TWI771761B - Method and device for processing medical image - Google Patents

Abstract

The present invention provides a method and a device for processing medical image. The method comprising steps of, firstly, receiving an inspected picture related to an subject receiving a plurality of recorded pictures related to the subject, comparing the inspected picture with the plurality of recorded pictures for determining at least one recorded pictures that is similar to the inspected picture, adjusting the at least one similar recorded pictures for generating a similar picture, and finally, comparing the inspected picture with the similar picture for determining at least one difference part between the similar picture and the inspected picture.

Description

Medical image processing method and medical image processing device

The present invention relates to an image processing technology, in particular to a medical image processing method and a medical image processing device which can compare and correct similar images similar to target images from a plurality of recorded images.

With the advancement of science and technology, various imaging medical testing instruments, such as ultrasonic scanning, MRI (Magnetic Resonance Imaging, MRI), Positron Emission Tomography (PET), Computed Tomography (Computed Tomography, CT), mammography and X-ray photography and other equipment to generate detection images. Such images can be used to diagnose certain diseases and also provide medical research.

The storage and transmission of inspection images basically follow the International Medical Imaging Standard of Digital Imaging and Communications in Medicine (DICOM), which is used in medical image processing, storage, printing, and transmission. Standard agreement. It contains definitions of file formats and network communication protocols. DICOM is an application protocol based on TCP/IP, and uses TCP/IP to communicate with each system. Between two medical instruments that can accept DICOM format, images and patient data can be received and exchanged through DICOM format files.

Through the DICOM image transmission protocol, users can obtain the historical image data stored in real-time and past examinations at any time. Through appropriate observation and comparison, differences can be found, and then the patient's physiological state can be judged. However, in the conventional technology, after obtaining the relevant detection images of the patient, the general medical personnel usually rely on the experience of the medical personnel to judge the past images and the current images through human vision. This diagnosis method is easy to follow. It depends on the experience of the medical staff, and misjudgment is prone to occur.

On the other hand, for example, the same subject may change in body shape such as fat or thin at different time points, or the medical images generated by different machines may have angular deviations, causing medical personnel to inspect before and after inspections. When imaging, it is not easy to find the images to be compared, and even because of different viewing angles, it is difficult to accurately and instantly compare the differences.

In view of the above, a medical image processing method and a medical image processing device are needed to solve the deficiencies of conventional techniques, improve the efficiency of medical image comparison and reduce misjudgment.

The present invention is a technology that can perform medical image processing. The image input terminal receives the detection image of the detection device, and through the image processing mechanism, the recorded image that is most similar to the detected image is selected from the past recorded images and processed. To generate similar images that can be compared and superimposed with the detected images, and automatically generate comparison information, it solves the problems of inconvenience and easy misjudgment by users in the conventional technology when comparing visually. Among them, when generating similar images, the present invention further provides a method of interpolating a plurality of recorded images to form similar images by interpolation, which avoids the conventional time-consuming imaging procedure of slicing through full 3D modeling. , to achieve the effect of improving the speed of operation.

In one embodiment, the present invention provides a medical image processing method, which includes the following steps. First, a detection image of a subject to be tested is received. Next, a plurality of recorded images of the subject is received. Afterwards, the detected image and the recorded images are compared to select at least one recorded image similar to the detected image from the recorded images. Then at least one recorded image is adjusted to generate a similar image. Finally, the detected image and the similar image are compared to determine that at least one different part is obtained.

In another embodiment, the present invention provides a medical image processing device, which includes an image processing unit and a first image output unit. The image processing unit is electrically connected to a database storing a plurality of recorded images, and is used for receiving a plurality of recorded images and a detection image from the image generating device, wherein the image processing unit detects a plurality of first images on the detection image Distinguishing feature points, and respectively detecting a plurality of second distinguishing feature points on a plurality of recorded images, and performing a comparison between the plurality of first distinguishing feature points in the detected images and the second distinguishing feature points in the plurality of recorded images respectively. The comparison is to determine and select at least one recorded image that is similar to the detected image from the plurality of recorded images, wherein the image processing unit adjusts the at least one recorded image to generate a similar image. The first image output unit is electrically connected to the image processing unit for outputting similar images.

