Quantitative 3D image analysis

In order to extract information from three-dimensional (3D) images such as CT and MRI, it is first necessary to outline the relevant structures (organs, rumours etc.). Simple methods such as thresholding only work when there is a one-to-one correspondence between the required structure and the image intensity. Otherwise, manual outlining is normally used, but that is tedious and time consuming, To solve this problem we have developed TOSCA, part of a pilot system for conformal radiotherapy planning which was funded by the European Commission (EC) in the AIM research programme. TOSCA has a number of interactive image segmentation algorithms which enable the user to outline 3D volumes with a minimum of effort. It also allows images acquired at different times, or from different modalities (e.g, CT and MRI), to be registered with each other so that they can both be utilized when defining the volumes. Once the volumes have been outlined, they can be visualized in 3D to study their spatial relationship or to plan radiation treatment. Quantitative analysis (area, volume, statistics etc.) is also facilitated, so that the response of a tumour to treatment, for example, can be assessed. In addition to describing the techniques employed, clinical feedback on the usefulness of the system is presented.