JP3689509B2 - Image correction processing method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method for correcting shading of a grayscale image, and in particular, a nuclear magnetic resonance (hereinafter referred to as “NMR”) signal from hydrogen, phosphorus, etc. in a subject is measured, and a nuclear density distribution, relaxation time distribution, etc. The present invention relates to an image correction processing method applied to a cross-sectional image of a human body obtained by a nuclear magnetic resonance imaging (MRI) method or the like for imaging the image and an MRI apparatus provided with such image correction processing means.
[0002]
[Prior art]
In MRI, protons, which are the main constituents of the subject, are imaged, and the spatial distribution of proton density and the relaxation phenomenon of excited states are imaged, so that the human head, abdomen, limbs, etc. The form or function is photographed two-dimensionally or three-dimensionally.
[0003]
Although this MRI image is widely used in clinical practice, there is a case where shading due to the apparatus appears strongly. In particular, when a local RF receiving coil is used, there are cases where shading caused by the sensitivity distribution of the receiving coil is remarkable and diagnosis is difficult. For example, the profile of the original image of the phantom shown in FIG. 9 is concave where it should be flat, and reflects the sensitivity distribution of the receiving coil. As means for avoiding this, shading correction has been proposed. As an example, a method has been proposed in which a low-frequency component of an image is extracted, and this is regarded as device-derived shading, and the original image is corrected (Axel et al., “Intensity correction in surface coil MR imaging” American Journal of・ Roentogenology, 148, 418-420, 1987).
[0004]
Although the above method is simple, the low-frequency component extracted from the original image has a deviation from the actual shading in the vicinity of the boundary between the subject and the background. As shown in FIG. 9, there is a defect that the peripheral portion of the subject becomes high brightness in the corrected image as shown in FIG.
[0005]
As a method for solving this problem, a method has been proposed in which a secondary image obtained by replacing the region outside the boundary between the subject and the background with the highest pixel value is extracted, and the low-frequency component of this image is extracted, thereby shading correction is performed. (Wold et al., “Phased Array Detector and Automatic Intensity Correction Algorithm for High-Resolution MR Imaging of the Human Brain,” Magnetic Resonance in Medicine, 34, 433-439, 1995).
[0006]
[Problems to be solved by the invention]
Although this method shows an effect with a cross-sectional image of the head, it cannot be applied to a subject with a complicated shape where the boundary between the subject and the background is complicated, and its application is a simple shape like the head Application to the subject (circular) has stopped. On the other hand, in MRI, since every part is imaged from every angle, the correction algorithm is required to be applied to an arbitrary image, but there is no algorithm that enables this.
[0007]
Accordingly, an object of the present invention is to provide a shading correction algorithm that can be applied to an arbitrary image.
[0008]
[Means for Solving the Problems]
The above problem is an image correction processing method for correcting shading of an original image that is a two-dimensional or three-dimensional gray image having a subject and a background.
And means (101) for segmenting the original image into a subject region and a background region,
And means (102) for extracting boundaries of the object and the background,
And means (103) for extracting picture elements in the vicinity of the boundary isotropically from the original image,
Of the picture element which is the extraction, and means for forming a picture element value of the picture element of the background area, the second image is replaced with the value calculated from the pixel value of the pixel of the subject region (104),
The second image, and means for applying a low pass filter (105),
Wherein it is solved by the image correction processing method and means for correcting the luminance of the original image with the third image after low pass filtering effect (106).
[0009]
By extracting the background near the boundary isotropically, it can be applied to images of all angles. Here, in the case of a two-dimensional image, isotropic extraction means that a pixel area within a square area or a circular area having a specific size with a boundary point as the center is extracted. In the case of a three-dimensional image, a picture element within a cubic region or a spherical region having a specific size centering on the boundary point is extracted.
[0010]
The means for replacing the pixel value of the background picture element with the signal value of the subject calculates the signal value of the subject based on the pixel value of the picture element that is the subject, and obtains this signal value as the background picture value. It is assumed to be an elementary picture element value. In this case, the signal value of the subject may be obtained based on the pixel value of the entire subject, but the average is based on the pixel value of the pixel that is the subject among the isotropically extracted picture elements. It can be obtained by calculating a value or a median value.
[0011]
The image obtained by applying the low-frequency pass filter to the second image obtained by replacing the background picture element in the vicinity of the boundary with the signal value of the subject as described above is substantially the same as the actual shading in the vicinity of the boundary. Since the original processing is isotropic, no matter which direction the original image has shading, the original image is corrected using this image, and ideal correction with overcorrection prevented is performed. be able to.
