Methods for Two Dimensional Nuclear Magnetic Resonance Imaging

a. Field of the Invention

The present invention relates to methods for obtaining improved two dimensional nuclear magnetic resonance (NMR) images from those disclosed in the art. The use of time varying magnetic field gradients such that the frequency history of the spins is every element of a two dimensional array is distinguishable from that of every other element in that array have been described in the art. However, the methods described in the art do not, in fact, provide tractible approaches for obtaining two dimensional NMR images from the use of such time varying magnetic field gradients.

b. Description of the Prior Art

A number of NMR techniques for spacial mapping of points, lines, and planes in a sample of matter are known in the art. The particular procedure whereby the use of time varying magnetic field gradients such that the frequency history of each element in an array is distinguishable from that of every other element in that array has been suggested in the patent to Likes (U.S. Pat. No. 4,307,343).

However, the methods described by Likes suffer from a number of shortcomings. First, Likes fails to recognize the necessity of an appropriate weighting function which is required for image reconstruction, and is necessitated by the use of gradient magnetic fields whose trajectories in phase space do not sample equal areas of that space during equal time intervals.

For example, beginning at column 10, line 27, Likes discloses the use of Lissajous, "bull's eye" and rosette trajectories as the time varying magnetic field gradients to achieve spatial differentiation. However, none of these trajectories map equal areas of space over equal time intervals. Accordingly, inaccuracies result from the use of the methods disclosed in the Likes patent to reconstruct the image, without compensating for the nonlinear relationship of the trajectory's path in phase space with respect to time.

It is therefore an object of the present invention to provide novel methods for NMR mapping of two dimensional samples of matter whereby magnetic field gradients map equal areas of phase space in equal intervals of time.

It is yet another object of the present invention to provide methods for NMR mapping whereby magnetic field gradients may equal areas of phase space in unequal intervals of time, which times can be obtained from an appropriate weighing function.

It is yet a further object of this invention to provide novel reconstruction methods for reconstructing two dimensional images from NMR information obtained from samples of matter while utilizing gradient magnetic fields having trajectories that do not sample equal areas of phase space during equal intervals of time.

Improvements to nuclear magnetic resonance imaging methods are presented. Standard nuclear magnetic resonance imaging methods comprise sensitizing a two dimensional slice of an object in the presence of a static magnetic field. Within this slice, the magnetization of the nuclear spins is caused to be rotated at each point transverse to the static field. A time varying magnetic field gradient is then applied to the sensitized two dimensional slice, and electromagnetic energy radiated from the two dimensional slice is detected. Spin information for each area element within a slice is then reconstructed from the received radiated electromagnetic energy. Each of these above steps are generally known to one of ordinary skill in the art, as disclosed by the Likes patent and the articles referenced therein. In one aspect, the improvement is the use of an appropriate weighting function in the reconstruction to weight the received radiated electromagnetic energy signals to compensate for the use of time varying magnetic field gradients that do not sample equal areas of phase space with equal time. In another independent aspect, almost any magnetic field gradient function which substantially covers two dimensional phase space may be used, that does not sample equal areas of phase space in equal time periods, and a procedure defined by which equal areas of phase space are mapped by sampling at unequal times. In yet another aspect, a uniquely specified class of time varying magnetic field gradient trajectories are defined which map equal area elements of two dimensional phase space in equal time intervals.

FIG. 1 is a graphic depiction of phase space parameters.

FIG. 2 is a graphic depiction of a trajectory which maps unequal areas of phase space over equal time periods.

FIG. 3 is a graphic depiction of an Archimedian spiral trajectory in phase space.

All nuclei with magnetic moments, an example being the hydrogen nucleus, when placed in a static magnetic field, precess (at a frequency .omega..gamma.) about the magnetic field direction. Applying an effective 90.degree. electromagnetic pulse to a nucleus in a magnetic field at a frequency close to the frequency of the angular precession induces a rotation of the magnetization to a direction transverse to the direction of the static magnetic field (the nucleus absorbs energy). The return of the magnetization to equilibrium generates radiation, which radiation may be sampled or detected. In a bulk sample containing many nuclei the time course for such radiation is a function of the environment of the nuclei, via the longitudinal relaxation time (T.sub.1) and the transverse relaxation time (T.sub.2). In general T.sub.1 is usually much larger than T.sub.2.

In two dimensional nuclear magnetic imaging the nuclear spin information is determined for individual area elements within a two dimensional slice of matter. It is critical in two dimensional imaging that the spin information from one element be distinguishable from the spin information from every other element within the slice. One method, suggested by Likes in U.S. Pat. No. 4,307,343, is to use time varying magnetic field gradients within the sensitized two dimensional slice so that the frequency history of the spins in each element in the slice is distinguishable from that of every other element. Variations on this general approach are more fully described below.

