| Summary: | Medical imaging is the main tool to extract a 3D modelling of the human body or specific organs within it. In order to accomplish this, various imaging modalities have been developed over the years, such as X-Ray Computed Tomography (CT), Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET). Each one is based on a particular energy source that passes through the body and on specific physical laws, which define the meaning of noise and the sensitivity of the imaging process. In all medical imaging systems the main goal is to increase resolution since higher resolution is a key factor in increased information content, which is critical for increased accuracy in the understanding of the anatomy and in the assessment of size and morphological structure of organs, for early detection of abnormalities, suspected pathologies and more.
In order to overcome the resolution limitations, one promising idea is to use signal processing techniques to enhance the spatial resolution. This approach proposes the acquisition of a high-resolution (HR) image from observed multiple low-resolution (LR) images. This image restoration approach is called super resolution (SR) image reconstruction (or restoration). It is the process of combining multiple low resolution images to form a high resolution image. The basic requirement in order to apply SR restoration techniques is the availability of multiple LR images captured from the same scene, which are sub-sampled (aliased) as well as shifted with subpixel precision. Each observed LR image is expressed as the result of a sequence of operators on the original HR image source, consisting of a geometrical warp, blurring and down-sampling.
The SR image reconstruction method consists of three stages, registration, interpolation and restoration (i.e., inverse procedure). In the registration stage, the relative shifts between LR images, with reference to a certain LR image, are estimated with fractional pixel accuracy. Accurate sub-pixel motion estimation is a very important factor in the success of the SR image reconstruction algorithm. Since the shifts between LR images are arbitrary, the registered HR image will not always match up to a uniformly spaced HR grid. Thus, non-uniform interpolation is necessary, to obtain a uniformly spaced HR image from a non-uniformly spaced composite of LR images. Finally, image restoration is applied to the up-sampled image to remove blurring and noise.
In order to evaluate the performance of SR reconstruction, a ‘simulate and correct’ approach to reconstruction is selected. First, simulated images of a computer generated phantom are formed and processed in order to comply with the observation model for the LR images. These are used as the images from which the HR image is constructed through the SR method. The iterative back-projection (IBP) algorithm suggested by Irani and Peleg has been chosen to be utilized, which belongs in the spatial domain methods and it is an easily and intuitively understood method. The results of the SR reconstruction are presented separately for the axial and the transaxial case. The evaluation relies on qualitative measures of image enhancement and on objective quantitative measures, such as the resolution (FWHM), the signal-to-noise ratio, the contrast ratio and the contrast-to-noise ratio.
The performed trials demonstrated improvement in both the axial and transaxial resolution. The super-resolution images also provide a significantly improved contrast ratio, which is important for improving sensitivity for detection of small details and features. The improvement in resolution can be achieved without using any hardware changes or any increase in the patient radiation procedure. An important contribution of super-resolution is also the reduction of partial volume effects in the reconstructed image. The loss in SNR, which is a typical characteristic of all resolution enhancement algorithms, was not that considerable to preclude the clinical application of super-resolution. The overall evaluation demonstrated that the SR reconstruction is a post-processing method, which can provide medical images of higher resolution and better contrast ratio, without increasing the amount of radiation or the duration of the scan.
|
|---|