[0012]
In addition, the MRI apparatus of the present invention irradiates a magnetic field generating means for generating each of a static magnetic field, a gradient magnetic field, and a high-frequency magnetic field, and a high-frequency magnetic field that generates nuclear magnetic resonance in a nuclear spin of a tissue constituting the subject. A sequencer for controlling the magnetism generating means; means for detecting a nuclear magnetic resonance signal generated from the subject; image processing means for signal processing of the nuclear magnetic resonance signal to reconstruct an image; and means for displaying an image. The image processing means includes a shading correction means for executing the image correction processing method.
[0013]
Since the MRI apparatus of the present invention is equipped with such a shading correction means, it can correct the shading of the MR image caused by the sensitivity distribution of the detecting means without overcorrection, and provides an image effective for diagnosis. it can.
[0014]
DETAILED DESCRIPTION OF THE INVENTION
An embodiment of the image correction processing method of the present invention will be described below with reference to the flowchart shown in FIG. FIG. 2A schematically shows an MRI image (original image) 200, where the subject 201 portion is a high-brightness gray image and the background portion 202 is a low-brightness gray image.
[0015]
In the image correction processing method of the present invention, first, a high luminance area of such a gray image 200 is extracted as a subject, and a low luminance area is extracted as a background to divide the area (means 101). In this area division, for example, the image is binarized with 50% of the maximum value of the pixel value as a threshold, with zero as the background and 1 as the subject. The binarized image thus obtained is shown in FIG. Note that the threshold for dividing the region is not limited to 50% of the maximum value, and a mode method, a percentage concentration separation method, and a known method can be adopted as a method for determining the threshold.
[0016]
Next, the boundary between the subject 201 and the background 202 is extracted based on the binarized image 300 (means 102). As a means 102 for extracting the boundary 401, a differential filter can be applied to the binarized image 300, and a Laplace filter is preferably used. As a result, an image (boundary image) 400 is obtained in which the pixel value of the boundary 401 is 1 and the other (background 402) has zero as shown in FIG. In this case, a known thinning process may be performed as necessary.
[0017]
Next, the picture elements near the boundary of the original image 200 are isotropically extracted (means 103). For this purpose, first, the boundary image 400 obtained by the means 102 is sequentially scanned as shown in FIG. 4A, and an address having a picture element value of 1 (that is, an address at the boundary) Extract the address of the picture element in the range. The specific range extracted here can be, for example, a square area of 5 × 5 or 7 × 7. FIG. 5A shows a 5 × 5 extracted region 400a. Here, 24 addresses are extracted around the address [3.3] which is the boundary 301 (hatched line).
[0018]
FIG. 5B shows the original image 200a of the extracted area (address). In the next means (104), among the picture elements of the original image in this area, the background picture element value is Replace with the signal value of the subject in the original image near the boundary. This process is hereinafter referred to as edge fill. For this edge fill processing, first, the picture elements in the extracted area are divided into a subject and a background. This is the subject if the pixel value of the binarized image 300 corresponding to each address is 1, and the background if it is 0. In FIG. 5B, the background picture elements are indicated by diagonal lines. Next, a signal value is calculated from the pixel value of the subject of the original image in the region. The signal value of the subject is obtained, for example, by calculating a simple average value or median value of the pixel values, and this signal value is used as the value of the background pixel value of the original image. In the embodiment shown in FIG. 5B, the signal value of the subject is calculated from the pixel values a25, a34, a35, a43, a44, a45, a52, a53, a54, a55 of the ten picture elements that are the subject. Are replaced with the pixel values of the 15 picture elements which are the background.
[0019]
The edge fill processing described above is performed over the entire boundary as shown in FIGS. 4B and 4C, and finally the background pixel values are replaced over the entire boundary as shown in FIG. 4C. An image (second image) 500 is obtained. Note that the edge fill processing may be performed for all the picture elements serving as the boundaries, but may be performed discretely.
[0020]
The second image 500 obtained in this way is applied with a low-frequency pass filter (hereinafter abbreviated as LPF) (means 105). Therefore, the second image is two-dimensionally Fourier transformed into a frequency space image, and an LPF such as a two-dimensional Butterworth filter, a two-dimensional Gaussian filter, or a two-dimensional Hanning filter is applied to the second image, and the two-dimensional inverse Fourier transform is performed again. In this way, the third image obtained by applying the LPF to the second image 500 performs edge fill on the outside of the original image, and therefore the shading is close to the original shading around the subject.
[0021]
Finally, the brightness of the original image 200 is corrected using the third image (shaded image) obtained by the means 105 (means 106). This correction is performed, for example, by dividing the original image 200 by the shading image. Thereby, a corrected image in which shading is corrected can be obtained.
