Method for providing high resolution, high contrast fused MRI images

Abstract

This present invention relates to an MRI scanning assembly and a method for fusing MRI images of a target thereby generating and providing high resolution, high contrast fused MRI images. The MRI images of the target are generated by different MRI devices operating at different magnetic field intensities. A method is also described for fusing MRI images generated by an MRI device operating with different operational parameters and operational protocols.

Claims

1. A magnetic resonance imaging (MRI) scanning assembly, comprising: a MRI scanning device operating at a first magnetic field intensity and generating a first plurality of MRI images of a target; the MRI scanning device operating at a second magnetic field intensity and generating a second plurality of MRI images of said target; and a processing unit configured to: identify at least two regions of interest in said first plurality of said MRI images, determine a monochrome level scaling factor between said at least two identified regions of interest, said monochrome level scaling factor is selected from a grey level, a red scale, green scale, and a blue scale, fuse said first plurality of MRI images and second plurality of MRI images based on the monochrome level scaling factor to generate a clear image representation of at least a portion of said target.

2. The MRI scanning assembly according to claim 1, wherein said first magnetic field intensity is less than 1.5 Tesla and said second magnetic field intensity is greater than 5 Tesla.

3. The MRI scanning assembly according to claim 1, further comprising at least one selected from a group consisting of (a) an MRI selector for selecting said MRI scanning device; (b) an image display unit for generating and displaying said at least said portion of said target.

4. A method for fusing MRI images of a target comprising: receiving a first plurality of MRI images of a target, said first plurality of MRI images generated by a MRI scanning device at a first magnetic field intensity; receiving a second plurality of MRI images of the target, said second plurality of MRI images generated by said MRI scanning device at a second magnetic field intensity; identifying at least two regions of interest in said first plurality of MRI images; determining a monochrome level scaling factor between said at least two identified regions of interest, said monochrome level scaling factor is selected from a grey level, a red scale, green scale, and a blue scale; fusing said first plurality of MRI images and said second plurality of MRI images based on the monochrome level scaling factor to generate a clear image representation of at least a portion of said target.

5. The method for fusing MRI images according to claim 4, wherein said first magnetic field intensity is less than 1.5 Tesla and said second magnetic field intensity is greater than 5 Tesla.

6. The method for fusing MRI images according to claim 4, wherein each one of said plurality of MRI images is associated with at least one member of a group of MRI device parameters.

7. The method for fusing MRI images according to claim 6, wherein said group of MRI device parameters is selected from the group consisting of MRI scanning parameters and MRI scanning protocols and any combination thereof and wherein said group of MRI scanning parameters comprises at least one of the following: pixel size and dwell time and said group of MRI scanning protocols comprises at least one of the following protocols: T1-weighted scan protocols and T2-weighted scan protocols.

8. The MRI scanning device according to claim 1, wherein said target is selected from comprises food products or industrial products.

9. The MRI scanning device according to claim 1, wherein said MRI scans comprise MRI scans for searching and investigating fluids.

10. The method for fusing MRI scans according to claim 4, wherein target is selected from food products or industrial products.

11. The MRI scanning assembly according to claim 1 wherein said target is selected from the group consisting of a group of cancerous cell, at least one anatomical organ, at least one cancerous anatomical organ and any combination thereof.

12. The method for fusing MRI scans according to claim 4 wherein said target is selected from the group consisting of a group of cancerous cell, at least one anatomical organ, at least one cancerous anatomical organ and any combination thereof.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) A preferred embodiment of the current invention is described hereinbelow with reference to the following drawings:

(2) FIG. 1 shows an MRI scanning assembly for fusing scanned MRI images generated at a low magnetic field intensity and generated at a high magnetic field intensity, according to a preferred embodiment of the present invention;

(3) FIGS. 2A and 2B compare a single slice generated from MRI scans of the rodent at the low magnetic field intensity and the high magnetic field intensity, respectively, in accordance with a preferred embodiment of the present invention;

(4) FIGS. 3A and 3B compares histograms of MRI images of the target taken at a high magnetic field intensity of approximately 1 Tesla and MRI images taken at a magnetic field intensity of approximately 7 Tesla, respectively, in accordance with a preferred embodiment of the present invention;

(5) FIGS. 4A and 4B compares the original slice from the high field MRI scan with the enhanced MRI scan, respectively, in accordance with a preferred embodiment of the present invention;

(6) FIGS. 5A, 5B, 5C and 5D show MRI images of a cross sectional slice of a cucumber typically generated by the low intensity magnetic field device at different in-slice pixel sizes, respectively, in accordance with a preferred embodiment of the present invention;

(7) FIG. 6 present a flow chart of a typical procedure for fusing multiple sets of images of a given volume of the target into a single enhanced image, in accordance with a preferred embodiment of the present invention, and

(8) FIGS. 7A-7C compare in grey and color, the results of combining multi-resolutions images as grey and color images, in accordance with a preferred embodiment of the present invention.

DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT OF THE PRESENT INVENTION

(9) Reference is now made to FIG. 1, which shows an MRI scanning assembly 10 for fusing scanned MRI images, according to a preferred embodiment of the present invention. The MRI scanning assembly 10 includes, inter alia, MRI scanning devices 12.sub.i=j and 12.sub.i=k, where 1≦i≦N and N is the number of MRI devices available for generating the MRI scans at different magnetic field intensities.

(10) Typically, in the MRI scanning assembly 10, the MRI scans of a target 14 are performed at different magnetic field intensities, such that the MRI devices 12.sub.i=j and 12.sub.i=k scan the target 14 at a different magnetic field intensities. In order for the devices 12.sub.i=j and 12.sub.i=k to scan the target 14, the target 14 is located on a moving platform 16 which transports the target 14 from a scanning location 18 at the device 12.sub.i=j to a scanning location 20 at the device 12.sub.i=k.

(11) The MRI scanning assembly 10 also includes an MRI selector 22 which selects the MRI devices 12.sub.i=j or 12.sub.i=k for scanning the target 14 at the scanning locations 18 and 20, respectively. Thus, for example, the selector 22 instructs the MRI device 12.sub.i=j to scan the target at the low magnetic field intensity.

(12) Typically, for example, the low magnetic field intensity is in a range of approximately 0.5 Tesla to approximately 1.5 Tesla. The device 12.sub.i=j scans the target 14 and generates a plurality of MRI target scans 24.

(13) Following the “low-field scan”, the target 14 is transported to the location 20 and the magnetic field selector 22 instructs device 12i=k to scan the target 14 at a higher magnetic field intensity scan. For example, for the “high-field scan”, the high magnetic field intensity is typically in a magnetic field intensity range of approximately 3 Tesla to 16 Tesla. The device 12.sub.i=k scans the target 14 and generates a plurality of MRI target scans 27.

(14) It is appreciated that the MRI devices 12.sub.i=j and 12.sub.i=k generate a plurality of MRI scans of the target 14 at each magnetic field intensity, as required. The plurality of scans is generated by changes in the MRI device parameters, such as the pulse sequence or a scanning protocol. The pulse sequence determines the dynamics of the magnetic moments and therefore determines the measured signal intensity.

(15) On completion of the respective MRI scans of the target 14, the MRI devices 12.sub.i=j and 12.sub.i=k transfer the scanned MRI data 24 and 27, respectively, to a processing unit 23 for processing, analyzing and interpretation by an operator. The processing unit 23 allows the operator to reconstruct the MRI scans, which are taken in “k-space” and convert the MRI images into “real-space”. The analysis of the reconstructed MRI scans is conducted in accordance with an analysis tool, as discussed below. The operator manually or with the aid of a computing device (semi-automatically), analyzes the reconstructed MRI scans. The analysis involves, for example, the operator outlining and segmenting a region of the reconstructed scan, such as an organ or a pathological feature. In addition, the operator, typically, makes measurements of the segmented region, such as the volume of region and the average signal strength.

(16) The processing unit 23 forwards the results of the analysis to an MRI image display device 25.

(17) Alternatively, the plurality of MRI scans 24 and 27 are generated by a single MRI scanning device 12.sub.i=1, which includes a typical device is the MRD scanning device, as described in U.S. Pat. No. 7,400,147, which is owned by the owner of the present application and is incorporated herein by reference in its entirety.

(18) In the single MRI device. different MRI images are generated by using different device parameters, such as pixel size and dwell time and different scanning parameters, such as a T1-weighted scanning protocol and a T2-weighted protocol.

(19) FIGS. 2A and 2B compares the MRI the resolution of MRI images obtained at a magnetic field intensity of approximately 7 Tesla with the resolution of MRI images obtained at a magnetic field intensity of approximately 1 Tesla, respectively, in accordance with a preferred embodiment of the present invention;

(20) Reference is now made to FIGS. 2A and 2B, which compares generated MRI scans of the rodent at the low magnetic field intensity and the high magnetic field intensity, respectively, in accordance with a preferred embodiment of the present invention. FIG. 2A shows a scan slice 36 of the small rodent with a tumor 38, taken at a magnetic field intensity of approximately 1 Tesla. FIG. 2B shows a scan slice 40 of the rodent with a tumor 39 at a magnetic field intensity of approximately 7 Tesla, at a different slice of the rodent.

