a Poisson distribution model of tracer decay, for a distribution of list data over the scan period is obtained, the model at least in part based on the list data from the scan period. List data, e.g. obtained from a second scan period, is compared with the model, by calculating a statistical measure of the self-consistency of the list data with reference to the model, e.g. a joint likelihood plot (208). The measure of self-consistency can be used to determine a feature in the list data relating to the incident or behaviour, such as a twitch, for example by a discontinuity or a minima (210) in the plot. Data can therefore be divided into before and after segments, or the scan operator can be alerted.