Various illustrative embodiments may be described more fully hereinafter with reference to the accompanying drawings, in which some illustrative embodiments are shown. However, the inventive concepts may be embodied in many different forms and should not be construed as limited to the exemplary embodiments set forth herein. Rather, these illustrative embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the inventive concept to those skilled in the art. Similar numbers always indicate similar elements. The medical image processing method and the medical image processing apparatus thereof will be described below with various embodiments in conjunction with the drawings. However, the following embodiments are not intended to limit the present invention.

First, please refer to FIG. 1 and FIG. 2 , wherein FIG. 1 is a schematic flowchart of an embodiment of a medical image processing method of the present invention; FIG. 2 is a schematic diagram of an embodiment of a medical image processing apparatus implementing image capture means of the present invention. In step 20 of the medical image processing method 2, firstly, the image generating device 4 receives the detection image of the subject. The image generating device 4 for outputting the detection image can be various medical detection equipment, such as: ultrasonic scanning device, magnetic resonance imaging (MRI), positron emission tomography (PET), computed tomography (Computed Tomography, CT), mammography and X-ray photography devices, etc. The detection images may be one or more static images, or may be continuous and dynamic detection images, or one or more static images in the detection images, which are not limited. In one embodiment, after the image generating device 4 generates a detection image about the subject (eg, one of the body parts of the subject), one of the detection images is selected from the detection images. One of the detection images is obtained by the image capturing method for subsequent comparison. In another embodiment, the image capturing means is performed by the medical image processing apparatus 3 shown in FIG. 2 . The medical image processing apparatus 3 includes an image capturing unit 30 , an image processing unit 31 and a first image output unit 32. The image capturing unit 30 is coupled to the image generating device 4 for receiving the detection image output by the image generating device 4 and extracting the detection image from the detection image.

Next, in step 21, a plurality of recorded images about the subject to be tested are obtained from the database, which are the past detected recorded images stored in the database, and are mainly used for comparison with the detected images. In this embodiment, the medical image processing device 3 is electrically connected to the cloud server 5 and the image processing unit 31 in the medical image storage and transmission system (Picture Archiving and Communication System, PACS) through the network connection unit 36 . The cloud server 5 has a database in it, and the image generating device 4 records all the detection images generated during the detection process for each test subject; Medical imaging records of the test. The medical image processing device 3 obtains the plurality of recorded images from the database 50 through a specific communication protocol. In one embodiment, the communication protocol may be Digital Imaging in Medicine (DICOM).

After that, go to step 22 to compare the detected image with the recorded images to select a recorded image similar to the detected image from the recorded images. The comparison method of this step is described as follows. As shown in FIG. 1B , step 220 is first performed to detect a plurality of first distinctive feature points on the detection image 9 . Please also refer to FIG. 3A , which is a schematic diagram of an embodiment of the detection image of the present invention. In the detection image 9 , C0 represents the first distinguishing feature point, which can be detected by the image processing unit 31 executing an algorithm. The automatic definition on image 9 is generated. The aforementioned algorithm, in one embodiment, may be a scale-invariant feature transform (SIFT), but is not limited thereto, for example: Speed-Up Robust Feature (SURF) ) and the like can also be implemented, which are well known to those with ordinary knowledge in the art, and will not be repeated here.

After the detection image 9 detects a plurality of first distinctive feature points C0 , step 221 is performed next to detect a plurality of second distinctive feature points from the plurality of recorded images stored in the database, respectively. In this embodiment, the medical image processing device 3 is electrically connected to the cloud server 5 and the image processing unit 31 in the medical image storage and transmission system (Picture Archiving and Communication System, PACS) through the network connection unit 36 . The cloud server 5 has a database in it, which stores the detection image records generated by the image generating device 4 during the detection process. The medical image processing device 3 obtains the plurality of recorded images from the database 50 through a specific communication protocol. In one embodiment, the communication protocol may be Digital Imaging in Medicine (DICOM). Please refer to FIG. 3B , which is a schematic diagram of a plurality of recorded images of the present invention. In one embodiment, the plurality of recorded images 9a to 9f shown in FIG. 3B represent the image records of the same subject in a certain examination in the past, and each recorded image 9a to 9f represents the images of different scanning layers on a specific organ. Detect images. In step 221, an algorithm is also used, and the present embodiment is a SIFT algorithm, which detects and generates a plurality of second distinctive feature points C1-C6 on each of the recorded images 9a-9f, but is not limited thereto.

