An Algorithm to Select the Optimal Combination of Dual Energy CT Images to Improve Target and Organs at Risk Visualization
N. M. Thorne1,2 *, E. W. Izaguirre 1,2 ,(1) Methodist University Hospital, Memphis,TN,(2) University of Tennessee Health Science Center, Memphis,TN
Presentations
(Saturday, 3/26/2022) [Central Time (GMT-5)]
Purpose: To develop an algorithm that constructs an optimal image based on the available dual energy image reconstruction kernels and energy channels in a commercial dual energy CT simulator to facilitate malignancy and organs at risk visualization and contouring of multiple tissues.
Methods: An unsupervised Gaussian Mixture Model (GMM) was used to automatically classify each material in a multi insert phantom based on material’s response to the dual energy scans. The developed algorithm determines the optimal linear combination of images that maximizes the contrast and minimize the dispersion between multiple materials or tissue pairs. The metric for the maximizing process is a generalized contrast to noise ratio (GCNR) function that accounts for the simultaneous maximization of the contrast in all the tissue pairs in the region of interest. The algorithm was validated using a phantom with tissue equivalent materials inserts and comparing the Signal to Nosie and Contrast to Nosie ratios between materials.
Results: The algorithm produces a single image data set, called the Multi Contrast Projection Image (MCPI), which is a linear combination of the Dual Energy channels and multiple reconstruction kernels and maximizes the GCNR function. This optimal MCPI was tested against all available dual energy CT simulator images and found to have the optimal CNR and SNR for tissue visualization and contouring.
Conclusion: We developed an algorithm, which creates an optimal image for contouring multiple tissues. The algorithm expands upon previous work that was limited to single tissue pairs and two image sets. The resulting MCPI showed an improved SNR and CNR. Furthermore, the algorithm can be easily implemented in a clinical workflow as it uses the existing available dual energy CT scanner images and protocols.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email