A Pan-Sharpening Approach to Multiresolution Image Fusion for Hybrid Counting/integrating CT
D Rigie*, P La Riviere, The University of Chicago, Dept. of Radiology, Chicago, IllinoisTH-A-103-7 Thursday 8:00AM - 9:55AM Room: 103
Purpose: We investigate the application of a pan-sharpening algorithm to a hybrid CT dataset containing low-resolution, photon-counting data and high-resolution, integrating data to generate images that inherit both the resolution properties of the integrating detector and the spectral information of the photon-counting detector.
Methods and Materials: Our 2D simulation studies include a realistic x-ray tube spectrum, a photon-counting detector with 5 energy bins, and an energy-integrating detector. We used several material-specific, ellipse-based phantoms to generate datasets with both detectors, reconstructing panchromatic 512x512 images from the integrating data and multispectral 128x128x5 images from the photon-counting data. We also separately reconstructed 512x512x5 images from the photon-counting data, which we regarded as the ground truth. We used a pan-sharpening algorithm from the remote-sensing field, which contains 5 tunable parameters that control the balance between resolution recovery and spectral distortion. The optimization-based algorithm seeks a high-resolution image that has the spatial detail of the panchromatic image and the spectral characteristics of the multispectral images. The parameters were selected to minimize the mean-squared error between the pan-sharpened images and the ground truth.
Results: The pan-sharpened images appear to have recovered nearly all of the detail of the panchromatic image without severe spectral distortion. There is some visible distortion near edges of high-contrast objects, such as bone. Using pan-sharpened and ground-truth images from two of the energy bins of the photon counting detector, we looked at the angular separation of calcium and iodine which agreed to within 0.2 degrees (6%).
Conclusions: Our preliminary studies suggest a possible solution to the count-rate limitations of photon-counting detectors. Using a high-resolution image, acquired by a conventional detector as a prior, one may be able to acquire low-resolution photon-counting data and then recover spectral images that possess both good spatial resolution and the desired spectral information.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by Toshiba Medical Systems.