(21) The scan slices 36 and 40 in FIGS. 2A and 2B, respectively, are not the same scan slice. Thus, the scan slices 36 and 40 are non-coherent. The MRI slices 36 and 40 do not include the same content and portion of the rodent. Thus, pixel-wise registration is not possible. The low resolution scan includes a higher SNR level than that of the high resolution scan (FIG. 2A). In order to enhance and improve the contrast between tissue masses of the high resolution scan, preferably, the low resolution scan (FIG. 2B) is fused with the high resolution scan of FIG. 2A thereby enhancing the contrast in FIG. 2B.

(22) The MRI scan in FIG. 2A is a low-resolution image of approximately 300 μM pixel size and with high contrast between certain tissues. FIG. 2A also shows a corresponding grey level scale 42. FIG. 2B shows a high resolution scan of approximately 100 μM resolution with a high SNR and low contrast between the corresponding high contrast tissues of FIG. 2A and a corresponding grey level scale 44.

(23) If, for example, the operator wishes to improve the contrast of a tumor area 41 (FIG. 2B), relative to the surrounding regions, thereby distinguishing and classifying the tumour area 41. The operator decides to enhance the contrast of the tumor area 41 and selects a corresponding area 38 in FIG. 2A.

(24) In operation, the operator and/or processing unit 23 identifies two corresponding regions of interest in FIGS. 2A and 2B that are required to be enhanced in FIG. 2B. For example, the operator and/or processing unit 23 selects the tumor 38 and a fatty white portion 46 in FIG. 2A. Similarly, preferably, the operator and/or processing unit selects corresponding regions 41 and 39 (white region) in FIG. 2B.

(25) In operation, a tissue (or tissues) which require enhancing the contrast thereof, is selected. The entire tissue region thereof is identified, typically, by a segmentation algorithm based on a previously defined seed voxel, as is known in the art. It is appreciated that the tissue region can also be identified manually by the operator. This identification procedure is performed separately for the images 36, (FIG. 2A) and 40 (FIG. 2B), since the images 36 and 40 cannot be registered. The images 36 and 40 are MRI images of different slices of the target 14. Now a reference region must be segmented in each of the images 3A and 3B. A reference region, such as a marker reference or a tissue reference, such as muscle portion 47 in FIG. 2A and a muscle portion 48 in FIG. 2B is identified and selected. It is appreciated that the reference tissues 47 and 48 are selected since these reference tissues images appear relatively similar in both spatial extent and in terms of signal strength.

(26) It is appreciated that for determining the reference regions the segmentation need not be accurate and does not have to include the entire spatial area of “reference object”. For each image, the reference object is used to generate a mean reference signal strength by averaging the signals of its included voxels.

(27) For the MRI scan images 36 and 40, a ratio between the mean grey levels of the segment to be enhanced with a reference segment is calculated from:
Ratio_1=mean (segment 38, image 1)/mean(ref segment 47 im 1, Image 1),
wherein im 1 is image 36 in FIG. 2A.

(28) For the MRI scan image 40, a second ratio is calculated as a mean of the grey values:
Ratio_2=mean(segment 41, image 2)/mean(ref segment 48 im 2, Image 2),
wherein im 2 is image 46 in FIG. 2A.

(29) From FIG. 2A, Ratio_1 is approximately 1.8 and from FIG. 2b, Ratio_2is approximately 1.

(30) A Scaling Factor SF is calculated from:
SF=RATIO_1/RATIO_2.

(31) Based on the mean grey values determined for the segments 36 and 40, SF is determined to be approximately 1.8 for the tumor segment.

(32) The grey values of the voxels included in segment 41 of the MRI scan image shown in 40 of FIG. 2B will be scaled by a value of 1.8 to form an enhanced high resolution as in FIG. 4B. Similar scaling is performed on other high contrast segments of image 36 of FIG. 2A. Reference is now made to FIGS. 3A and 3B, which compare histograms of MRI images of the target taken at a low magnetic field intensity of approximately 1 Tesla and MRI images taken at a high magnetic field intensity of approximately 7 Tesla, respectively, in accordance with a preferred embodiment of the present invention.

