Ultrasonic imaging device

Abstract

Set thresholds for brightness classification to correct the brightness of ultrasound images. [Solution] The processor calculates a brightness histogram 20 of the B-mode tomography image 10 and calculates a set of multiple (e.g., two) Gaussian functions that approximate the brightness histogram 20 (multi-distribution function analysis 104). Next, the processor 1002 determines a threshold for class division based on a Gaussian function selected from among these multiple Gaussian functions based on predetermined conditions (classification threshold calculation 106). The processor 1002 also divides the B-mode tomography image 10 into multiple segments and classifies each segment into multiple classes based on these thresholds (class classification 110).

Description

[Technical Field]

[0001] This disclosure relates to an ultrasonic image processing apparatus, and more particularly to brightness correction. [Background technology]

[0002] Among the processes performed by ultrasound diagnostic equipment is a process called brightness correction or gain correction (hereinafter collectively referred to as brightness correction), which is used to appropriately correct the brightness and contrast of ultrasound images.

[0003] For example, if the brightness of deep areas is low due to ultrasonic attenuation, making observation difficult, it is desirable to increase the brightness of the deep areas. Conversely, if the signal is weak and noise is noticeable even after increasing the brightness, it is desirable to decrease the brightness. Also, if structures with high signal strength are saturated with high brightness, it is desirable to decrease the brightness. Brightness correction is performed in accordance with these objectives.

[0004] Brightness correction can be performed manually by the user or automatically by the device. Manual brightness correction generally involves either adjusting the brightness uniformly across the entire screen or adjusting it for each depth. The depth-based adjustment is called TGC (Time Gain Control) or STC (Sensitivity Time Control).

[0005] Some automatic brightness correction functions not only reduce the effort required from the user, but also enable two-dimensional corrections that are difficult to set manually.

[0006] Furthermore, the following technologies are known for automatically performing brightness correction on ultrasound images.

[0007] The ultrasound diagnostic apparatus disclosed in Patent Document 1 generates an equalization offset pattern that equalizes the brightness levels of a tomographic image frame. In this generation process, the tomographic image frame is divided into multiple sub-areas, the average brightness of each sub-area is calculated, and the brightness difference between the average brightness of each sub-area and the reference brightness, which is the average brightness of the entire tomographic image frame, is calculated. An adjustment value is determined based on the brightness value histogram of each sub-area and the brightness value histogram of the entire tomographic image frame, and the offset value of each sub-area is determined by multiplying the brightness difference by the adjustment value. Then, an equalization offset pattern including offset values ​​corresponding to all pixels in the tomographic image frame is calculated based on the offset values ​​of each sub-area. The gain for each pixel is determined using the equalization offset pattern thus generated. According to this method, the sub-area can be divided into low-brightness, medium-brightness, and high-brightness regions, and the gain level of only the medium-brightness region can be equalized intensively.

[0008] The ultrasound diagnostic apparatus disclosed in Patent Documents 2 and 3 divides a B-mode tomographic image into multiple regions, determines the peak brightness from the brightness histogram for each region, and calculates the brightness difference between the determined peak brightness and a brightness target value defined for each region. Then, a representative brightness difference value is obtained by weighting and adding the brightness differences of each region using the weights assigned to each region, and the gain is automatically adjusted repeatedly until the representative brightness difference value is sufficiently close to the target value. [Prior art documents] [Patent Documents]

[0009] [Patent Document 1] Japanese Patent Publication No. 2010-068987 [Patent Document 2] Japanese Patent Publication No. 2007-117167 [Patent Document 3] Japanese Patent Publication No. 2007-117168 [Overview of the project] [Problems that the invention aims to solve]

[0010] In a method of performing correction for each brightness range such as low brightness, medium brightness, and high brightness, it is important to appropriately set a threshold value for dividing the brightness range. However, no prior art that proposes a method for setting an appropriate threshold value is known.

[0011] The present disclosure proposes a technique for setting a threshold value for classification in the case of classifying each region of an ultrasonic image into a plurality of classes related to brightness and performing brightness correction for each class.

Means for Solving the Problem

[0012] The ultrasonic image processing apparatus according to the present disclosure includes a processor. The processor approximates a histogram of the brightness of the ultrasonic image to be corrected by a combination of a plurality of distribution functions, determines one or more threshold values for brightness based on at least one of the plurality of distribution functions, divides the ultrasonic image to be corrected into a plurality of regions, classifies the plurality of regions into a plurality of classes based on the one or more threshold values, and for each of the plurality of regions, performs brightness correction of each pixel in the region according to the target brightness corresponding to the class to which the region belongs.

[0013] Here, the processor may select one distribution function from the plurality of distribution functions and determine the one or more threshold values based on the selected distribution function.

