JP2005270491A - Cardiac function quantitative evaluation method in cine mri - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce an effect of a setting of a threshold (a window level) when displaying an image, provide an automatic processing method of a moving image in a diagnosis of a ventricular diastolic function, and provide a method precisely evaluating the ventricular diastolic function from a ventricular volume curve obtained from Cine MRI in the ventricular function evaluation by Cine MRI. <P>SOLUTION: This method is so constituted that after Cine MRI is processed with a V (Variance dependent)-filter, the image processed with the V-filter is undergone with a differentiation and a thinning processing to form the boundary between the ventricular lumen blood and the endomyocardium and determine the volume of the ventricular lumen. This method is quantitatively evaluates a local myocardial wall movement and the ventricular function using the the ventricular lumen volume curve approximated by Fourier series and its differential curve. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention automatically determines the boundary between intraventricular blood and the myocardial intima from a moving image obtained by electrocardiogram-synchronized cine (Cine) magnetic resonance imaging (cine MRI), and the obtained ventricular volume curve is obtained. The present invention relates to a method for quantitatively evaluating ventricular contraction function and ventricular dilation function by performing Fourier series approximation.

  Diagnosis and treatment of heart disease requires accurate diagnosis of ventricular function and local myocardial wall motion. Echocardiography and ventricular angiography have been used to diagnose ventricular function and myocardial wall motion, but cine MRI is superior in accuracy and reproducibility to echocardiography, ventricular angiography, and cardiac pool scintigraphy. (For example, refer to Patent Document 1). In cardiac function evaluation using cine MRI, high-resolution moving images in multiple sections from the apex to the base of the ventricle are taken, and the ventricular volume and myocardial weight are calculated from the three-dimensional image using the Simpson method (disc summation method). Since measurements are taken, accurate measurements can be obtained even in cases with ventricular deformation or wall motion abnormalities. In echocardiography, the imaging section is limited by lung air and bone, but in MRI, the imaging section can be freely selected. Furthermore, even in the case of obesity or emphysema where the image quality is poor by echocardiography, a clear image can be obtained by using cine MRI.

Japanese translation of PCT publication No. 2002-541963

  Furthermore, from cine MRI, not only ventricular volume and contractile function (ventricular ejection fraction etc.) but also an index related to ventricular diastolic function (diastolic maximum filling speed) can be obtained. The ventricular diastolic function is a more sensitive index than the ventricular contractile function in the determination of the pathophysiology and prognosis of heart failure, and shows an abnormality at an earlier stage than the contractile function in myocardial ischemia.

  By the way, in order to accurately evaluate the diastolic function, it is necessary to obtain a ventricular volume curve with high time resolution. Recently, respiratory arrest cine MRI in which one heartbeat is divided into 40 time phases can be obtained, and ventricular diastolic function evaluation by cine MRI is also possible. However, the quantitative evaluation of the ventricular diastolic function requires accurate tracing of the boundary between the intraventricular cavity and the myocardium in hundreds of cine MRI images, which requires a great deal of labor and time for image analysis. Therefore, development of a method capable of automatically and accurately detecting the boundary between intraventricular blood and myocardial intima in cine MRI and quantitatively analyzing both the contractile function and the dilated function has been strongly desired.

  Until now, in ventricular function evaluation by cine MRI, a method of manually determining and visually tracing the boundary between the heart lumen and the myocardial intima has been used. However, the conventional manual tracing method is strongly influenced by the experience level and skill of the judge, and the result may vary depending on the threshold (window level) setting when displaying an image. Furthermore, in the diagnosis of ventricular diastolic function, it is necessary to trace the endocardial border for hundreds of moving images, and automation of the process is essential for clinical use.

  In addition, the ventricular volume curve obtained from cine MRI includes fine fluctuations due to measurement errors, and if a measured value such as the maximum diastolic filling rate is directly obtained from the differential curve of the ventricular volume curve, it depends on the noise component. There is a problem that accurate results cannot be obtained due to fluctuations. So far, an automatic measurement method of cine MRI has been proposed, but it is difficult to accurately evaluate the extended function with the existing method.

