ASSOCIATION-BASED MULTIMODAL IMAGE SEQUENCE ANALYSIS WITH WAVELETS IN RADIOTHERAPY

摘要

The monitoring of the patients position during treatment in radiotherapy is an important task to detect deviations from the reference position defined in treatment planning. Unfortunately, information from in-treatment data is still poor compared to pre-treatment data from diagnostics and simulation. Often, portal images (megavoltage X-ray images) are the only existing images during irradiation, which are of extremely low contrast. One approach for the visualization of useful in-treatment information is a dynamic enhancement of Electronic Portal Images (EPIs) by using offline image data as a-priori knowledge. Generally spoken, several static and dynamic images, captured by different imaging techniques, are fused to generate useful representations of the considered anatomical region. In the approach proposed in this paper, we use an associative memory applied to the wavelet transform of multimodal image data sets. The aim is to improve the feature generation and alignment during the dynamic fusion process.