WO2006023354A1

WO2006023354A1 - Use of multiple pulse sequences for 3d discrimination of sub-structures of the knee

Info

Publication number: WO2006023354A1
Application number: PCT/US2005/028503
Authority: WO
Inventors: Saara Marjatta Sofia Totterman; Jose Tamez-Pena
Original assignee: Virtualscopics, Llc
Priority date: 2004-08-18
Filing date: 2005-08-10
Publication date: 2006-03-02
Also published as: WO2006023354B1; EP1792265A1

Abstract

To image substructures in the knee, two sets of image data are taken, e.g., T1 fat-suppressed and T2 non-fat-suppressed MRI images. The two sets of image data are fused to obtain a fused multispectral image volume. The image volume can be displayed or subjected to automatic segmentation and feature extraction to resolve substructure boundaries that are poorly resolved in one of the original sets of image data.

Description

USE OF MULTIPLE PULSE SEQUENCES FOR 3D DISCRIMINATION OF

SUB-STRUCTURES OF THE KNEE Reference to Related Application

[0001] The present application claims the benefit of U.S. Provisional Patent Application No. 60/602,315, filed August 18, 2004, whose disclosure is hereby incorporated by reference in its entirety into the present disclosure. Field of the Invention

[0002] The present invention is directed to a technique for imaging the knee and more specifically to a technique for fusing multiple sets of imaging data of the knee to improve delineation of boundaries within the knee. Description of Related Art

[0003] Imaging of the knee cartilage and meniscus are important in many diagnostic assessments of conditions ranging from sports injuries to progressive osteoarthritis. MRI is used to obtain direct visualization of the cartilage (Eckstein, et al, Osteoarthritis and Cartilage, 10, 922-928, 2002; Recht et al, Radiology 187, 473-478, 1993; Peterfy et al, Radiology 192, 485-491, 1994, Eckstein et al, Magnet. Reson. Med 36:256-265, 1996). Typically, a single MRI imaging protocol is employed to obtain a series of images forming a 3D image set. For example, Eckstein et al (2002) employed a spoiled 3D gradient-echo sequence (Flash = fast low angle shot) with selective water excitation (RF amplitude ratios 1-2-1 ; TR= 17.2 ms, TE= 6.6 ms; FA=20 degrees). The selection of pulse sequence is critical for optimizing the high contrast visualization of the objects of interest, while optimizing signal-to-noise, particularly to provide confidence in tracing the boundaries of structures like the cartilage and the meniscus, which have complicated 3D shapes. [0004] Other sequences have been selected for good visualization of the structures of the knee. For example, McGibbon et al (Osteoarthritis and Cartilage 11, 483-493, 2003) employed a frequency selective fat-suppressed 3D SPGR (spoiled gradient- recall in the steady state) pulse sequence with TR=55ms, TE=13.5ms and flip angle = 45 degrees).

[0005] Raynauld et al (Osteoarthritis and Cartilage 11, 351-360, 2003) employed a 3D FISP sequence with fat suppression and TR=42ms, TE=7ms, flip angle= 20 degrees.

[0006] Xia (Osteoarthritis and Cartilage 11, 473-474, 2003) emphasizes the need for short TE times in the imaging protocol, specifically with TE times around 3.2 ms for accurate delineation of the cartilage and a more precise measurement of cartilage thickness.

[0007] A problem exists, however, with the use of any single pulse sequence used to form images of the knee. There are multiple structures and fluids within the knee, each having its own characteristic appearance in any of Tl weighted images, T2 weighted images, proton density weighted images, gadolidium enhanced images, fat suppressed images, or other particular MR images.