In order to solve this problem, the present inventors performed a variant dependent filter (hereinafter referred to as a V-filter) process on an electrocardiogram-synchronized cine magnetic resonance imaging (cine MRI) image of the heart. Invented a method for determining the volume of the intraventricular lumen by performing differentiation and thinning on the image after V-filter processing to form a boundary between the intraventricular blood and the myocardium.
Further, a time change curve of the intraventricular volume in each cross section of the heart is obtained, a Fourier series approximation is performed on the time change curve of the intraventricular volume, and the intraventricular volume curve and its differential curve approximated by the Fourier series Invented a method for quantitative evaluation of local myocardial wall motion and ventricular function by evaluating ventricular contraction function and ventricular dilatation function.

  The method for quantitatively determining local myocardial wall motion and ventricular function in electrocardiogram-synchronized cine magnetic resonance imaging according to the present invention described above is based on the following. It becomes possible to detect automatically without setting etc. Furthermore, from the obtained ventricular volume curve, an index relating to the ventricular contraction function and ventricular dilation function can be automatically measured while suppressing the influence of statistical fluctuations.

Embodiments of a method for automatically obtaining a boundary between intraventricular blood and the myocardial intima from a moving image of an electrocardiogram-synchronized cine MRI according to the present invention and quantitatively evaluating an index relating to a ventricular contraction function and a ventricular dilation function Will be described with reference to the drawings.
By the way, the quantitative evaluation method according to the present invention uses electrocardiogram synchronization or real-time ultra-high-speed imaging, and collects any of the gradient echo method, the steady state free precession method, the echo planar method, and the spin echo method. It is intended for a cine MRI moving image of the heart, which is imaged using the method.

First, a method for automatically obtaining the boundary between intraventricular blood and the myocardium from the electrocardiogram-synchronized cine MRI moving image will be described with reference to FIG. In order to designate the approximate position of the ventricle in the cine MRI, a circle surrounding the ventricle is manually set for one of the reconstructed ventricular short axis cine MRI.

  Next, the signal intensity in the vicinity of the septum and the side wall of the heart lumen blood in cine MRI is detected, and the signal detection sensitivity non-uniformity correction for the signal non-uniformity of the heart lumen blood due to the coil sensitivity is performed using this value. . Then, in order to automatically detect the boundary between intraventricular blood and the myocardial intima, V-filter processing is performed.

  The V-filter is a filter that performs smoothing while preserving the boundary. As shown in FIG. 3, the neighborhood of the target pixel is divided into four regions, and the average and variance are calculated in each region. An area where the variance is minimized is selected, and the average value is assigned to the target pixel. Since the selection area is clearly switched on both sides of the boundary, a sharp boundary can be obtained. By performing V-filter processing, noise is removed, the boundary is emphasized, and contour extraction becomes easy. Further, the V-filter processing has a feature that image change is reduced even if the threshold setting at the time of differentiation processing described later varies slightly.

  The image subjected to the V-filter processing is subjected to differentiation processing to detect the boundary between intraventricular blood and the myocardial intima. The boundary between intraventricular blood and myocardial intima in cine MRI is a portion where the signal intensity of the image changes, and differential calculation for extracting the change in signal is used for contour extraction. In order to obtain the first derivative for the image, the differential values in the x direction and the y direction are calculated using the Sobel differential operator shown in equations 1 and 2.

ここでｆ_ｘをｘ方向の微分値、ｆ_ｙをｙ方向の微分値とする。両方向の微分値を求めた後、数式３によりｘ及びｙ方向を合成した微分値ｆ_ｘｙを算出する。画像の各画素について算出し、微分画像を作成する。微分オペレータは合計すると０となり、平坦な領域では出力は０となる。したがって信号変化のある領域、つまり輪郭領域が描出される。

Here the f _x differential value in the x direction, the f _y a differential value in the y direction. After obtaining the differential value in both directions, a differential value f _xy obtained by combining the x and y directions is calculated according to Equation 3. Calculation is performed for each pixel of the image to create a differential image. The differential operators total 0 and the output is 0 in a flat region. Therefore, an area with a signal change, that is, an outline area is drawn.