[0014] Further, the plurality of distribution functions are two distribution functions, and in the selection process, when the height of the peak of the distribution function on the low brightness side among the two distribution functions is a predetermined number of times or more higher than the height of the peak of the distribution function on the high brightness side among the two distribution functions, and the brightness corresponding to the peak of the distribution function on the low brightness side is a predetermined value or more, the processor may select the distribution function on the low brightness side.

[0015] Also, as the brightness correction, the processor may correct the brightness of each pixel so that the average of the brightnesses between regions belonging to the same class is equal.

[0016] Furthermore, a predetermined distance or greater may be provided between adjacent target brightness levels in each class.

[0017] Furthermore, the processor may perform pre-correction on the ultrasound image before correction and treat the pre-corrected ultrasound image as the ultrasound image to be corrected. In this case, the pre-correction may include lateral pre-correction, which increases the brightness of the ultrasound image before correction as it approaches the edges in the lateral direction.

[0018] Furthermore, the pre-correction includes a depth-direction pre-correction that increases brightness in the depth direction as the depth increases, compared to the ultrasound image before correction, and the upper limit of the brightness increase due to the lateral pre-correction is greater than the upper limit of the brightness increase due to the depth-direction pre-correction. It would be acceptable to do so.

[0019] Furthermore, the processor may determine, based on the recognition result of the cross-section represented by the ultrasound image, at least one of the types of distribution functions to be used as the plurality of distribution functions, and the number of distribution functions that constitute the combination of the plurality of distribution functions.

[0020] Furthermore, the processor may determine the calculation conditions for one or more thresholds based on the recognition result of the cross-section represented by the ultrasonic image, and determine the one or more thresholds according to the calculation conditions. [Effects of the Invention]

[0021] According to the ultrasonic image processing apparatus described herein, the threshold for class separation can be set according to the histogram of brightness of the ultrasonic image. [Brief explanation of the drawing]

[0022] [Figure 1] This figure shows an example of the hardware configuration of a computer that performs the processing of the embodiment. [Figure 2]This diagram illustrates the general flow of the process in the embodiment. [Figure 3] This diagram illustrates the detailed flow of the classification process. [Figure 4] This figure shows an example of approximating the brightness histogram of a B-mode tomography image with a pair of Gaussian functions. [Figure 5] This diagram illustrates the different cases involved in selecting a reference Gaussian function when determining the lower and upper limits of intermediate brightness. [Figure 6] This figure illustrates the data changes from B-mode tomographic images to the acquisition of segment-specific classification results. [Figure 7] This diagram illustrates the detailed flow of the brightness correction process. [Figure 8] This figure shows another example of the detailed flow of the brightness correction process. [Figure 9] This diagram illustrates the general flow of processing for modified examples, including pre-correction processing. [Figure 10] This diagram illustrates the arrangement of segments in the lateral and depth directions. [Figure 11] This diagram illustrates the general flow of processing for modified examples, including pre-correction processing. [Figure 12] This figure illustrates a set of brightness histograms from B-mode tomography images of the liver and heart, and a Gaussian function that approximates them. [Modes for carrying out the invention]

[0023] Embodiments of this disclosure will be described below with reference to the drawings. The ultrasound image processing apparatus in this embodiment generates an ultrasound image to be displayed by an ultrasound diagnostic apparatus. The generated ultrasound image is, for example, a B-mode tomography image.

[0024] The image generation process performed by the ultrasound image processing apparatus of this embodiment is performed, for example, by a computer. In one example, this computer is built into the ultrasound diagnostic apparatus. In this example, the ultrasound image processing apparatus of this embodiment is either the ultrasound diagnostic apparatus itself or an image processing system built into the ultrasound diagnostic apparatus.

[0025] In another example, the image generation process in this embodiment may be performed by a computer external to the ultrasound diagnostic device. In this example, the external computer is connected to the ultrasound diagnostic device via a communication path such as a data communication network. In this example, the external computer may be a single computer, or it may consist of multiple computers that cooperate to perform the processing through communication via a data communication network.

[0026] In another example, the process may be carried out through the cooperation of a computer built into the ultrasound diagnostic device and an external computer connected to the ultrasound diagnostic device.

[0027] Figure 1 shows an example of a computer hardware configuration that is built into or connected to an ultrasound diagnostic device. The illustrated computer has a circuit configuration in which a processor 1002, a memory (main memory) 1004 such as random access memory (RAM), an auxiliary storage device 1006 which is a non-volatile storage device such as flash memory, SSD (solid state drive), or HDD (hard disk drive), a controller that controls interfaces with various input / output devices 1008, and a network interface 1010 that controls connection to a network such as a local area network are connected via a data transmission path such as a bus 1012. For example, a program describing the processing of this embodiment is installed on the computer and stored in the auxiliary storage device 1006. The program stored in the auxiliary storage device 1006 is executed by the processor 1002 using the memory 1004, thereby realizing the ultrasound image processing device of this embodiment.