Next, step 222 is performed, and the plurality of first distinguishing feature points C0 of the detection image 9 are compared with the second distinguishing feature points C1 to C6 in the plurality of recorded images 9a to 9f, respectively, so as to be recorded by the plurality of records. Among the images 9a to 9f, the most similar images to the detection image 9 are determined. In an embodiment of step 222, through an algorithm, such as a SIFT or SURF algorithm, the present embodiment is a SIFT algorithm, the detection image 9 and each of the recorded images 9a-9f are calculated, that is, each The second distinctive feature points C1 to C6 of each of the recorded images 9a to 9f are matched with the first distinctive feature point C0 of the detection image 9, and each recorded image is obtained by SIFT calculation. 9a to 9f are relative to the similarity density of the detected image, and the higher the similarity density is, the greater the number of matched first and second feature points. For example, in Fig. 3B, the recorded image 9e has the highest similarity density, so after the calculation in step 222, the recorded image 9e is the most similar recorded image.

Returning to FIG. 1A , after the recorded image 9e closest to the detected image is obtained after step 22, step 23 is performed to adjust the outline, orientation or scale of the recorded image to generate a similar image. In an embodiment of this step, the adjustment procedure is mainly to perform an adjustment calculation on the most similar recorded image found by the aforementioned comparison, so as to obtain a similar image with the same viewing angle, orientation, outline or scale as the detected image. In one embodiment, the recorded image obtained through step 22 may have a difference in viewing angle orientation, outline or scale with the detected image, for example, the state shown in FIG. 3C may be present. Therefore, through the adjustment procedure of step 23, the similar images can be corrected through adjustment calculation. In one embodiment, as shown in FIGS. 3C and 3D , after the recorded image in FIG. 3C is processed in step 23 , the image state shown in FIG. 3D is obtained. The calculation method of step 23 mainly generates a transposition matrix by at least four or more similar second feature points CS1-CS4 on the recorded image 9e obtained in step 22, and this transposition matrix can convert the image into a space ( x, y, z), so that the recorded image 9e becomes a similar image 9se with the same viewing angle, orientation, outline or scale as the detected image 9 through the transformation of the transposition matrix. It can be seen from the similar image 9se in FIG. 3D that the detection image 9 shown in FIG. 3A has a similar viewing angle. It should be noted that the adjustment algorithm is a conventional technique and will not be repeated here.

Next, returning to FIG. 1A , after obtaining a similar image most similar to the detected image in step 23 , step 24 is performed to compare the detected image and the similar image to determine that at least one different part is obtained. In an embodiment of step 24, a scale can be superimposed on the similar image and the detection image in the same scale respectively to facilitate the user to identify, as shown in (a) of FIG. 3E As shown in Figure ~(b), the unit length has a certain number of pixels, eg, 100 pixels (pxls) in Figure 3E, but not limited thereto. In another embodiment of step 24, as shown in (c) of FIG. 3E, the similar image 9se and the detection image 9 are calculated, and only the difference map of the difference area 900 between the two images is individually highlighted. Like 9x for user reference. Alternatively, as shown in (d) of FIG. 3E, the difference image 9x and the detection image 9 are superimposed on each other to generate a superimposed image 9', and a similar image is framed on the detection image. The position of the difference area 900 is highlighted to facilitate the user's reference.

It should be noted that, in the process of comparing the detected image with the recorded image stored in the database, there may be two situations. The first is to use the similar density of the entire detected image and the recorded image as mentioned above. Compare to select the single recorded image with the highest density similar to the detection pattern. In the second case, it is possible to compare one or more than two recorded images that are most relevant to the detection image. The following describes the embodiment of the second case.

As shown in FIGS. 4A to 4C , wherein FIG. 4A is a schematic flowchart of another embodiment of the medical image processing method of the present invention, FIG. 4B is a schematic flowchart of an embodiment of calculating and generating a plurality of similar recorded images, and FIG. 4C is A schematic diagram of the process flow of an embodiment of generating similar images. In the medical image processing method 2a shown in FIG. 4A, the manners of steps 20a-21a are the same as the manners of the foregoing steps 20-21, and details are not repeated here. Step 22a in this embodiment, as shown in FIG. 4B, further includes defining a comparison area between the detected image and the recorded image, and comparing the detected image and the recorded images, so as to obtain the data from the recorded images. At least one recorded image similar to the detected image is selected from the images. In one embodiment, step 22a may further include step 220a, where the detection image 9 (as shown in FIG. 5A ) and each of the recorded images 9a to 9f (as shown in FIG. 5B ) are respectively defined as a plurality of regions. , in this embodiment, four areas are divided by the center of the length and width of the image to form four quadrants, wherein the detection image 9 is defined as four first comparison areas 90-93, and the first quadrant is the first comparison area 90 , the second quadrant is the first comparison area 91 , the third quadrant is the first comparison area 92 , and the fourth quadrant is the first comparison area 93 . The recorded images 9a to 9f respectively define a plurality of second comparison regions 90a to 93a to 90f to 93f corresponding to the first comparison regions 90 to 93, wherein the first quadrant is the second comparison regions 90a to 90f, the second The quadrants are the second comparison regions 91a to 91f, the third quadrant is the second comparison regions 92a to 92f, and the fourth quadrant is the second comparison regions 93a to 93f.