(33) In FIG. 3A, the pixel populations of the image 36 are shown as a function of the Grey Level (GL) for two different portions of the rodent in the low-field scan. In FIG. 3A, the grey level histograms for a tumor 38 (FIG. 2A) and the reference rodent muscle region 47 (FIG. 2A) are identified by the respective shaded portions 28 and 30, respectively. The y-axis is the proportion of each GL relative to the overall population of pixels in the studied segment (regions 38 and 47).

(34) FIG. 3A shows that the tumor 38 and muscle regions 47 are clearly distinguishable by their different grey level values. The separable regions 28 and 30, due to the high contrast of the MRI scans obtained at the low magnetic field intensity.

(35) In FIG. 3B, the grey level histograms for the tumor 41 and the reference rodent muscle region 48 are identified by the respective shaded portions 32 and 34, respectively.

(36) In FIG. 3B, the pixel fractions are compared as a function of the Grey Level (GL) for the tumor 39 (FIG. 2B) and the reference muscle region 48 (FIG. 2B) of the rodent. In FIG. 3A, the grey level histograms for the tumor 38 and the reference rodent muscle region 47 are identified by the respective shaded portions 28 and 30, respectively.

(37) FIG. 3B shows that the tumor and muscle regions are not clearly distinguishable by their different grey level values, due to the relatively low relative contrast of the MRI scans obtained at the high magnetic field intensity.

(38) Reference is now made to FIGS. 4A and 4B, which compares the high resolution high-field MRI scan and the enhanced high resolution high-field scan, respectively, in accordance with a preferred embodiment of the present invention.

(39) FIG. 4A shows the image 50 from the high-field scan and indicates regions of interest 52 and 54 as well as a reference region 56. FIG. 4B shows the results of scaling the image 40 (FIG. 2B) with the SF's determined from the low field and high field MRI scans. A tumor region 62, a white region 64 and a reference region 66 are clearly distinguishable and have a high level of contrast with the remaining portions of the image 60.

(40) FIGS. 4A and 4B shows the contrast enhancement of the two regions, namely, the tumor region and the fat tissue region (white region).

(41) Thus, using this method a clear image of rodent slice is obtained with the tumor and fatty region being well distinguished from other tissues imaged in the slice.

(42) Reference is now made to FIGS. 5A, 5B, 5C and 5D, which show MRI images of a cross sectional slice of a cucumber typically generated by the low intensity magnetic field device at different in-slice pixel sizes, respectively, in accordance with a preferred embodiment of the present invention;

(43) FIG. 5A shows the cross section image of the cucumber 200 generated at an in-slice pixel size of 0.25 mm (high resolution) and a group of cucumber seeds 204 are clearly distinguishable from a cucumber background 202. However, due to the low SNR, the image of the cucumber seeds 204 is not clearly distinguishable from the cucumber background 202. A group of seeds 206 are not clearly distinguishable from the background 202. In addition, FIG. 5A shows a group of seeds 201 located on the periphery of the cucumber 200.

(44) FIG. 5B shows a cross section image of the cucumber 1000 generated at an in-slice pixel size of 0.5 mm (medium resolution) and the group of seeds 1204 is clearly imaged. The image of the group of seeds 1206 is clearer. Due to edge effects, the border between the cucumber flesh 1202 and the group of cucumber seeds 1204 is not clearly defined. The group 206 is not clearly distinguishable from the cucumber background 202. Due to the decrease in the resolution, FIG. 5B does not clearly identify a group of seeds located on the periphery of the cucumber 1000.

(45) FIG. 5C shows a cross section image of the cucumber 2000 generated at an in-slice pixel size of 1 mm (low resolution) and groups of cucumber seeds 2204 and 2206 are not clearly seen and the image is very blurred. Due to the further decrease in the resolution, FIG. 5C does not clearly identify a group of seeds located on the periphery of the cucumber 1000.

(46) FIG. 5D shows a combined image 3000 of the high resolution (0.25 in-slice pixel size) and medium resolution (0.5 mm in-slice pixel size). The group of seeds 3214 is distinguishable from the cucumber background 3212 and the group of seeds 3216 is barely distinguishable from the cucumber background 3212. However, due to noise, the edges of the seeds 3214 and 3216 are not clearly discernible. FIG. 5D shows a group of seeds 3218 located on the periphery of the cucumber 3000. To summarize the resolution in 3000 in FIG. 5D has the full resolution of cucumber 200 in FIG. 5A.

(47) Reference is now made to FIG. 6, which presents a flow chart of a typical procedure 300 for fusing multiple sets of images of a given volume of the target 14, into a single enhanced image, in accordance with a preferred embodiment of the present invention. The procedure 300 is controlled by the processing unit 23, wherein the images are taken at the same slice of the target 14.