[0028] Figure 2 shows an example of the processing flow performed by the processor 1002 of the ultrasound image processing apparatus according to this embodiment on an ultrasound image. Below, an example of processing a B-mode tomography image 10 as an example of an ultrasound image is shown.

[0029] In this process, the processor 1002 first performs a class classification process 100 on the B-mode tomography image 10 input from an ultrasound diagnostic device or the like. The class classification process 100 classifies each pixel of the B-mode tomography image 10 into several classes with different brightness levels. For example, it may classify them into three classes: low brightness, medium brightness, and high brightness. The class classification process 100 generates a classification result 12 indicating the class of each pixel. The detailed procedure for the class classification process 100 will be explained in detail later.

[0030] The processor 1002 refers to the classification result 12 and performs a brightness correction process 200 on the B-mode tomography image 10. In the brightness correction process 200, the brightness of each pixel in the B-mode tomography image 10 is corrected according to the class of that pixel. In this brightness correction process 200, the contrast adjustment value 16 for each class is referred to. The contrast adjustment value 16 for each class is a brightness offset determined to provide sufficient contrast (in this case, brightness difference) between adjacent classes. For example, if it is desired that the class "low brightness" should have a brightness of at least 10, and that there should be a brightness difference of about 40 between the class "low brightness" and the "mid-range brightness" class, then the contrast adjustment values ​​for the class "low brightness" and the "mid-range brightness" class should be set to 10 and 50, respectively. The contrast adjustment value 16 for each class is predetermined by the system developer or user, taking into account the desired image quality.

[0031] The brightness correction process 200 generates a brightness-corrected image 14. The brightness-corrected image 14 is displayed, for example, on a screen provided in an ultrasound diagnostic device.

[0032] Referring to Figure 3, an example of the classification process 100 of this embodiment will be described. The classification process 100 consists of several sub-processes. These sub-processes include luminance histogram calculation 102, multi-distribution function analysis 104, classification threshold calculation 106, segment division 108, and classification 110.

[0033] In the classification process 100, the processor 1002 first performs a luminance histogram calculation 102. In the luminance histogram calculation 102, a luminance histogram 20 of the input B-mode tomography image 10 is obtained. The luminance histogram 20 is a graph that shows the number of pixels in the B-mode tomography image 10 that have a given luminance for each luminance.

[0034] The processor 1002 performs a multi-distribution function analysis 104 on the luminance histogram 20. The multi-distribution function analysis 104 is a process that finds a set of n distribution functions (where n is an integer greater than or equal to 2) that approximate the luminance histogram 20. In other words, the multi-distribution function analysis 104 is a process that fits the luminance histogram 20 with a set of n distribution functions.

[0035] Various statistical distribution functions can be used, such as Gaussian functions and Lorentz functions. A set of n distribution functions is typically a set of n identical distribution functions, but it may also be a mixture of different types of distribution functions. For example, the process of fitting a histogram with a set of multiple Gaussian functions is called multi-Gaussian fitting.

[0036] The process of fitting the luminance histogram 20 with multiple distribution functions (for example, two Gaussian functions) can be performed using conventionally known algorithms.

[0037] The approximate Gaussian function 22 is obtained by multi-distribution function analysis 104. The approximate Gaussian function 22 is a set of multiple distribution functions that approximate the luminance histogram 20. Below, we will explain using the case where the luminance histogram 20 is approximated by a set of two Gaussian functions as an example. In this example, the approximate Gaussian function 22 includes information that identifies those two Gaussian functions. The information that identifies the Gaussian functions is a set of three values: the maximum value α, the mean value μ, and the standard deviation σ of the Gaussian function.

[0038] Figure 4 illustrates a B-mode tomographic image 10 and its analysis results 400. The analysis results 400 show a curve 402 representing the luminance histogram 20 calculated from the B-mode tomographic image 102 by the luminance histogram calculation 102. A set of two Gaussian functions 404 and 406 that approximate this luminance histogram 20 is obtained by multi-distribution function analysis 104. μ1 and σ1 are the mean and standard deviation of Gaussian function 404, respectively, and μ2 and σ2 are the mean and standard deviation of Gaussian function 406, respectively. By adding these Gaussian functions 404 and 406 together, a dashed curve 408 that approximates the luminance histogram 20 is obtained.

[0039] Returning to the explanation of Figure 3, the next step is for the processor 1002 to perform the classification threshold calculation 106. The classification threshold calculation 106 is a process that calculates thresholds for dividing the luminance range from the data of the approximate Gaussian function 22. For example, if the luminance range is divided into three ranges: low luminance, intermediate luminance, and high luminance, the classification threshold calculation 106 calculates a first threshold that separates the low luminance and intermediate luminance, and a second threshold that separates the intermediate luminance and high luminance.

