Patent Application
20080144959
1. Field of the Invention
The present disclosure relates to image processing, and more particularly to a system and method for pre-processing scans as a preparation for segmentation.
2. Description of Related Art
Computed Tomography (CT) has dominated the area of noninvasive coronary imaging, in part due to the ability to segment and visualize the coronaries automatically form high resolution, high contrast-to-noise CT data. However, an equivalent in Magnetic Resonance (MR) datasets is very difficult due to the inherent qualities of MR images: lower contrast, lower spatial resolution, and more noise, including motion.
Therefore, a need exists for a system and method for pre-processing MR scans as a preparation for segmentation
According to an embodiment of the present disclosure, a computer-implemented method for pre-processing image data of a three-dimensional volume includes providing the image data of a vessel, applying a super-sampling filter to the image data to generate super-sampled image data having an increased resolution as compared to the image data, applying an unsharp masking filter to the super-sampled image data for increasing contrast of an edge of the vessel, applying a de-noising filer for removing noise surrounding the edge after applying the unsharp masking filter, and storing the image data after applying the de-noising filter thereto.
Applying the super-sampling filter comprises upsampling the image data for increasing the resolution of the image data, determining a Fourier transform of the image data after upsampling, applying a low-pass filter to the Fourier transform, multiplying voxel values of the image data by a Fourier gain after applying the low-pass filter, applying an inverse Fourier transform (FT) to the image data to determine an interpolated image after multiplying voxel values, and outputting the interpolated image.
Upsampling comprises inserting a zero value between each pair of adjacent voxels of the image data.
Applying the unsharp masking filter comprises filtering the image data by:
where a negative of a Laplacian filter α controls a shape of the Laplacian, with a default α=0.2.
The de-noising filter is edge-preserving.
The method further includes segmenting stored image data.
According to an embodiment of the present disclosure, a program storage device is provided readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for pre-processing image data of a three-dimensional volume. The method steps include providing the image data of a vessel, applying a super-sampling filter to the image data to generate super-sampled image data having an increased resolution as compared to the image data, applying an unsharp masking filter to the super-sampled image data for increasing contrast of an edge of the vessel, applying a de-noising filer for removing noise surrounding the edge after applying the unsharp masking filter, and storing the image data after applying the de-noising filter thereto.
According to an embodiment of the present disclosure, a computer-implemented method for pre-processing Magnetic Resonance (MR) image data includes providing the MR image data, super-sampling the MR image data to generate super-sampled MR image data, increasing contrast of an edge in the super-sampled MR image data, removing noise about the edge to generate pre-processed MR image data, and segmenting the pre-processed MR image data.
Super-sampling includes upsampling the MR image data for increasing a resolution of the image data, determining a Fourier transform of the MR image data after upsampling, applying a low-pass filter to the Fourier transform, multiplying voxel values of the MR image data by a Fourier gain after applying the low-pass filter, applying an inverse Fourier transform (FT) to the MR image data to determine an interpolated image after multiplying voxel values, and outputting the interpolated image, wherein the interpolated image is the super-sampled MR image data.
Preferred embodiments of the present invention will be described below in more detail, with reference to the accompanying drawings:
Segmentation of Magnetic Resonance Imaging (MRI) scans of coronary arteries is made difficult by low contrast, high noise, and low resolution conditions typically associated with the MRI scans. To improve segmentation results, according to an embodiment of the present disclosure, a set of filters is applied to the MRI scans for pre-processing before segmentation.
According to an embodiment of the present disclosure, an optimized pre-processing method substantially improves MRI scans for use in segmentation, and provides a basis for semi-automatic segmentation of the coronary tree in MR. User interaction is also minimized, reducing the number of seed points needed. Pre-processing filters are implemented, which run sequentially on the MRI data for improving segmentation.
Referring to
In MRI, the diameter of the coronary vessel can be as low as 2 pixels in basal slices. Super-sampling 102 to a higher resolution adds more pixels and facilitates front propagation throughout the entire artery. The unsharp masking filter 109 enhances the edges of the vessel to substantially prevent leaking. De-noising 110 is performed by using a 3D anisotropic diffusion filter for removing the noise surrounding the vessel.
The resulting image can then be processed by a segmentation method 111. The results of the segmentation method include a vessel centerline, which can then be used to navigate the original data.
Referring to the super-sampling 102; super-sampling 102 comprises upsampling 103, a Fourier transform (FT) 104, a low-pass filter 105, multiplication by a Fourier gain 106, an inverse Fourier transform 107, and outputting an interpolated image 108.