Then proceed to step 221a, and compare the first distinguishing feature point C0 of each first comparison area 90-93 with the corresponding second comparison area 90a-93a-90f-93f in each recording image 9a-9f The second distinguishing feature points C1 to C6 are calculated. In an embodiment of step 221a, each of the first comparison regions 90-93 of the detection image 9 is compared with the corresponding second comparison regions 90a-93a of each of the recorded images 9a-9f through the SIFT algorithm Calculation is carried out to 90f~93f. During the calculation process, the second distinguishing feature points C1~C6 of each of the recorded images 9a~9f in each of the second comparison areas 90a~93a~90f~93f and the corresponding first The first distinguishing feature point C0 of the comparison areas 90 to 93 can be calculated to obtain similar concentrations of each of the second comparison areas 90a to 93a to 90f to 93f, that is, to represent the second comparison of each of the recorded images 9a to 9f. The degree of similarity between the regions 90 a to 93 a to 90 f to 93 f and the corresponding first comparison regions 90 to 93 in the detection image 9 is detected.

Next, step 222a is performed to determine at least one recorded image most similar to the detected image 9 according to the calculation results of the second comparison areas 90a-93a to 90f-93f in each of the recorded images 9a-9f. The result of step 222a has two aspects. The first aspect is that after the calculation, the second comparison areas of the four quadrants with the highest density all appear in the same recorded image; the second aspect is this The main feature introduced in the embodiment is that the quadrant with the highest concentration is distributed on different recorded images, which will be described in detail below. As shown in FIG. 4A, after step 22a, step 23a is performed to adjust the contour, orientation or scale of at least one recorded image to generate a similar image. As shown in FIG. 4C , in the adjustment procedure of step 23a, step 230a is used to first determine whether all the second comparison areas with the highest similar density appear in the same recorded image. In one embodiment, as shown in FIG. 5C, each of the recorded images 9a-9f in the figure has a similar density value M (M _{90a -90f} , M _{91a -91f} , M _{92a -92f} , M _{93a -93f} ), which are respectively composed of similar concentration values corresponding to four second comparison regions 90a~93a to 90f~93f, and it can be seen from the schematic diagram of FIG. 5C that each second comparison region 90d~93d of the record image 9d The similarity density is M (9, 8, 9, 8) which are the highest, so step 231a is performed to take the recorded image 9d with the highest similarity density in all the comparison areas as the most similar recorded image. Then, the adjustment procedure of step 232a is performed, for example, the adjustment of the image angle of view, which is mainly to perform adjustment and calculation on the recorded image most similar to the detection image in step 231a, so as to obtain the same angle of view, outline or proportion as the detection image. similar images. In one embodiment, the recorded image after step 231a may have a difference in viewing angle orientation, outline or scale between the detected image, such as the state shown in FIG. 3C . Therefore, similar images can be corrected through the adjustment algorithm through the adjustment procedure of step 232a, and the image state shown in FIG. 3D can be obtained. The adjustment algorithm is a conventional technique and will not be repeated here.

In another embodiment, as shown in Figure 5D, it can be clearly seen that similar concentrations M occur on different recorded images. In this embodiment, the similar concentrations of the second comparison regions 90d to 93d of the recorded image 9d are respectively M(8, 7, 6, 9), wherein the second comparison regions 90d and 93d have higher similar concentrations , while the second comparison regions 90e to 93e of the recorded image 9e have similar concentrations M(7, 8, 9, 6), wherein the second comparison regions 91e and 92e have higher similar concentrations. In this case, since there is no single recorded image that is most similar to the detected image 9, after step 231a, step 233a will be performed to generate a similarity image by interpolating the images with higher similarity density picture. In this step, there are many ways of interpolation. For example, in one embodiment, the feature values of the pixels (x, y) at each position of the recorded image 9d in FIG. 5D are used, such as grayscale values. , brightness value or contrast value, etc., and the eigenvalue of the corresponding pixel (x, y) in the recorded image 9e are added and divided by two. A new image obtained after the corresponding pixels of the two recorded images are calculated is a similar image.