  Next, contour extraction is performed using a binary image by image threshold processing after differentiation. When performing contour extraction, since the obtained contour is not one pixel wide, a one pixel wide contour is created by thinning processing. If it matches any of the eight patterns shown in FIG. 4, the target pixel is deleted. By repeating this process until there are no more pixels that match the pattern in FIG. 3, the region having a plurality of pixel widths is deleted from the outside to create a one-pixel width contour.

  Next, a method of creating a ventricular volume curve from the boundary between intraventricular blood and myocardial intima determined in the cine MRI determined by the above method, and automatically measuring an index relating to ventricular contraction function and ventricular dilation function. explain. The ventricular volume curve is obtained by plotting the time from the ECG-synchronized R wave on the horizontal axis and the ventricular volume on the vertical axis. The maximum value in the end diastole of the ventricular volume curve is the end-diastolic volume (EDV), the minimum value in the systole is the end-systolic volume (ESV), the ventricular diastole volume and the ventricular systole. The difference in volume is called ejection volume (SV), the product of ejection volume and heart rate is called cardiac output (CO), and SV / EDV is called ventricular ejection fraction (EF). It is used for functional evaluation of the heart. Indicators such as ejection volume ejection rate indicate ventricular contraction pump function.

  In the first derivative curve of the ventricular volume curve, the systolic minimum value is called the maximum ejection rate (PER) and the diastolic minimum value is called the maximum filling rate (PFR). Speed reflects ventricular dilation. The ventricular diastolic function is a more sensitive index than the ventricular contraction function in the determination of the pathophysiology and prognosis of heart failure, and the ventricular diastolic function is abnormal at an earlier stage than the contraction function even in myocardial ischemia. The left ventricular maximum filling rate is obtained as the minimum value of the first derivative curve of the ventricular volume curve. Since this differential curve is easily affected by volume measurement errors in each time phase of the cine MRI moving image, an accurate maximum filling speed measurement result cannot be obtained simply by calculating the minimum value of the differential curve.

  This time, the inventor invented a method of creating a ventricular volume curve from cine MRI and automatically quantitatively measuring indexes related to ventricular contraction function and ventricular dilation function (see FIG. 2). After performing a differential process with the V-filter on the cine MRI, the maximum value of the differential value is scanned radially from the center point of the intraventricular lumen blood, so that the boundary between the intraventricular lumen blood and the myocardial intima in the image is obtained. A region of interest was created.

  This process is performed for each slice from the apex to the base, and the number of pixels in the boundary between the intraventricular blood and the myocardium is obtained, and the volume for each pixel is calculated to calculate the ventricular volume at each time phase. did. The obtained left ventricular volume curve was approximated by a Fourier series having a fourth or higher order term.

  Using the left ventricular volume curve approximated by the Fourier series and its differential curve, the ventricular end-diastolic volume (EDV), end-systolic volume (ESV), ejection volume (SV) : stroke volume), cardiac output (CO), ventricular ejection fraction (EF), maximum ejection speed (PER), maximum filling rate (PFR) The time from reaching the index and the electrocardiogram R wave to the end-systolic volume (ESV), the maximum ejection rate (PER), and the maximum filling rate (PFR peak filling rate) was calculated.

  ECG-synchronized cine MRI was performed on eleven cases of heart disease patients (mean age: 67.8 years) consisting of 9 males and 2 females.

<Test method>
The GE 1.5T high-speed MR device for heart (CV / i) and the phased array coil for heart made by GE are used for the MRI apparatus, and one heartbeat (RR wave interval) is divided into 20-32 under respiratory arrest. And imaged. Steady state method (TR = 4.3 ms TE = 1.7 ms, slice thickness: 10 mm, FOV: 31.9 × 31.9 cm) was used for the imaging pulse sequence.