[0008] For example, the Tl fat suppressed sequence provides good delineation of bone/cartilage boundary, but poor delineation of cartilage/synovium as well as cartilage/meniscus boundary. The T2 non-fat suppressed sequence, on the other hand, provides poor delineation of bone/cartilage boundary but excellent delineation of cartilage/meniscus, cartilage/synovium and cartilage/fluid boundaries. There is thus a tradeoff in the ability to image the above-noted boundaries. [0009] Those tradeoffs will be further detailed in the following sections: [0010] Normal Cartilage Appearance on Spin Echo [0011] The laminar appearance of weight-bearing cartilage in spin-echo images and its relationship to the directional orientation of collagen fibers and to the concentration of proteoglycans in the cartilage have been documented in the literature. That appearance is visually best demonstrated on proton density spin echo (SE) images which show an intermediate signal intensity deep layer and a higher signal intensity superficial layer. Due to its relatively short T₂ relaxation time, increasing T₂-weighting decreases the signal intensity of the deep layer. An increase in T₂ of the deep layer of the cartilage has been correlated to the early changes of degeneration of cartilage. That can vary from changes seen in increase in the T₂ in T₂ map through the visually detected high signal intensity areas in the T₂-weighted spin-echo images. [0012] While spin-echo imaging provides very good contrast for the cartilage and its contrast behavior is well understood, it has limitations in image resolution due to the limitation in signal to noise ratio.

[0013] Normal Cartilage Appearance on Gradient Echo

[0014] Three-dimensional gradient echo imaging with high signal to noise ratio allows use of higher resolution. However, in cartilage imaging with 3D gradient echo sequences, the emphasis has been to maximize the cartilage signal intensity or eliminate any nonuniformity in it with the purpose to provide maximal signal intensity for cartilage segmentation for cartilage volume and thickness measurements. [0015] OA Changes in Cartilage- Spin echo and gradient echo images [0016] Regarding osteoarthritis changes in the cartilage, clinical studies have shown that in PD and T₂W SE images, the irregularity of the surface of the cartilage as well as small cartilage defects are visualized. The 3D fat suppressed SPGR sequence obscures the surface changes; however, those are clearly visible in T₂W GRE images. [0017] Separating cartilage from Menisci [0018] The menisci, with their very short Ti, appear as very low signal intensity structures in clinically used spin-echo images. However, in TjW fat suppressed 3D SPGR images, shortening the TE increases the meniscal signal intensity to the degree that in very short TE images (TE=~3 ms), the meniscus appears as a high signal intensity structure which is inseparable from high signal intensity cartilage. The meniscal tears and meniscal degeneration, which are clinically evaluated using TiW or PD SE images, where they appear as high signal intensity, have that same appearance also in T[W 3D SPGR images. Further, the ligamentous structures and synovium cannot be visually separated from cartilage in the middle of the knee in fat suppressed or water excited SPGRE images.

Summary of the Invention

[0019] It will be apparent from the above that a need exists in the art for an imaging technique which provides good delineation of the above boundaries without the above-noted tradeoff. It is therefore an object of the invention to provide such a technique.

[0020] The invention is based on the inventors' discovery that, by fusing or combining the image information from two different MRI pulse sequences, one can improve the overall discrimination and delineation of the boundaries of the components of the knee.

[0021] For example, if the TR is kept constant, both SPGR and GRE images with TE shorter than 4 ms show the meniscus as a high signal intensity structure which is inseparable from cartilage. Increasing TE in both sequences decreases the signal intensity of the menisci, making them visually separable from cartilage.

[0022] Increasing the TE from 9 ms decreases the signal of the deep layer of the cartilage and starts showing the laminar appearance of the cartilage. That phenomenon is seen earlier with 3D GRE than 3D SPGR sequences.

[0023] When the TR and TE are kept constant and the flip angle is increased from 10° to 40°, the signal intensity of the deep layer of the cartilage decreases, bringing back the laminar appearance of the cartilage. However, the changes in the flip angle have to be more drastic than the changes in TE.

[0024] Increasing the flip angle from 10° to 40° increases the signal intensity of the fluid. At the same time, when the signal intensity of the cartilage decreases, there is an improvement in contrast between the cartilage and the fluid.