In addition to the aforementioned method of adding and dividing the feature values of the two recorded images by two to determine similar images, in another embodiment of the calculation in step 233a, the recorded image 9d and the recorded image 9e can be Each of the second comparison regions 90d to 93d and 90e to 93e defines a weight value, wherein the second comparison region with a larger number of matching second distinctive feature points has a larger weight value. Then, the pixel feature values of the pixels in each of the second comparison areas 90d to 93d and 90e to 93e in the recorded image 9d and the recorded image 9e are calculated according to the weights they have to obtain similar images. The above-mentioned calculation method will be described below.

As shown in FIGS. 6A and 6B , it is assumed that the two recorded images of the second comparison area having relatively high similar density respectively are the recorded images 9d and 9e shown in FIG. 5D . In the second comparison area 90d and the second comparison area 93d in the recorded image 9d, that is, the two second comparison areas 90d and 93d along the positive X-axis and along the Y-axis, the similar density is higher than that in the recorded image 9e The second comparison regions 90e and 93e, while the recorded image 9e has similar concentrations in the second comparison region 91e and the second comparison region 92e, that is, the two second comparison regions 91e and 92e extending along the Y axis in the negative X axis The corresponding second comparison areas 91d and 92d are higher than the recorded image 9d.

In the above-mentioned case, as shown in FIG. 6B , since the second comparison area with high similarity density is located in the right half of the recorded image 9d and the left half of the recorded image 9e, the left half and the left half of the recorded image 9e are used in the calculation. The right half of the weight distribution is adjusted, so the decision of the weight is determined by the distance x between the Y axis. According to the aforementioned rules, the weight distribution of the second comparison area in each quadrant is as follows: 90d 91d 92d 93d w11=(W+x)/2W w12=(Wx)/2W w13=(Wx)/2W w14=(W+x)/2W 90e 91e 92e 93e w21=(Wx)/2W w22=(W+x)/2W w23=(W+x)/2W w24=(Wx)/2W Among them, w11, w12, w13, and w14 represent the weights of the second comparison areas 90d to 93d in the recorded image 9d, respectively, and w21, w22, w23, and w24 represent the weights of the second comparison areas 90e to 93e in the recorded image 9e, respectively. W is the half image height of the recorded images 9d and 9e in the X-axis, and the unit is the number of pixels. x represents the absolute distance between each pixel Pd and Pe and the Y axis on the recorded images 9d and 9e. Therefore, according to the above weights, as shown in FIGS. 6C and 6D , the similar image 9s in FIG. 6D is an image synthesized by the weight calculation of the recorded images 9d and 9e in FIG. 6C . Taking the pixel Pixel 90s on the area 90s of the similar image 9s in FIG. 6D as an example, the characteristic value it has, such as the grayscale value, is the pixel Pixel of the second comparison area 90d corresponding to the area 90s in the recorded image 9d The eigenvalues G90d(x, y) of 90d and the eigenvalues G90e(x,y) of the pixels Pixel 90e in the second comparison area 90e corresponding to the area 90s in the recorded image 9e are added with the weighted eigenvalues. Its calculation equation is expressed as G90s(x,y)= w11*G90d(x,y)+ w21*G90e(x,y). In the same way, the feature value G91s(x,y) of each pixel in the area 91s corresponding to the second comparison area 91d and 91e on the new similar image 9s is expressed as G91s(x,y)=w12*G91d(x,y) )+w22*G91e(x,y) and so on G92s(x,y)=w13*G92d(x,y)+w23*G92e(x,y) and G93s(x,y)= w14*G93d( x,y)+w24*G93e(x,y). After the above-mentioned interpolation calculation, the gray scale of each pixel on the new similar image 9s can be obtained, thereby forming the similar image as shown in FIG. 6D . The similar images 9s interpolated in the above-mentioned manner have the advantages of fast computation and no need to perform slicing through full 3D modeling.