[0040] An example of the classification threshold calculation process 106 will be explained with reference to Figure 5. The example in Figure 5 shows a case where the luminance histogram 20 is approximated by a pair of two Gaussian functions and the range of luminance is classified into three classes: low luminance, medium luminance, and high luminance. Here, for the sake of explanation, the maximum value, mean, and standard deviation of the Gaussian function 404, whose peak (i.e., maximum value) is on the low luminance side, are set to α1, μ1, and σ1, respectively, and the maximum value, mean, and standard deviation of the Gaussian function 406, which is on the high luminance side, are set to α2, μ2, and σ2, respectively.

[0041] In this example, the relationship between the two Gaussian functions included in the approximate Gaussian function 22 is divided into two cases, Case 1 and Case 2. Then, depending on which case the relationship between the two Gaussian functions falls into, one of the two Gaussian functions is selected to be used to determine the threshold. In one example, if condition A "α1 / α2>3 and μ1>30" is met, it is determined to be Case 1, and otherwise it is determined to be Case 2. The subcondition α1 / α2>3 in condition A means that the maximum value α1 of the low-luminance Gaussian function is significantly higher than the maximum value α2 of the high-luminance Gaussian function. The number 3 shown as the comparison target in this subcondition is merely an example. Also, the subcondition mean μ1>30 means that the mean value μ1 of the Gaussian function is not too low within the luminance range of 0 to 255, which is represented by 8 bits. The number 30 shown as the comparison target is merely an example.

[0042] For the luminance histogram 20 that is determined to be Case 1 as a result of this case classification, the classification threshold is determined based on the low-luminance Gaussian function 404. For example, two thresholds, namely the lower limit of the intermediate luminance and the upper limit of the intermediate luminance, are determined according to Equation 1 below.

[0043]

number

[0044] As can be seen from Equation 1, in Case 1, the width γ corresponds to the standard deviation σ1 from the mean μ1 of the low-luminance Gaussian function 404. low1 The lower limit of the intermediate brightness is defined as a brightness level σ1 lower than the mean value. Also, a range corresponding to the standard deviation σ1 is defined from the mean value μ1. high1 The upper limit of the intermediate brightness is defined as a brightness level σ1 higher than the lower limit. The "intermediate brightness" class is the range from this lower limit to the upper limit. The range below the lower limit of the intermediate brightness is the "low brightness" class, and the range above the upper limit of the intermediate brightness is the "high brightness" class.

[0045] In Case 1, the peak value (i.e., maximum frequency) of the low-luminance Gaussian function 404 is significantly higher than the peak value of the high-luminance Gaussian function 406. Moreover, the luminance of the peak of the low-luminance Gaussian function 404 is not too low. In such cases, the area near the peak of the high-luminance Gaussian function 406 is thought to represent a group of pixels of tissue that causes strong reflection, such as the diaphragm and blood vessel walls, while the area near the peak of the low-luminance Gaussian function 404 is thought to better represent a group of pixels with intermediate luminance corresponding to the parenchymal tissue, such as the liver and thyroid gland. Thus, when considering the overall frequency distribution of luminance in B-mode tomography images, the low-luminance Gaussian function 404 is thought to better represent a group of pixels with intermediate luminance corresponding to the parenchymal tissue. Therefore, in Case 1, the lower and upper limits of the intermediate luminance are determined based on the low-luminance Gaussian function 404.

[0046] Furthermore, for the luminance histogram 20 that was determined to be Case 2 as a result of the case classification, the classification threshold is determined based on the high-luminance Gaussian function 406. For example, the lower limit and upper limit of the intermediate luminance are determined according to Equation 2 below.

[0047]

number

[0048] As can be seen from Equation 2, in Case 2, the width γ corresponds to the standard deviation σ2 from the mean μ2 of the Gaussian function 406 on the high-luminance side. low2 The lower limit of the intermediate brightness is defined as a brightness level lowered by σ². Additionally, a range corresponding to the standard deviation σ² is set from the mean μ². high2 The upper limit of the intermediate brightness is set to a brightness level that is σ² higher. Similar to Case 1, the lower limit of the intermediate brightness separates the low brightness and intermediate brightness classes, and the upper limit of the intermediate brightness separates the intermediate brightness and high brightness classes.

[0049] The brightness distribution in B-mode tomography images of organs such as the heart and liver often falls under Case 2.

[0050] For example, in a cross-section primarily involving the liver, the majority of the B-mode tomographic image consists of the liver tissue parenchyma with a certain level of brightness, while areas with lower brightness representing blood in large blood vessels within that parenchyma occupy a certain area. Therefore, the brightness distribution of the parenchyma portion is represented by the high-brightness Gaussian function 406 with a high peak, and the brightness distribution of the blood portion is represented by the low-brightness Gaussian function 404 with a low peak. In this case, it is appropriate to define the range of intermediate brightness representing the tissue of interest based on the high-brightness Gaussian function 406.