[0025] With increasing TE, and more so with increasing flip angle, the adjacent synovium becomes separable from cartilage. [0026] With increasing TE, the T₂ lesions in the cartilage become apparent. That can also be seen with an increasing flip angle, however, to a much lesser degree.

[0027] Based on results from the inventors' studies, a preferred embodiment uses fat suppressed 3D-GRE or SPGRE sequence with a TR/TE/flip angle of 39 ms/7ms/20° and 3D GRE sequence with TR/TE/flip angle of 29 ms/14 ms/40 when using the GE echo speed magnet. The length of TR in that protocol reflects the limitation of GE echo-speed magnet. On a Philips Intera, water excitation sequence can be acquired using the TR as short as 20 ms.

[0028] Thus, the overall steps in the present invention include:

[0029] - employing a first pulse sequence to image the knee and produce a 3D set of images (volume 1);

[0030] - employing a second pulse sequence to image the knee and produce a second

3D set of images (volume 2);

[0031] - registering the two volumes to produce a single multispectral image volume; and

[0032] - displaying the multispectral image volume with improved boundary delineation, or utilizing the multispectral volume in an automatic segmentation and feature extraction algorithm for analysis of the substructures.

[0033] Examples of registration and fusion and multispectral analysis are given in

WO 01/75483 by Tamez-Pena et al. Examples of automatic segmentation and feature extraction are given in U.S. Patent No. 6,169,817 by Parker et al. Both of the above references are hereby incorporated by reference in their entireties into the present disclosure. Brief Description of the Drawings

[0034] A preferred embodiment of the present invention will be set forth in detail with reference to the drawings, in which:

[0035] Fig. 1 shows a flow chart of operational steps in the present invention;

[0036] Fig. 2 shows images of the knee before and after fusion; and

[0037] Fig. 3 shows a schematic diagram of a system on which the invention can be implemented.

Detailed Description of the Preferred Embodiment

[0038] A preferred embodiment of the present invention will be set forth in detail with reference to the drawings.

[0039] A flow chart of operations performed in the preferred embodiment is shown in Fig. 1. In step 102, a first pulse sequence is used to image the knee and to produce a first set of 3D images. In step 104, a second pulse sequence is used to image the knee and to produce a second set of 3D images. In step 106, the two volumes are registered to produce a single multispectral image volume. The registration can be accomplished through techniques such as those disclosed above or through any other suitable techniques. The resulting single multispectral image volume is stored (e.g., in RAM or on persistent storage such as a hard drive). Then, either or both of steps 108 and 110 can be carried out. In step 108, the single multispectral image volume is displayed with improved boundary delineation. In step 110, an automatic segmentation and feature extraction algorithm, such as that disclosed above or any other suitable algorithm, is used to analyze substructures in the knee. [0040] As a particular example of the presnet invention, a human knee was imaged with the following two pulse sequences:

[0041] A Phillips 1.5T MRI scanner was employed. A human subject's knee was scanned using a knee surface coil. A first pulse sequence was employed with the following characteristics: GRE 3D scanning sequence with fat suppression, slice thickness 1.5mm, TR=39ms, TE=7.8ms, flip angle = 20 degrees. A second pulse sequence was used with the following characteristics: GRE 3D scanning sequence, slice thickness 1.5mm, TR=29ms, TE=14ms, flip angle = 40 degrees. The two images and the fused image are shown in Fig. 2. [0042] The results are summarized below in Table I below. Note that with any single pulse sequence, there are some critical boundaries that have poor definition, because of the particular appearance of different components in any single pulse sequence. However, in the fused image, each and every major boundary of importance can be seen with high confidence.

[0043] Table I

[0044] The fused image information is displayed in color, by assigning different color values to the information contained in each individual scan. The result has greater discrimination and contrast between different substructures which would otherwise be indistinct on conventional single sequence images of the knee.