[0022]
In the above description, an example in which the correction processing method of the present invention is applied to a two-dimensional image has been described. However, the present invention can be similarly applied to a three-dimensional image. In this case, after extracting the contour (boundary), a three-dimensional region is isotropically extracted around the boundary picture element, and edge fill processing is performed on this region. Also in this case, edge fill may be performed discretely instead of all the picture elements that are the boundaries.
[0023]
In addition, the isotropically extracted region may be a circular region instead of a square region in the case of two dimensions. In the three-dimensional case, a cubic region or a spherical region can be used.
[0024]
Furthermore, the correction processing method of the present invention can be applied to correction of an image having shading due to the apparatus, but the case where it is applied to image processing in an MRI apparatus will be described below.
[0025]
FIG. 6 is a diagram showing an overall outline of an MRI apparatus to which the present invention is applied. This MRI apparatus is roughly classified into a central processing unit (CPU) 1, a sequencer 2, a transmission system 3, and generation of a static magnetic field. A magnet 4, a gradient magnetic field generation system 21, a reception system 5, and a signal processing system 6 are provided.
[0026]
The CPU 1 inputs photographing parameters from an input device such as a keyboard 22 in advance, and controls each of the sequencer 2, the transmission system 3, the reception system 5, and the signal processing system 6 according to a program. The sequencer 2 operates based on a control command from the CPU 1, and sends various commands necessary for collecting tomographic image data of the subject 7 to the transmission system 3, the gradient magnetic field generation system 21, and the reception system 5. .
[0027]
The transmission system 3 includes a high-frequency transmitter 8, a modulator 9, and an irradiation coil 11 as a high-frequency coil. The modulator 9 amplitude-modulates a high-frequency pulse from the high-frequency transmitter 8 according to a command from the sequencer 2. The high frequency pulse thus amplified is amplified through the high frequency amplifier 10 and supplied to the irradiation coil 11 so that the subject 7 is irradiated with a predetermined pulsed electromagnetic wave.
[0028]
The static magnetic field generating magnet 4 is for generating a uniform static magnetic field in an arbitrary direction around the subject 7. In addition to the irradiation coil 11, a gradient magnetic field coil 13 that generates a gradient magnetic field and a reception coil 14 of the reception system 5 are installed inside the static magnetic field generating magnet. The gradient magnetic field generating system 21 has a configuration in which a gradient magnetic field can be applied independently in Cartesian coordinate axis directions orthogonal to each other, a gradient magnetic field power supply 12 for supplying a current to the gradient magnetic field coil, and a gradient magnetic field power supply 12. It is comprised by the sequencer 2 which controls.
[0029]
The receiving system 5 includes a receiving coil 14 as a high frequency coil, an amplifier 15 connected to the receiving coil 14, a quadrature phase detector 16, and an A / D converter 17, and receives an NMR signal from the subject 7. When the coil 14 detects, the signal is converted into a digital quantity via the amplifier 15, the quadrature phase detector 16, and the A / D converter 17, and is sampled by the quadrature phase detector 16 at the timing according to the command from the sequencer 2. The data is converted into two series of collected data and sent to the CPU 1.
[0030]
The signal processing system 6 includes an external storage device such as an optical disk 19 and a magnetic disk 20 and a display 18 such as a CRT. When data from the receiving system 5 is input to the CPU 1, the CPU 1 performs signal processing and image processing. Processing such as reconstruction is executed, and a desired cross-sectional image of the subject 7 is displayed on the display 18 and recorded on the magnetic disk 20 of the external storage device.
[0031]
The CPU 1 also has an image correction processing algorithm as shown in FIG. 1 and corrects image shading caused by the sensitivity distribution of the receiving coil 14.
[0032]
Comparison of the correction processing method according to the present invention and the conventional correction processing method using shading that is obtained by applying an LPF to the original image, using a phantom image taken with a solenoid type receiving coil in such an MRI apparatus as an example. To explain. The graph shown in FIG. 7 is a graph in which the horizontal axis represents the x-direction position of the AA ′ cross section of the phantom 600 and the vertical axis represents the signal value. In FIG. This profile is concave due to the sensitivity distribution (shading) of the receiving coil 700. In the conventional method, the LPF is directly applied to the original image to estimate the sensitivity distribution of the coil. In this case, the sensitivity distribution is present at the edge portion of the subject (the portion indicated by the oblique lines) as indicated by the dotted line in the figure. It becomes lower and a difference from the original sensitivity distribution occurs. On the other hand, in the correction processing method of the present invention, edge fill processing is performed on the original image, and the signal value of the subject isotropically expanded outside the subject. Then, the estimated sensitivity distribution does not decrease and almost matches the actual sensitivity distribution.