According to the methods of the present invention, a two dimensional slice of an object in the presence of a static magnetic field of an NMR spectrometer is sensitized by standard procedures. In the sensitized slice, the magnetization of the spins is transverse to the static field. A time varying magnetic field gradient is then applied within the slice so the frequency history of the spins in each element of the two dimensional slice is distinguishable from that of every other element in the slice. The nuclei in the slice then radiate electromagnetic energy which is sampled, and the sampled electromagnetic energy is converted by reconstruction means to spin information for each element in the two dimensional slice. Each of these steps, excitation, spatial differentiation and image reconstruction, can be carried out in accordance with the general principles disclosed in the Likes patent, subject to the improvements set forth below. In one method of the present invention, the effect of time varying magnetic fields that do not sample equal areas of phase space in equal time is compensated for during the reconstruction phase. In another method of the present invention, the effect of time varying magnetic fields that do not sample equal areas of phase space in equal time is compensated for during the sampling phase (i.e., sampling is performed at non-uniform time intervals so that such sample represents an equal area of phase space). In other methods of the present invention, a unique class of magnetic field gradients is used, which provides time varying magnetic fields that sample equal areas of phase space during equal time intervals.

According to the first method of the present invention, any gradient magnetic field trajectory which substantially covers two dimensional phase space may be used to map that space. During image reconstruction, an appropriate weighting function is applied to account for the failure of the gradient field to pass through equal area increments in equal time intervals.

According to the second method of the invention, a weighting function is used to determine the time periods at which the radiated electromagnetic energy is to be sampled. Even though the trajectories of the magnetic field gradients do not pass through equal phase space areas during equal time intervals, the sampling periods over which the radiated electromagnetic energy is detected can be altered to compensate for the spatially nonlinear characteristics of the gradients. Examples 2 and 3 below apply this aspect of the invention to a sinusoidal and Archimedian spiral trajectory, respectively. The detection of the radiated electromagnetic energy is performed at the termination of each unequal time interval such that each sample of radiated electromagnetic energy corresponds to an equal area of phase space. As shown below, these time intervals are determined with reference to the trajectory and weighting function. According to a third embodiment of the invention, a specific subclass of Archimedian spiral gradient field trajectories has been determined that map equal area elements in equal time intervals.

One method of the present invention involves the use of an appropriate weighting function to weight the received radiated electromagnetic energy signals to compensate for the non-uniform time varying magnetic field gradients that do not sample equal area increments of phase space with equal time.

In accordacne with this first method, a two dimensional slice of a sample is sensitized by standard procedures, such as those disclosed in the Likes patent beginning at column 3, line 44, bridging to column 4, line 2. Sensitizing the slice of the object induces transverse magnetization in the sensitized region of the sample. A time varying magnetic field having gradients in the x and y directions is then applied to the sensitized region of the sample in order to spatially differentiate the radiated electromagnetic signals as disclosed in the Likes patent at column 4, line 2, to column 5, line 53. The spatially differentiated radiated electromagnetic energy signals can be represented by the two dimensional Fourier transform F(f.sub.1,f.sub.2) as described in Ernst in "NMR Fourier Zeugma--Topography," Vol. 18, pp. 69-83 of the Journal of Magnetic Resonance (1975), as follows:

wherein .rho.(x,y) is the spin information in the (x,y) plane; wherein the phases of the applied gradient field, as shown in FIG. 1, are defined f.sub.1 =c.sub.1 t.sub.1, and f.sub.2 =c.sub.2 t.sub.2 ; and wherein the applied gradient field is defined for the x direction (H.sub.x =c.sub.1 x) and for the y direction (H.sub.y =c.sub.2 y.) The terms c.sub.1 and c.sub.2 are constants which are proportional to the gradient fields in the x and y directions, respectively. These well known principles are disclosed in the Likes Patent at Column 4, lines 26 to 68. As described in the Likes Patent at column 4, lines 26 to 68, the two functions f.sub.1 (t) and f.sub.2 (t) are determined from the magntic gradient fields H.sub.x (t) and H.sub.y (t), respectively, where ##EQU1## where gamma is the gyromagnetic ratio.

The reconstructed form of the local spin density (.rho.) at point (x,y) can be determined from the inverse Fourier transform of the radiated electromagnetic energy signals.

The inverse Fourier transform is defined as: ##EQU2## wherein the summation is performed over time increments (.DELTA.t.sub.1, .DELTA.t.sub.2) in (f.sub.1, f.sub.2) space.