[0051] In the case of the heart, the low-luminance areas corresponding to the blood in the atria and ventricles occupy a large portion of the B-mode tomography image, while the high-luminance areas corresponding to tissues such as the atria and ventricles are relatively small. In this case, the areas corresponding to blood are represented by the low-luminance Gaussian function 404 with a high peak, and the areas corresponding to tissues such as the atria are represented by the high-luminance Gaussian function 406 with a low peak. In this case, there are a relatively large number of pixels corresponding to tissues such as the atria. Therefore, the peak height (α1) of the low-luminance Gaussian function 404 is less likely to be significantly higher than the peak height (α2) of the high-luminance Gaussian function 406 (for example, satisfying one of the case classification conditions, α1 / α2>3). Also, since the luminance of the blood flow area is low, the luminance of the peak of the low-luminance Gaussian function 404 is likely to be low (for example, not satisfying one of the case classification conditions, μ1>30). Therefore, the luminance histogram of the cardiac B-mode tomography image 10 is likely to be classified as case 2.

[0052] In both the case of the heart and the liver, the parenchymal portion of the body tissue of interest in the B-mode tomographic image 10 is more strongly represented by the high-intensity Gaussian function 406; therefore, it is preferable to define the range of intermediate brightness based on the high-intensity Gaussian function 406.

[0053] Returning to the explanation of Figure 3, the processor 1002 executes segment division 108. Segment division 108 is the process of dividing the B-mode tomography image 10 into multiple segments. A segment is a rectangular area consisting of M × N pixels (M and N are positive integers). Segment division 108 generates a divided image 24. The divided image 24 is data representing the image of each segment. For example, the divided image 24 can be represented by a set of information indicating the position of the vertical and horizontal dividing lines that separate each segment, along with the B-mode tomography image 10. The method of representing the divided image 24 is not limited to this.

[0054] Next, the processor 1002 performs classification 110. In classification 110, each segment within the divided image 24 is classified into a class.

[0055] In one example, in the classification process 100, first, pixel-level classification is performed. That is, the processor 1002 classifies each pixel of the segmented image 24 into three classes: low luminance, medium luminance, and high luminance according to the luminance. In this classification, the intermediate luminance lower limit value and the intermediate luminance upper limit value exemplified above are used as the threshold values for separating the classes. Next, for each segment, the processor 1002 determines the class with the largest number of pixels in that segment and sets that class as the class of the segment. The classification result 12 obtained by this classification 110 is provided to the luminance correction process 200 (see FIG. 2).

[0056] FIG. 6 shows the data transition by the segment division 108 and the classification 110. The B-mode tomographic image 10 becomes the segmented image 24 by the segment division 108. Then, by the classification 110, the classification result 12 indicating the class of each segment of the segmented image 24 is obtained.

[0057] FIG. 7 shows an example of the luminance correction process 200. In this example, the processor 1002 first performs the luminance correction amount calculation 202. In the luminance correction amount calculation 202, the luminance correction amount of each segment is calculated based on the classification result 12. In this calculation, the contrast adjustment value 16 for each class is referred to.

[0058] In the luminance correction amount calculation 202, the processor 1002 first calculates the luminance reference value for each class. This calculation is represented by, for example, the following Equation 3.

[0059]

Equation

[0060] Here, # is the identifier of the class. Also, μ target# is the luminance reference value of class #, which is the value to be obtained by this calculation. The luminance reference value μ of class # target#This is the target value for the average brightness of the segment group corresponding to class #. mean(#) is the average brightness of the segment group corresponding to class #, and is determined based on the classification result 12. That is, since the class of each segment in the B-mode tomography image 10 can be determined from the classification result 12, mean(#) can be obtained by calculating the average brightness of the pixels in the segment group corresponding to class #. C (#) This is a predetermined contrast adjustment value of 16 for class #.

[0061] In Equation 3, the mean(#) of class # is multiplied by the contrast adjustment value C of class #. (#) By adding these together, the luminance reference value μ for each class is obtained. target# Determine the luminance reference value μ for each class. target# By defining it in this way, it is possible to create a gap between the average brightness values ​​of adjacent classes that is roughly equal to the difference in contrast adjustment values ​​between those classes.

[0062] Next, the processor 1002 uses the brightness reference value μ for each class that was determined earlier. target# The brightness correction amount 30 for each segment is calculated using the following: Brightness reference value μ target# This corresponds to the target brightness for the class #.