It should be noted that the above-mentioned example is a calculation formula in which the two second comparison regions in the positive X-axis or the Y-axis of the negative X-axis have high similar concentrations. In another embodiment, as shown in FIGS. 6E and 6F , in this embodiment, the second comparison regions with high similar density are located at the two second comparison regions on the positive Y axis and the positive and negative X axes in the recorded image 9d 90d and 91d, and the comparison regions 92e and 93e on both the negative Y-axis and the positive and negative X-axis of the recorded image 9e have high similar density. Since the second comparison areas with high similarity density are located in the upper half of the recorded image 9d and the lower half of the recorded image 9e respectively, the calculation is performed by using the weight distribution of the upper half and the lower half. Therefore, the weight is determined. is determined by the distance y from the X axis. According to the aforementioned rules, the weight distribution of the second comparison area of each quadrant is as follows, so its weight relationship is as follows: 90d 91d 92d 93d w11=(H+y)/2H w12=(H+y)/2H w13=(Hy)/2H w14=(Hy)/2H 90e 91e 92e 93e w21=(Hy)/2H w22=(Hy)/2H w23=(H+y)/2H w24=(H+y)/2H Among them, w11, w12, w13, and w14 represent the weights of the second comparison areas 90d to 93d in the recorded image 9d, respectively, and w21, w22, w23, and w24 represent the weights of the second comparison areas 90e to 93e in the recorded image 9e, respectively. H is the half image width in the Y axis of the recorded images 9d and 9e, and the unit is the number of pixels. y represents the absolute distance between each pixel Pd and Pe and the X axis on the recorded images 9d and 9e. Therefore, according to the above-mentioned weight distribution, as shown in FIG. 6G , the feature value of each pixel in the region 90s1 corresponding to the second comparison regions 90d and 90e on the new similar image 9s1 , such as the grayscale value, can be expressed as G90s1 ( x,y)=w11*G90d(x,y)+w21*G90e(x,y). Similarly, the feature value of each pixel in the region 91s1 corresponding to the second comparison region 91d and 91e on the new similar image 9s1 is G91s1(x,y)=w12*G91d(x,y)+w22*G91e(x,y) ) , and so on. After the above-mentioned interpolation calculation, the similar image 9s1 can be obtained.

If a single recorded image has three relatively high similar densities, as shown in FIG. 6H , in this embodiment, the second comparison regions 90d to 92d of the recorded image 9d have relatively high similar densities, while the recorded image 9d has relatively high similar densities. The second comparison area 93e in the image 9e has a high similarity density, so its weight relation is as follows: 90d 91d 92d 93d w11=(L+r)/2L w12=(L+r)/2L w13=(L+r)/2L w14=(Lr)/2L 90e 91e 92e 93e w21= (Lr)/2L w22=(Lr)/2L w23=(Lr)/2L w24=(L+r)/2L Among them, w11, w12, w13, and w14 represent the weights of the second comparison areas 90d to 93d in the recorded image 9d, respectively, and w21, w22, w23, and w24 represent the weights of the second comparison areas 90e to 93e in the recorded image 9e, respectively. L is one half of the sum H of the half width W and half height of the recorded images 9d and 9e, that is, 0.5(H+W), and the unit is the number of pixels. r represents the absolute distance between each pixel Pd and Pe on the recorded images 9d and 9e and the origin. Therefore, according to the above-mentioned weight assignment, new similar images can be obtained by interpolation. It should be noted that the method of determining the weight is not limited by the distances of x, y, r and W, H and L as described above, and those skilled in the art can define the rules of the weight according to their own needs.

Returning to FIG. 4C , after obtaining the similar image through step 233a, the viewing angle orientation, outline or scale adjustment procedure of step 234a is performed. In this step, the similar image in step 233a and the detected image are mainly adjusted and calculated to obtain a similar image with the same viewing angle, orientation, outline or scale as the detected image. In one embodiment, the similar image obtained through step 233a may have a difference in viewing angle orientation, outline or scale with the detected image, such as the state shown in FIG. 3C . Therefore, similar images can be corrected through adjustment calculation through the viewing angle orientation, outline or scale adjustment procedure in step 234a, and the image state shown in FIG. 3D can be obtained. The adjustment algorithm is a conventional technique and will not be repeated here. Finally, step 24a is performed to compare the detected image with the similar image to determine that at least one different part is obtained (as shown in FIG. 3E ). The procedure of this step is as described in the aforementioned step 24, which is not repeated here.