  First, in one of the ventricular short-axis cine MRI, a circle including the ventricular myocardium was set and the position of the ventricle was designated. Next, the coil sensitivity nonuniformity was corrected using blood signal intensities near the septum and the side wall. Next, in order to automatically detect the boundary between intraventricular blood and the myocardial intima, V-filter processing and differentiation processing were performed. Next, the maximum value of the differential value was scanned radially from the center point of the intraventricular lumen blood, and the boundary between the intraventricular lumen blood and the myocardial intima in each image was obtained. This process was performed for each image from the apex to the base, and the number of pixels of intraventricular blood was determined to calculate the ventricular volume. The obtained left ventricular volume curve was approximated by a fourth order Fourier series. By using the left ventricular volume curve approximated by Fourier series and the differential curve of the curve, the index of the left ventricular contraction function and the expansion function was obtained.

<Case 1>
FIG. 5 shows the processing results in the case of lower wall acute myocardial infarction. From the image obtained by superimposing the created ROI on the left ventricular short-axis tomographic image, it was confirmed that the created ROI substantially coincided with the edge of the left ventricular cavity. The calculated left ventricular ejection fraction was 49.0%, which was lower than the normal value of 62 ± 5%, and a slight decrease in cardiac function was observed. In addition, by performing Fourier series approximation of the fourth-order term, a good left ventricular volume curve and its first-order differential curve from which noise components are effectively removed are obtained, and the left ventricular diastolic filling rate is automatically and quantitatively determined. It became possible to ask.

  The influence of the conventional method (threshold method) and the present method (V-filter method) on obtaining the intraventricular volume from cine MRI images will be described. As shown in FIG. 6, in this method, a constant intraventricular cavity volume is always obtained as compared with the conventional method, and visually matches the intraventricular lumen well. In the other 10 cases, the borderline between the left ventricular blood and myocardium automatically created by this method was in good agreement with the left ventricular border, confirming the effectiveness of this method in clinical cases. .

<Summary>
Since cine MRI images are not easily affected by bones and air and have high spatial resolution, indices of cardiac function evaluation calculated from the left ventricular volume curve obtained by this method can be obtained by ECG-synchronized SPECT and echocardiography. It is considered that the left ventricular function can be evaluated with higher accuracy than the parameter value and with higher accuracy.
Furthermore, the region of interest necessary for determining the left ventricular volume is automatically set. Therefore, there are advantages such as reduction of the load on the operator, improvement of reproducibility between observers, and reduction of processing time.

It is a figure which shows the flow of the process for determining automatically the boundary of the ventricle and the myocardial intima in ECG synchronous cine MRI. It is a figure which shows the parameter | index regarding a ventricular function. It is a figure which shows 4 area | regions near the attention pixel in V-filter process. It is a figure which shows 8 patterns which delete a focused pixel in thinning process. It is a figure which shows the left ventricular function automatic measurement process result in a myocardial infarction case. It is a figure which shows the comparison of this method (V-filter method) and the conventional method (threshold method) about the influence of the threshold value which acts on the intraventricular cavity volume.

Claims (2)

  In a method for determining local myocardial wall motion and ventricular function using electrocardiogram-synchronized cine (Cine) magnetic resonance imaging (cine MRI) of the heart, means for performing Variance dependent filter (V-filter) processing on the cine MRI image; And means for performing differential processing and thinning processing on the image after the V-filter processing to form a boundary between intraventricular blood and the myocardial intima, and determining a volume of the intraventricular lumen. Quantitative evaluation method of local myocardial wall motion and ventricular function. A method for obtaining a time-varying curve of the intraventricular lumen volume in each cross section of the heart using the method of claim 1 and performing Fourier series approximation on the time-varying curve of the ventricular lumen volume, and a heart approximated by the Fourier series Quantitative evaluation of local myocardial wall motion and ventricular function, comprising means for evaluating ventricular contraction function and ventricular dilatation function using a chamber volume curve and a differential curve of the chamber volume curve Method.

Images