[0063]

number

[0064] Here, i is the segment identification number. Class(i) is a function that, when the identification number i is input, refers to the classification result 12 and returns the identifier # of the class corresponding to i. Δ(i) is the brightness correction amount for segment i, and is the value we want to obtain using this formula. mean(i) is the average of the brightness values ​​of the pixels within segment i.

[0065] As shown in Equation 4, the luminance correction amount Δ(i) of segment i is equal to the luminance reference value μ of the class # of segment i. target#This is obtained by subtracting the average brightness mean(i) of the segment i from the given value. In this way, the brightness correction amount 30 (i.e., Δi) for each segment of the B-mode tomography image 10 can be determined.

[0066] Next, the processor 1002 performs the application of the brightness correction amount 204. That is, the processor 1002 adds the brightness correction amount Δi of the segment i to which the pixel belongs to the brightness value of each pixel in the divided image 24. This generates the brightness-corrected image 14.

[0067] Figure 8 shows a modified example of the brightness correction process 200. In Figure 8, the same reference numerals are used for processes or data similar to those shown in Figure 7.

[0068] In this modified version, the processor 1002 performs pixel-level smoothing 206 on the luminance correction amount 30 for each segment. This smoothing is performed, for example, by averaging the luminance correction amounts of the pixel and its surrounding pixels to determine the luminance correction amount for that pixel. Smoothing 206 gives the luminance correction amount 32 for each pixel. After smoothing 206, the processor 1002 adds the luminance correction amount for that pixel from the luminance correction amount 32 for each pixel of the B-mode tomography image 10 to the luminance value of that pixel. This generates the luminance-corrected image 14.

[0069] Next, with reference to Figure 9, a modified version of the process in Figure 2 will be described. The modified version in Figure 9 is obtained by adding a pre-correction process 300 to the process in Figure 2.

[0070] As you approach the ends of the lateral direction of the imaging range, that is, the direction in which the ultrasonic beam is scanned, the intensity of the ultrasound decreases due to poor probe contact, etc., and the brightness decreases. The pre-correction process 300 includes lateral pre-correction to compensate for this brightness reduction at both ends of the lateral direction.

[0071] In this modified example, a pre-correction process 300 is applied to the B-mode tomography image 10 before correction to generate a pre-corrected image 40. Subsequent processes after the pre-correction process 300, such as the class classification process 100 and the brightness correction process 200, are performed on the pre-corrected image 40 instead of the B-mode tomography image 10. The processing content of the class classification process 100 and the brightness correction process 200 may be the same as that of the embodiment described above.

[0072] Lateral pre-correction may aim to maintain the average brightness at the same depth. Lateral pre-correction may be performed, for example, on a segment-by-segment basis as described above. In this case, segment division for the B-mode tomographic image 10 is performed in the pre-correction process 300. The class classification process 100 then utilizes the results of the segment division in the pre-correction process 300.

[0073] An example of lateral pre-correction at the segment level will be explained with reference to Figure 10. Figure 10 shows an example where a B-mode tomographic image 10 is divided into m rows and n columns of segments. j indicates the segment number along the depth direction, and k indicates the segment number along the lateral direction. Therefore, segment jk is the segment in the jth row and kth column.

[0074] In this example, the processor 1002 calculates the luminance difference g of segments jk according to, for example, the following equation 5. 0_jk Calculate.

[0075]

number

[0076] Next, the processor 1002 calculates the luminance correction amount g of segment jk for lateral direction correction according to, for example, the following equation 6. 1_jk Calculate.

[0077]

number

[0078] In lateral pre-correction, the processor 1002 calculates the brightness correction amount g according to Equation 6. 1_jk This value is subtracted from the brightness of each pixel within segment jk. This correction calculation allows us to increase the brightness at both ends in the lateral direction without changing the average brightness at the same depth j before and after the correction.

[0079] The above explains an example of lateral pre-correction.

[0080] In addition to the brightness reduction at both ends of the lateral direction addressed by lateral pre-correction, ultrasound images have the characteristic that brightness decreases as the depth increases. This brightness reduction in the depth direction may be corrected using conventionally known TGC. For example, the pre-correction process 300 may perform TGC as a depth-direction pre-correction on the input B-mode tomographic image 10 to which TGC has not been applied. In this case, the order in which depth-direction pre-correction and lateral pre-correction are performed does not matter. As another example, the pre-correction process 300 may receive a B-mode tomographic image 10 to which TGC has already been applied as input and perform lateral pre-correction on it.

[0081] Additionally, upper limits may be set for the magnitude of the luminance correction amount for depth-direction pre-correction and lateral pre-correction, ensuring that the calculated luminance correction amount does not exceed these limits. This reduces the possibility of excessive correction occurring.