(48) In step 302, the MRI scanned images are interpolated in order to generate voxels of the same geometrical voxel size, as is known in the art.

(49) In step 304, registration of the images acquired from the same acquisition mode is performed. The registration procedure ensures that the voxel representations of the images to be fused represent the same region of interest of the target.

(50) In step 306, the registered images are averaged to form a single image for each acquisition mode. This image includes a multiplicity of slices.

(51) In step 308, the combined images from the distinct acquisition modes used to image the registered target. A typical registration method is “The Lukas-Kanade Optical Flow Method”, as is known in art and described in “An Iterative Image Registration Technique with an Application to Stereo Vision”, B. D. Lucas and T. Kanade (1981), published in the Proceedings of Imaging Understanding Workshop, pages 121-130. Since the distinct image acquisition modes may have a different appearance, other methods known in the art for registering multi-modality images may be used. These can be based on maximizing mutual information of images patches as is known in the art.

(52) In step 310, the registered MR images of different acquisition modes are fused according to any of the well know fusion methods. In the following steps of FIG. 6, a method that is suited to variable resolution images acquisitions is outlined:

(53) In order to fuse the different acquisition modes of averaged images, they are divided into two types: high resolution images and low resolution images. The high resolution images are combined to form a single monochrome image as follows:

(54) The pixel values are combined using some weighting which can be assigned by a variety of methods, such as a principal component analysis. The method is known in the art and described in “Principal Component Analysis”, by I. T. Jolliffe, Series: Springer Series in Statistics, 2nd ed., Springer, N.Y., 2002, XXIX, 487 p. 28 illus. ISBN 978-0-387-95442-4. This combined monochrome image controls the brightness and/or intensity of the fused colored image while the low resolution images will control the spectral resolution of the fused image.

(55) The steps for fusing these high resolution and low resolution images to colored image are further outlined in FIG. 6.

(56) In step 314, each low-resolution image acquisition mode is assigned a color channel: for example, red, green and blue for three acquisition modes. The low resolution image is transformed to the HSV (hue, saturation, value) basis.

(57) In step 318, the intensity channel (value) is associated with the high resolution monochrome image and/or combined with the low-resolution intensity channel, for example by the Brovey method, as is known in the art.

(58) In step 320, the resulting image is transformed back to RGB space to form a colored fused final image of the target.

(59) Reference is now made to FIGS. 7A-7C, which compare in grey and color, the results of combining multi-resolutions images as grey and color images, in accordance with a preferred embodiment of the present invention.

(60) FIG. 7A shows a high resolution image 500 of a cross section of a cucumber 501 and groups of cucumber seeds 502 and 504. FIG. 7A is similar to the high resolution scan shown in FIG. 5A.

(61) FIG. 7B shows a high resolution image 520 of a cross section of the cucumber 501 and groups of cucumber seeds 502 and 504. In the image 510, the high resolution image (FIG. 5A), the medium resolution image (FIG. 5B) and the low resolution image (FIG. 5C) are combined by a IHS method as described in “Application of the IHS Color Transform to the Processing of Multisensor Data and Image Enhancement”, (Haydn, R., Dalke, G. W. and Henkel, J.:. Proc. of the International Symposium on Remote Sensing of Arid and Semiarid Lands, Cairo, pp. 599-616, 1982.), The image 520 is clearer than the colored image 510.

(62) FIG. 7C shows a high resolution image 520 of a cross section of the cucumber 501 and groups of cucumber seeds 502 and 504. In the image 520, the high resolution image (FIG. 5A), the medium resolution image (FIG. 5B) and the low resolution image (FIG. 5C) are combined by the Brovey method, as is known in the art (step 314, FIG. 6). The image 530 clearly distinguishes between the groups of cucumbers 502 and 504 and the groups of cucumbers 502 and 504 are clearly distinguishable from the cucumber background. 501.

(63) In the foregoing description, embodiments of the invention, including preferred embodiments, have been presented for the purpose of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise form disclosed. Obvious modifications or variations are possible in light of the above teachings. The embodiments were chosen and described to provide the best illustration of the principals of the invention and its practical application, and to enable one of ordinary skill in the art to utilize the invention in various embodiments and with various modifications as are suited to the particular use contemplated. All such modifications and variations are within the scope of the invention as determined by the appended claims when interpreted in accordance with the breadth they are fairly, legally, and equitably entitled.