However, f.sub.1 and f.sub.2 may both be defined as a function of a single time parameter (t) wherein the gradient fields are defined H.sub.x =c.sub.1 x and H.sub.y =c.sub.2 y cos t with the corresponding phases f.sub.1 =c.sub.1 t and f.sub.2 =c.sub.2 sin t as shown in FIG. 2. In order to ensure mapping of the entire slice, c.sub.2 is defined to be much larger than c.sub.1. As shown in FIG. 2, this gradient field function does not map through equal areas .DELTA.A of (f.sub.1, f.sub.2) phase space in equal time intervals, that is, sin t is not linear in t. Thus, an appropriate weighting function (W(t)) must be applied to correctly reconstruct spin information. As a result, the corrected inverted Fourier transform is given by:

wherein for the above defined gradient fields W(t)=c.sub.1 c.sub.2 cos t and the summation is performed over equal time intervals .DELTA.t and wherein W(t).DELTA.t.sup.2 =.DELTA.A.

Alternatively after the weighting function has been determined, one may map for equal areas in phase space by sampling the radiated electromagnetic energy at unequal time intervals. In accordance with this second method, the weighting function is used to determine the unequal time intervals at which to sample the radiated electromagnetic energy at intervals which correspond to equal areas. By sampling at unequal time periods which have been correlated by the weighting function to equal areas mapped by the gradients, the detected electromagnetic energy need not be corrected by the weighting function prior to reconstructing the image. Rather, compensation is achieved at the sampling phase instead of the image reconstruction phase. The unequal time interval is obtained by requiring the area element .DELTA.A.sub.i to be constant (.DELTA.A.sub.i =.DELTA.A):

Solving for .DELTA.t.sub.i one then obtains: ##EQU3## It should be understood that nearly any gradient field function that substantially covers two dimensional phase space may be substituted for the above functions and may be appropriately weighted on reconstruction. Alternatively, almost any gradient field function that substantially covers two dimensional phase space may be used to map equal areas in phase by imposing an equal area condition on the sampling time intervals, as shown above for one method.

As shown by the Likes Patent at column 11, line 25, and referring to FIG. 3 of the drawings, the phase space trajectory of the Archimedian spiral is given by the equation

The trajectory of the Archimedian spiral as given by equation 6 can also be defined in terms of f.sub.1 and f.sub.2 phase space wherein:

Referring again to FIG. 3, each area .DELTA.A.sub.i ; over which the trajectory of the Archimedian spiral maps during a given time period can be defined as a trapazoid having the dimensions 1/2(r.sub.i+1 -r.sub.i-1) and r.sub.i d.theta.. Accordingly, the area .DELTA.A.sub.i can be defined:

where r.sub.i =c.theta..sub.i, r.sub.i+1 =c(.theta..sub.i +2.pi.) and r.sub.i-1 =C(.theta..sub.i -2.pi.), substituting these values for r.sub.i, r.sub.i+1 and r.sub.i-1 into equation 9 yields:

For the general case where .theta. is defined to be equal to t, the area .DELTA.A.sub.i is given by:

Referring to FIG. 3, the trajectory of the above-defined Archimedian spiral is dependant on time because it accelerates as it radiates outwardly, and as a result, will not map through equal areas of phase space .DELTA.A.sub.i, .DELTA.A.sub.i-1 over equal time periods. Accordingly, use of the methods disclosed by the Likes Patent without compensating for the spatial nonlinearity of the trajectory of the Archimedian spiral over time will result in inaccuracies.

However, by using the methods of the present invention, the area .DELTA.A can be required to be constant, and by defining .theta. to again equal t, equation 11 can be used to determine unequal time intervals .DELTA.t.sub.i over which equal areas of phase space can be mapped:

Thus, applicant has recognized that the general Archimedian spiral in fact does not map through equal areas of phase space over equal time intervals, and that by mapping at unequal time intervals, accurate data can be collected and improved images reconstructed.

In addition, applicant has discovered that a select subset of the Archimedian spiral can be defined which in fact does map through equal areas in equal time. If .DELTA.A.sub.i and .DELTA.t.sub.i are both required to be constant (thereby assuring that equal areas are mapped during equal time intervals), and referring again to Equation 10 it is determined that

Since .DELTA.A, .DELTA.t and c are all constants, solving for .theta. yields

where A and B are arbitrarily chosen constants. Substituting the value of .theta. from equation 14 into equation 10 verifies that the specifically defined Archimedian spiral will in fact map through equal areas of phase space over equal time periods as shown in equation 15 below.

Referring to FIG. 3, it can be seen that the area element given by equation 10 is not valid for the first cycle of the spiral from 0 to 2.pi.. Accordingly, the area .DELTA.A.sub.o over which the trajectory passes during the first 2.pi. cycle must be separately determined. Using basic geometry, and with respect to the above-identified Archimedian spiral, this area is determined to be:

The foregoing is intended to illustrate the methods of the present invention and are not intended to limit its scope. As one skilled in the art would recognize, many modifications may be made to the present invention which fall within its scope and spirit. Specifically, one would recognize that the two dimensional methods described herein may be readily modified to perform three dimensional NMR imaging.