| Summary: | Purpose:
Medical imaging systems are used to scan patients to obtain valuable information for diseases diagnosis and assisting treatment. An ideal scanner should be sensitive enough to detect any trace amount of abnormal tissue at its early stage. With the continuous development of high-tech treatment systems such as Tomotherapy (manufactured by Tomo HD), the high-resolution imaging system is favorable to reduce the damage of normal tissue due to the image guidance of Mega-voltage beam before treatment. In this study, a software approach was presented to improve image resolution without hardware upgrade of a scanner.
Methodology
A programming technique “Super Resolution Technique” was used and demonstrated in an example of CT. It utilized several similar images with known relative shifts between them. (They can be positional or angular shifted and taken at the same time frame as far as possible). Those images are of low resolution and can be reconstructed to form a higher resolution image. A Super Resolution program was written by MATLAB to prove the method. The experiments 1 to 4 were purely computer-based simulations and experiment 5 used a LightSpeed VCT scanner for real scans. For the computer-based experiments, a few low resolution images have been attempted and registration steps were explored for image reconstruction. A resolution target, USAF1951, was called from MATLAB and used to examine the resolving power before and after image processing based on Super Resolution algorithm. Image-image subtraction was used to compare pre-processing and post-processing images. The number of non-zero pixels was used to access the percentage of similarity. For the experiment using LightSpeed VCT scanner, a GE VCT QA phantom was used to test the performance of the technique.
Result
From the experiments using USAF1951, it was found that: the minimum resolvable line pairs had improved from family -1 element 6 to family 0 element 2 (2 elements improvement) after applying “Super Resolution Technique” as shown in the experiment 1. An xy directional shifting of the pre-processing images resulted in a better reconstructed image than x-axis shifting or y-axis shifting in terms of resolution, shown in the experiment 2. The experiment 3 concluded that the more the pre-processing images, the better the reconstructed image would be. The experiment 4 showed that the shifts of pre-processing images greater than the detector size could still result in a higher resolution image. The experiment 5 revealed that applying “Super Resolution Technique” to a real CT scanner could not give an obvious improvement in resolution, but the image background noise had reduced.
Conclusion
It was concluded that the “Super Resolution Technique” could improve the image resolution and reduce the background noise at expense of more imaging time and more dose from the additional view. In case of hardware upgrade of imaging device is not practicable, Super Resolution could help improve the image quality. === published_or_final_version === Medical Sciences === Master === Master of Medical Sciences
|
|---|