Upsampling 103 includes inserting a zero value between each pair of voxels of the image data, thereby increasing a resolution of the image. A Fourier transform in applied to upsampled image data. The low-pass filter 105 is applied to a Fourier transform of the image data. The image data output by the low-pass filter is multiplied by a gain (e.g., 8) to increase brightness. The image is resorted by inverse Fourier transform. The interpolated image, the output of the inverse Fourier transform, is output 108 to the unsharp masking filter 109.
Referring to the unsharp masking filter 109; the filter may be expressed as:
such that, a negative of the Laplacian filter α controls the shape of the Laplacian, with a default α=0.2.
Referring to the de-noising filter 110; the de-noising filter 110 detects edges and interpolates along the edges. The de-noising filter 110 is edge-preserving, e.g., the de-noising filter 110 does not interpolate across the edges. One of ordinary skill in the art would recognize that various de-noising filters may be implemented.
Experiments implementing a method according to an embodiment of the present disclosure used five datasets (n=5), which were navigator gated, axial whole-heart coronary MR scans acquired using a TrueFISP pulse sequence, using T2-preparation, on a Siemens 1.5T Avanto® scanner. The parameters were FOV/Matrix=190×320, TR/TE=3.9/1.5 ms, Flip angle=90, 120 slices, 1 mm thick, TA=10:10 mins.
For each dataset, the centerline coordinates of the vessel were overlaid on a volume-rendered image, for example, see
Although the LCA (left coronary artery) and LCX (left circumflex artery) needed roughly 5 to 7 seedpoints, combined, to track through all the branching, segmentation error was minimal.
According to an embodiment of the present disclosure, pre-processing the MR images prior to segmentation improves segmentation results. In each of the five datasets, the RCA was successfully segmented with only one user-defined seedpoint. Overall, the pre-processing increased the CNR (contrast-to-noise ratio) and the spatial resolution of the MR volumes, as a positive step toward fully satisfying the challenge of the coronary segmentation in MR.
Referring to block 111 and image segmentation; for an exemplary segmentation of the coronary arteries, a multi-layered fast-marching method may be employed. The method can be divided into three hierarchical levels, each comprised of interchangeable sub-methods: the Voxel level, the Segment level, and the Tree level.
At the Voxel level, two scores are determined, including a multiscale vesselness measure and a gray-level statistics score. The multiscale vesselness measure represents the geometric similarity to a tubular shape, and is determined according to the second-order variations of the Hessian matrix. The gray-level statistics score is determined based on an adaptive estimation of the intensity distribution in the vessel. The weight parameters yielding the optimal segmentation results for MR are found. After multiplication by the weights, these two scores are summed, and a combined score s(p)[0→1] is determined. If the combined score is above a threshold T, then the voxel is accepted into the segmentation.
The Segment level partitions the segmentation mask into representative units, using shape analysis.
At the Tree level, arterial branching points are detected and the actual tree structure of the arterial tree is reflected.
One of ordinary skill in the art would appreciate that other segmentations may be applied to the pre-processed image data according to an embodiment of the present disclosure.
It is to be understood that the present invention may be implemented in various forms of hardware, software, firmware, special purpose processors, or a combination thereof. In one embodiment, the present invention may be implemented in software as an application program tangibly embodied on a program storage device. The application program may be uploaded to, and executed by, a machine comprising any suitable architecture.
Referring to
The computer platform 401 also includes an operating system and microinstruction code. The various processes and functions described herein may either be part of the microinstruction code or part of the application program (or a combination thereof), which is executed via the operating system. In addition, various other peripheral devices may be connected to the computer platform such as an additional data storage device and a printing device.
It is to be further understood that, because some of the constituent system components and method steps depicted in the accompanying figures may be implemented in software, the actual connections between the system components (or the process steps) may differ depending upon the manner in which the present invention is programmed. Given the teachings of the present disclosure provided herein, one of ordinary skill in the related art will be able to contemplate these and similar implementations or configurations.
Having described embodiments for a system and method for pre-processing scans as a preparation for segmentation, it is noted that modifications and variations can be made by persons skilled in the art in light of the above teachings. It is therefore to be understood that changes may be made in embodiments of the present disclosure that are within the scope and spirit thereof.
This application claims the benefit of Provisional Application No. 60/726,598 filed on Oct. 14, 2005 in the United States Patent and Trademark Office, the contents of which are herein incorporated by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 60726598 | Oct 2005 | US |