Please refer to FIG. 7 , which is a schematic structural diagram of an embodiment of the medical image processing apparatus of the present invention. The medical image processing apparatus 3 in this embodiment includes an image capturing unit 30 , an image processing unit 31 and a first image output unit 32 . The image capturing unit 30 in this embodiment is coupled to the image generating device 4 through an analog-to-digital converter (ADC) 37 , and the image capturing unit 30 receives the detection image output by the image generating device 4 , and extract the inspection image from the inspection image. In this embodiment, the image capturing unit 30 is an FPGA device, which is used to perform image correlation calculation. The image generating device 4 may be an ultrasound scanning device, a magnetic resonance imaging (MRI), a positron emission tomography (PET), a computed tomography (CT), a mammography, or an X-ray. Light photography device. The image generating device 4 in this embodiment includes a detection device 40 , a computing host 41 electrically connected to the detection device 40 , and a second display device 42 electrically connected to the computing host 41 .

The image processing unit 31 is electrically connected to the image capture unit 30 for receiving the detection image output from the image capture unit 30 . The image processing unit 31 is further connected to the network 100 through the network interface 36 , and is connected to the cloud server 5 through the network 100 . The cloud server 5 has a database 50 which stores a plurality of recorded images. The above-mentioned recorded images are image records left by the patient when the image generating device 4 performs detection. The medical image processing device 3 obtains a plurality of recorded images from the database 50 and transmits them to the image processing unit 31 for subsequent calculation processing. The image processing unit 31 may be a system on silicon (SOC) with computing capability. In this embodiment, the image processing unit 31 is configured to execute the calculation processes shown in FIGS. 1A to 1B and FIGS. 4A to 4C , which are as described above and will not be repeated here. The image capturing unit 30 and the image processing unit 31 are also coupled to a memory unit 38, such as a flash memory, a dynamic random access memory, or a combination of the foregoing. The first image output unit 32 is electrically connected to the image processing unit 31 and the first display device 39 , and the first image output unit 32 is used for outputting the calculated image to the first display device 39 for display.

The medical image processing device 3 further includes an input interface unit 33 , which is electrically connected to the image processing unit 31 and the input device 34 , such as a keyboard, a mouse, a trackball, or a touch panel. The input interface unit 33 receives the input signal input from the input device 34 and is used to control the mode of the image output from the image processing unit 31 , as shown in FIGS. 7( a ) to ( d ). In addition, in another embodiment, the medical image processing device 3 further includes a second image output unit 35 electrically connected to the image capture unit 30 for returning the image to the computing host 41 of the image generating device 4 . The computing host 41 can output the returned image via the second display device 42. It should be noted that the image returned by the medical image processing device 3 to the computing host 41 may be a detected image, a recorded image, a detected image, a similar image, or a similar image and a detected image for comparison processing The subtracted image or superimposed image obtained after the procedure enables the image generating device 4 to also display the image processed by the medical image processing device 3, so as to help the examiner to make real-time judgment.

To sum up the above, the present invention describes the detection image generated by the detection device through the embodiment, compares the image in the detection image with the past recorded image through the image comparison mechanism, and then selects or selects from the past recorded image. The similar images are obtained by interpolation, and then the detected images can be compared, superimposed or subtracted to generate new image information for the user to make reference for medical judgment. The aforementioned automatic comparison method can solve the problems of inconvenience and easy misjudgment of users when comparing through visual inspection in the conventional technology. In addition, the embodiments of the present invention further describe the calculation mechanism of similar images obtained by interpolation, solve the problem of calculation efficiency of obtaining similar images by slicing through full 3D modeling in the past, and achieve the effect of fast calculation.

The above descriptions are merely for describing the preferred embodiments or examples of the technical means adopted by the present invention to solve the problems, and are not intended to limit the scope of the patent implementation of the present invention. That is, all the equivalent changes and modifications that are consistent with the context of the scope of the patent application of the present invention, or made in accordance with the scope of the patent of the present invention, are all covered by the scope of the patent of the present invention.