[0082] In this case, the upper limit of the luminance correction amount for lateral pre-correction may be set to a value greater than the upper limit of the luminance correction amount for depth direction pre-correction. Since the drop in luminance at the lateral edges is greater than the drop in luminance due to attenuation in the depth direction, setting a larger upper limit for lateral pre-correction can compensate for the large drop in luminance at the lateral edges.

[0083] Next, further modifications will be described with reference to Figures 11 and 12. Figure 11 shows the processing performed by processor 1002 in this modification. In Figure 11, the same reference numerals are used for the same processing and data as in Figure 2.

[0084] As shown in Figure 11, in this modified example, the processor 1002 performs cross-sectional recognition processing 600 on the B-mode tomographic image 10, and uses the cross-sectional recognition results obtained from this processing in the class classification processing 100.

[0085] The cross-sectional recognition process 600 recognizes which cross-section of the subject the ultrasound image (B-mode tomographic image 10 in this example) represents. The cross-section of the subject is identified, for example, by a combination of organs within the subject and the positional relationship of the cross-section to those organs. Furthermore, ultrasound diagnostic guidelines for each body part, such as the heart, abdomen, and breast, specify the cross-sections that should be imaged for the diagnosis of that particular body part.

[0086] The cross-sectional recognition process 600 is implemented, for example, using a machine learning model. The machine learning model is constructed using methods such as neural networks. For this construction, a large number of pairs of ultrasound images and information identifying the cross-section shown in the ultrasound image (e.g., the identification name of the cross-section) are prepared as training data. The machine learning model is then trained by providing the ultrasound images as input data and the information identifying the cross-section as training data. The machine learning model, which has been sufficiently trained in this way, is used in the cross-sectional recognition process 600. When an ultrasound image is input to the trained machine learning model, it determines the cross-section shown in the image and outputs information identifying the determined cross-section as the cross-sectional recognition result. Based on this cross-sectional recognition result, the classification process 100 is controlled. Since each organ has a specific trend in the frequency distribution of brightness in B-mode tomography images, the classification process 100 performs processing according to that trend, depending on the type of organ shown in the cross-sectional recognition result.

[0087] For example, the type of distribution function used in the multi-distribution function analysis 104 may be determined according to the cross-sectional recognition results. Alternatively, the number of distribution functions used in the multi-distribution function analysis 104 may be determined according to the cross-sectional recognition results.

[0088] As another example, the calculation conditions for the upper and lower limits of the intermediate brightness may be determined according to the cross-sectional recognition results. For example, the coefficient γ in equations 1 and 2 described above may be used as these calculation conditions. low1 gamma high1 and γ low2 gamma high2 There is.

[0089] For example, Figure 12 shows examples of a B-mode tomography image 10A of the liver and its brightness distribution analysis results 400A, and a B-mode tomography image 10B of the heart and its brightness distribution analysis results 400B.

[0090] In the liver analysis result 400A, the brightness histogram 402A is approximated by a pair of Gaussian functions 404A and 406A. In the case of the liver, the peak of the high-brightness Gaussian function 406A is higher, and this Gaussian function 406A is thought to correspond to the parenchymal portion of the liver tissue.

[0091] On the other hand, in the cardiac analysis result 400B, the peak of the low-luminance Gaussian function 404B is higher than that of the two Gaussian functions 404B and 406A that approximate the luminance histogram 402B, but the high-luminance Gaussian function 406B corresponds to the part of the cardiac tissue.

[0092] Furthermore, compared to the liver, the B-mode tomographic image 10B of the heart has a smaller area of ​​parenchyma, resulting in a smaller ratio of intermediate brightness. Therefore, the coefficient γ in equations 1 and 2 low and γ hign The value for the heart is set to a smaller value than that for the liver.

[0093] The above describes a modified method that utilizes the cross-sectional recognition results. It is also possible to apply the pre-correction process 300, as explained with reference to Figures 9 and 10, to this modified method.

[0094] In the example described above, the luminance range was classified into three classes: low luminance, mid-luminance, and high luminance, but this is merely one example. The method presented in this disclosure is also applicable when the luminance range is classified into two classes, or even into four or more classes.

[0095] In this embodiment, each process is executed on any computer. Furthermore, any computer may execute these processes using a processor as hardware, a program as software, or a combination thereof. In that case, the processor is configured to work in cooperation with the program to execute the various processes in this embodiment, and can function as a unit or means in this embodiment. Also, the execution order of the processes by the processor is not limited to the order described and may be changed as appropriate. Any computer may be a general-purpose computer, a computer designed for a specific purpose, a workstation, or any other system capable of executing each process.

