(Sunday, 7/29/2018) 4:00 PM - 4:55 PM
Room: Room 205
Purpose: Radiomic features have the potential to uncover disease characteristics that fail to be appreciated by the naked eye. One major challenge facing radiomics are image acquisition parameters, which determine radiomic features. One method to solve this challenge is to standardize image acquisition parameters. Another is to use post-image processing approaches. The question is then how to evaluate effectiveness of those post-imaging methods. Direct comparison of radiomic features is time consuming and impractical due to the large number of features. We attempt to investigate if there is any relationship between NPS and radiomic features and if NPS can be used for standardization evaluation.
Methods: A water phantom was scanned in a Siemens Force. CT images were reconstructed with the Advanced Modeled Iterative Reconstruction (ADMIRE Bl64, Bl57, Br40) at different strength settings (1 to 5). Unsubtracted and subtracted noise power spectra (NPS) were measured. CT images from the lung cancer patient were also reconstructed with Bl64, Bl57, and Br40. The imaging biomarker explorer (IBEX) was used to extract image features. The ROI was held consistent among different kernels.
Results: The peak frequency differences of Bl64-Br40, Bl57-Br40, and Bl64-Bl57 are approximately 4.25, 2.55, and 1.75, respectively. The majority of image features of the lung cancer patient show consistent changes, with small difference with Bl64-Bl57 and larger difference with Bl64-Br40. This implies that NPS may indicate image feature changes. In addition, the change in image features may be estimated simply from the change in the peak frequency and magnitude of the NPS. A few image features, such as cluster shade, information measure correlation and basic pixel intensity, do not follow this pattern.
Conclusion: There is a consistent relationship between the NPS and radiomic features, which may be potentially used to evaluate the standardization methods of CT images acquired with various acquisition parameters.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