[0045] Because of the above-noted resolution limitations of spin-echo imaging, it is preferred not to use spin-echo images for volumetric measurement for cartilage imaging. [0046] Fig. 3 shows a block diagram of a system 300 on which the present invention can be implemented. The system 300 receives the first and second sets of image data from one or both of an input device 302 for receiving previously stored image data or an input device 303, such as an MRI device as described above, that takes the image data. Thus, the system 300 can be integrated into an imaging device or can be a stand-alone device that receives the image data, e.g., over a communication line or on a storage medium. Either way, the input data are supplied to a workstation 304 having storage 306 (RAM, persistent storage such as a hard drive, etc.), a processor or multiprocessor unit 308, and graphics rendering capabilities 310. The graphics rendering output (e.g., the fused image or the result of segmenation and extraction) is supplied to an output 312, which can be a monitor for viewing the image, a printer, or any other suitable device. As will be understood in the art, the storage 306 is available for storing, temporarily or persistently, any or all of the first and second sets of image data, the fused image and the result of segmentation and extraction. [0047] As noted above, three-dimensional gradient echo imaging with a high signal to noise ratio allows use of higher resolution. However, in cartilage imaging with 3D gradient echo sequences, the emphasis has been to maximize the cartilage signal intensity or eliminate any nonuniformity in it with the purpose to provide maximal signal intensity for cartilage segmentation for cartilage volume and thickness measurements Therefore, when pulse sequence comparison studies were done, the TE' s or flip angles, which decreased the signal intensity of the deep layer to the extent that the laminar appearance of the cartilage started appearing, were considered disadvantageous. That is also the reason why a 3D SPGRE sequence was chosen and 3D GRE sequences, which provided T₂ information of the tissues were not considered as viable options. Those studies also concentrated on increasing the contrast between bone and cartilage and in eliminating chemical shift artifact. Therefore, TiW fat suppressed 3D SPGR sequence, which produced high signal cartilage contrasting well against the low signal intensity bone marrow, was chosen as an ideal cartilage imaging sequence. However, in the segmentation process, the cartilage should present good contrast against other adjacent structures as well. [0048] The inventors' early cadaver work on MR imaging of the wrist showed that cartilage, the ligaments and triangular fibrocartilage which histologically corresponds to menisci, were best seen and separated from each other when a 3D GRE sequence with TE of 15-18 and flip angle of 20-30 was used.

[0049] While a preferred embodiment of the present invention has been set forth in detail above, those skilled in the art who have reviewed the present disclosure will readily appreciate that other embodiments can be realized within the scope of the invention. For example, while MRI has been disclosed, other suitable imaging technologies can be used. Also, disclosures of specific numerical values, pulse sequences, and the like should be construed as illustrative rather than limiting. Therefore, the present invention should be construed as limited only by the appended claims.

Claims

We claim:

1. A method for imaging knee structures in a patient, the method comprising:

(a) using a first image taking procedure to take a first set of image data of the knee structures;

(b) using a second image taking procedure which is different from the first image taking procedure to take a second set of image data of the knee structures; and

(c) fusing the first set of image data and the second set of image data to form a set of fused image data.

2. The method of claim 1, further comprising storing the set of fused image data.

3. The method of claim 1, wherein the first image taking procedure and the second image taking procedure are magnetic resonance imaging procedures.

4. The method of claim 3, wherein the first image taking procedure uses a first pulse sequence, and wherein the second image taking procedure uses a second pulse sequence.

5. The method of claim 4, wherein the first pulse sequence is a Tl fat- suppressed pulse sequence, and wherein the second pulse sequence is a T2 non-fat- suppressed sequence.

6. The method of claim 5, wherein the first pulse sequence and the second pulse sequence have different flip angles.

7. The method of claim 1, further comprising displaying the set of fused image data as a fused image.

8. The method of claim 7, wherein the fused image is displayed in a plurality of colors.

9. The method of claim 8, wherein the plurality of colors comprise a first color representing a contribution to the fused image from the first set of image data and a second color representing a contribution to the fused image from the second set of image data.