[0096] A processor may consist of one or more hardware components, and the type of hardware is not limited. For example, a processor may consist of a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a programmable logic device such as an FPGA (Field Programmable Gate Array), a dedicated circuit for executing a specific process such as an ASIC (Application Specific Integrated Circuit), a GPU (Graphic Processing Unit), or an NPU (Neural Processing Unit). Furthermore, the type of hardware may be a combination of different types of hardware. When multiple hardware components are configured to execute one or more processes of a given processor, these components may reside in physically separate devices or in the same device. Also, in any embodiment, the order of each process performed by the processor is not limited to the order described above and may be changed as appropriate. Hardware is composed of electrical circuits (circuitry) that combine circuit elements such as semiconductor elements.

[0097] Furthermore, the program may be firmware or software such as microcode. Alternatively, the program may be, for example, a set of program modules, each function of which may be implemented by a processor configured to perform its respective function. The program may be program code or multiple code segments stored on one or more non-temporary computer-readable media (e.g., storage media or other storage). The program may be divided and stored on multiple non-temporary computer-readable media located on physically separate devices. Program code or code segments may represent any combination of procedures, functions, subprograms, routines, subroutines, modules, software packages, classes, or instructions, data structures, or program statements. Program code or code segments may be connected to other code segments or hardware circuits by sending and receiving information, data, arguments, parameters, or memory contents. [Explanation of Symbols]

[0098] 10 B-mode tomography image, 12 Classification results, 14 Brightness-corrected image, 16 Contrast adjustment value, 20 Brightness histogram, 22 Approximate Gaussian function, 24 Segmented image, 100 Class classification processing, 102 Brightness histogram calculation, 104 Multi-distribution function analysis, 106 Classification threshold calculation, 108 Segment division, 110 Class classification, 200 Brightness correction processing.

Claims

1. The processor comprises, The histogram of the brightness of the ultrasound image to be corrected is approximated by a combination of multiple distribution functions. Based on at least one of the aforementioned multiple distribution functions, one or more threshold values ​​for luminance are determined. The ultrasound image to be corrected is divided into multiple regions, Based on the aforementioned threshold of one or more, the multiple regions are classified into multiple classes. For each of the aforementioned multiple regions, brightness correction is performed on each pixel within that region according to the target brightness corresponding to the class to which the region belongs. Ultrasonic imaging device.

2. The aforementioned processor, Select one distribution function from the aforementioned multiple distribution functions, Based on the selected distribution function, one or more thresholds are determined. The ultrasonic image processing apparatus according to feature 1.

3. The aforementioned multiple distribution functions are two distribution functions, The processor, in the selection process, selects the low-luminance distribution function if the peak height of the low-luminance distribution function among the two distribution functions is several predetermined times higher than the peak height of the high-luminance distribution function among the two distribution functions, and the luminance corresponding to the peak of the low-luminance distribution function is greater than or equal to a predetermined value. The ultrasonic image processing apparatus according to claim 2.

4. The ultrasonic image processing apparatus according to claim 1, characterized in that the processor corrects the brightness of each pixel so that the average brightness of regions belonging to the same class becomes equal as the brightness correction.

5. A predetermined distance or greater is provided between adjacent target brightness levels of the same class. The ultrasonic image processing apparatus according to feature 1.

6. The aforementioned processor, Perform pre-correction on the uncorrected ultrasound image. The ultrasound image that has undergone the aforementioned pre-correction is treated as the ultrasound image to be corrected. The aforementioned pre-correction includes lateral pre-correction, which increases the brightness of the ultrasound image before correction, with respect to the lateral direction, closer to the edges. The ultrasonic image processing apparatus according to feature 1.

7. The aforementioned pre-correction includes, for the ultrasound image before correction, a depth-direction pre-correction that further increases the brightness in the depth direction as the depth increases. The upper limit of the brightness increase due to the lateral pre-correction is greater than the upper limit of the brightness increase due to the depth direction pre-correction. The ultrasonic imaging apparatus according to feature 6.

8. The aforementioned processor, Based on the recognition result of the cross-section represented by the ultrasound image, at least one of the following is determined: the type of distribution function to be used as the plurality of distribution functions, and the number of distribution functions that constitute the combination of the plurality of distribution functions. The ultrasonic image processing apparatus according to feature 1.

9. The aforementioned processor, Based on the recognition result of the cross-section represented by the ultrasound image, the calculation conditions for the one or more thresholds are determined. Determine the one or more thresholds according to the calculation conditions above. The ultrasonic image processing apparatus according to feature 1.

10. The histogram of the brightness of the ultrasound image to be corrected is approximated by a combination of multiple distribution functions. Based on at least one of the aforementioned multiple distribution functions, one or more threshold values ​​for luminance are determined. The ultrasound image to be corrected is divided into multiple regions, Based on the aforementioned threshold of one or more, the multiple regions are classified into multiple classes. For each of the aforementioned multiple regions, brightness correction is performed on each pixel within that region according to the target brightness corresponding to the class to which the region belongs. A program that causes a computer to perform a process.

Images