[0033]
FIG. 8 shows the result of correcting the shading of the phantom MR image using the sensitivity distribution estimated by the correction processing method of the present invention. The profile of the original image had non-uniformity {(maximum signal value−minimum signal value) / minimum signal value = b / a} of 0.33, but it was 0.11 (b ′ / a ′) by shading correction. The edge enhancement in the conventional corrected image as shown in FIG. 9 is not observed.
[0034]
The correction processing method of the present invention is effective for MR images of various parts such as a cervical image, a shoulder image, an indirect jaw image, a head image, and a spine image, and is effective for shading correction of various local coils and multiple coils. is there.
[0035]
【The invention's effect】
As can be seen from the above description, according to the correction processing method of the present invention, the boundary of the original image is extracted, the background picture element near the boundary is extracted isotropically, and the signal value is replaced. By performing (edge fill processing), overcorrection of the edge portion is eliminated, and ideal shading correction can be performed. In addition, since the subject area is expanded isometrically, an ideal correction can be made regardless of the orientation of the subject in any direction, and it can be applied to subjects of any part and any angle.
[Brief description of the drawings]
FIG. 1 is a flowchart showing an embodiment of the present invention.
2A and 2B are diagrams for explaining processing of the unit 101 in the flowchart of FIG. 1, in which FIG. 2A is a diagram showing an original image, and FIG. 2B is a diagram showing a binarized image;
FIG. 3 is a view for explaining processing of means 102 in the flowchart of FIG. 1;
FIGS. 4A and 4B are diagrams for explaining the processing of the means 103 and 104 in the flowchart of FIG. 1, in which FIG. 4A is a diagram showing region extraction from a boundary image, and FIGS. The figure which shows substitution of a pixel value.
FIG. 5 is a diagram illustrating an embodiment of edge fill according to the present invention.
FIG. 6 is a block diagram showing an overall outline of an MRI apparatus to which the present invention is applied.
FIG. 7 is a diagram illustrating the principle of the present invention.
FIG. 8 is a graph showing an example of processing by the correction processing method of the present invention.
FIG. 9 is a graph showing processing by a conventional correction processing method.
[Explanation of symbols]
1 .... CPU (image processing means)
2 ... Sequencer 21 ... Gradient magnetic field generation system (magnetic field generation means)
3 .... Transmission system 4 .... Static magnetic field magnet (magnetic field generating means)
5 .... Receiving system 6 .... Signal processing system (image processing means)
200 …… Original image
201 ・ ・ ・ ・ ・ Subject
202 ・ ・ ・ ・ ・ Background
300... Binarized image
400 ... Border image
401: Boundary
500 ... second image

Claims (2)

In an image correction processing method for correcting shading of an original image which is a two-dimensional or three-dimensional gray image having a subject and a background
Means (101) for dividing the original image into a subject area and a background area;
Means (102) for extracting a boundary between the subject and the background;
Means (103) for isotropically extracting picture elements in the vicinity of the boundary from the original image;
Means (104) for forming a second image by replacing a pixel value of a picture element in a background area with a value calculated from a picture element value of a picture element in a subject area among the extracted picture elements;
Means (105) for applying a low-frequency pass filter to the second image;
Means (106) for correcting the luminance of the original image using the third image after the low frequency pass filter action.   Magnetic field generating means for generating each of a static magnetic field, a gradient magnetic field, and a high-frequency magnetic field, and a sequencer for controlling the magnetic field generating means for irradiating a high-frequency magnetic field that causes nuclear magnetic resonance to occur in a nuclear spin of a tissue constituting the subject Magnetic resonance imaging comprising: means for detecting a nuclear magnetic resonance signal generated from the subject; image processing means for signal processing the nuclear magnetic resonance signal to reconstruct an image; and means for displaying the image In the apparatus, the image processing means divides the original image into a subject area and a background area ( 101 ) And means for extracting a boundary between the subject and the background ( 102 ) And means for isotropically extracting picture elements in the vicinity of the boundary from the original image ( 103 )When,
Means for forming a second image by replacing a pixel value of a pixel in a background area with a value calculated from a pixel value of a picture element in a subject area among the extracted picture elements ( 104 ) And means for applying a low-frequency pass filter to the second image ( 105 ) And means for correcting the luminance of the original image using the third image after the action of the low-frequency pass filter ( 106 And a shading correction means including a magnetic resonance imaging apparatus.

Images