10. The method of claim 1, further comprising performing automatic segmentation and feature extraction on the fused image data.

11. The method of claim 10, wherein the automatic segmentation and feature extraction are used to identify substructure boundaries in the knee structures.

12. The method of claim 1 1, wherein the substructure boundaries comprise a bone/cartilage boundary and at least one of a cartilage/meniscus boundary, a cartilage/synovium boundary, and a cartilage/fluid boundary.

13. The method of claim 1, wherein the fused image data comprise multispectral image data.

14. The method of claim 13, further comprising performing multispectral analysis on the multispectral image data.

15. A system for imaging knee structures in a patient, the system comprising: an input device for inputting a first set of image data of the knee structures and a second set of image data of the knee structures; and a processor, in communication with the input, for fusing the first set of image data and the second set of image data to form a set of fused image data.

16. The system of claim 15, further comprising storage, in communication with the processor, for storing the set of fused image data.

17. The system of claim 15, further comprising a display, in communication with the processor, for displaying the set of fused image data as a fused image.

18. The system of claim 17, wherein the fused image is displayed in a plurality of colors.

19. The system of claim 18, wherein the plurality of colors comprise a first color representing a contribution to the fused image from the first set of image data and a second color representing a contribution to the fused image from the second set of image data.

20. The method system claim 15, wherein the processor also performs automatic segmentation and feature extraction on the fused image data.

21. The system of claim 20, wherein the automatic segmentation and feature extraction are used to identify substructure boundaries in the knee structures.

22. The system of claim 21, wherein the substructure boundaries comprise a bone/cartilage boundary and at least one of a cartilage/meniscus boundary, a cartilage/synovium boundary, and a cartilage/fluid boundary.

23. The system of claim 15, wherein the fused image data comprise multispectral image data, and wherein the processor also performs multispectral analysis on the multispectral image data.

24. A system for imaging knee structures in a patient, the system comprising: an image taking apparatus for:

(a) using a first image taking procedure to take a first set of image data of the knee structures; and

(b) using a second image taking procedure which is different from the first image taking procedure to take a second set of image data of the knee structures; and a processor, in communication with the image taking apparatus, for fusing the first set of image data and the second set of image data to form a set of fused image data.

25. The system of claim 24, further comprising storage, in communication with the processor, for storing the set of fused image data.

26. The system of claim 24, wherein the image taking device is a device for taking magnetic resonance image data, and wherein the first image taking procedure and the second image taking procedure are magnetic resonance imaging procedures.

27. The system of claim 26, wherein the first image taking procedure uses a first pulse sequence, and wherein the second image taking procedure uses a second pulse sequence.

28. The system of claim 27, wherein the first pulse sequence is a Tl fat- suppressed pulse sequence, and wherein the second pulse sequence is a T2 non-fat- suppressed sequence.

29. The system of claim 28, wherein the first pulse sequence and the second pulse sequence have different flip angles.

30. The system of claim 24, further comprising a display, in communication with the processor, for displaying the set of fused image data as a fused image.

31. The system of claim 30, wherein the display displays the fused image in a plurality of colors.

32. The system of claim 31, wherein the plurality of colors comprise a first color representing a contribution to the fused image from the first set of image data and a second color representing a contribution to the fused image from the second set of image data.

33. The system of claim 24, wherein the processor also performs automatic segmentation and feature extraction on the fused image data.

34. The system of claim 33, wherein the automatic segmentation and feature extraction are used to identify substructure boundaries in the knee structures.

35. The system of claim 34, wherein the substructure boundaries comprise a bone/cartilage boundary and at least one of a cartilage/meniscus boundary, a cartilage/synovium boundary, and a cartilage/fluid boundary.

36. The system of claim 24, wherein the fused image data comprise multispectral image data, and wherein the processor also performs multispectral analysis on the multispectral image data.