2. 2a: Medical Image Processing Methods 20~24: Steps 20a~24a: Steps 220~222: Steps 220a~222a: Steps 230a~233a: Steps 3: Medical image processing device 30: Image capture unit 31: Image processing unit 32: The first image output unit 33: Input interface unit 34: Input device 35: The second image output unit 36: Network connection unit 37: Analog-to-digital conversion unit 38: Memory unit 39: The first display device 4: Image generation device 40: Detection device 41: Computing host 42: Second display device 5: Cloud server 50:Database 9: Detect the image 9a~9f: Recording images 9s: Similar Images 9': overlay image 9sc: 90~93: The first comparison area 90a~93f: Second comparison area 90sc: Simulated Image 900: Difference area 100: Internet C0: The first distinguishing feature point C1~C6: The second distinguishing feature point Pd, Pe: pixel

FIG. 1A is a schematic flowchart of an embodiment of a medical image processing method of the present invention. FIG. 1B is a schematic flow chart of an embodiment of comparing the detected image and the recorded image according to the present invention. FIG. 2 is a schematic diagram of an embodiment of the medical image processing apparatus of the present invention. FIG. 3A is a schematic diagram of an embodiment of the detection image of the present invention. FIG. 3B is a schematic diagram of a plurality of recorded images of the present invention. FIG. 3C is a schematic diagram of an embodiment of a recorded image before performing adjustment calculation in the present invention. FIG. 3D is a schematic diagram of an embodiment of a similar image generated after the adjustment calculation is performed in the present invention. FIG. 3E is a schematic diagram of an embodiment of a comparison process between a similar image and a detected image. FIG. 4A is a schematic flowchart of another embodiment of the medical image processing method of the present invention. FIG. 4B is a schematic flow chart of another embodiment of comparing the detected image and the recorded image according to the present invention. FIG. 4C is a schematic flowchart of an embodiment of generating a similar image. 5A is a schematic diagram of a detection image having a plurality of comparison regions according to the present invention. FIG. 5B is a schematic diagram of a recorded image having a plurality of comparison regions according to the present invention. 5C and 5D are schematic diagrams of similar concentration distributions, respectively. 6A and 6B are schematic diagrams of recorded images with high similar concentrations in different comparison regions, respectively. 6C and 6D are schematic diagrams of an embodiment of the similar images obtained by the adjustment calculation of the present invention. 6E to 6G are schematic diagrams of another embodiment of the similar images obtained by the adjustment calculation of the present invention. FIG. 6H is a schematic diagram of another embodiment of the similar image obtained by the adjustment calculation of the present invention. FIG. 7 is a schematic diagram of a system structure of the medical image processing apparatus of the present invention applied to medical image processing.

2: Medical Image Processing Methods

20~24: Steps

Claims (19)

A medical image processing method, comprising: receiving a detection image of a test subject; receiving a plurality of recorded images of the test subject; Selecting at least one recorded image that is similar to the detected image; adjusting the contour, orientation or scale of the at least one recorded image that is similar to the detected image to generate a similar image; and comparing the detected image image and the similar image to determine and obtain at least one different part. The medical image processing method according to claim 1, wherein the step of comparing the detected image and the recorded images further comprises: detecting feature points of the detected image to obtain a plurality of first distinguishing feature points; Detecting feature points of the recorded images to obtain a plurality of second distinctive feature points for each of the recorded images respectively; and comparing the first distinctive feature points in the detected images with the recorded images For the second distinguishing feature points in the image, the greater the number of matches between the first distinguishing feature points and the second distinguishing feature points, the higher the similarity. The medical image processing method according to claim 2, wherein an algorithm is used to detect the plurality of first and second distinguishing feature points. The medical image processing method as claimed in claim 2, further comprising: Defining a plurality of different first comparison areas for the detection image, so that each of the first comparison areas includes some of the first distinguishing feature points; and obtaining the detection images from the recorded images a first record image and a second record image, and the first record image and the second record image respectively define a plurality of different second comparison areas corresponding to the plurality of first comparison areas , so that each of the second comparison regions includes some of the second distinguishing feature points. The medical image processing method according to claim 4, wherein the step of adjusting the first recorded image and the second recorded image further comprises: each of the first recorded image and the second recorded image The two comparison areas define a weight value, wherein the second comparison area with a larger number of matches of the second distinctive feature points has a larger weight value; and each of the first recorded image and the second recorded image has a larger weight value. The pixel features in the two comparison regions are calculated according to the weights in the second comparison region to generate the similar image. The medical image processing method according to claim 5, wherein the determination method of the weight value is determined according to the distance between each pixel and a reference, and the reference is the center of the first or second recorded image, through the The horizontal or vertical axis of the center of the first or second recorded image. The medical image processing method as claimed in claim 1, wherein the step of adjusting the at least one recorded image further comprises interpolating a first recorded image and a second recorded image that are similar to the detected image, to produce the similar image. The medical image processing method as claimed in claim 1, wherein the step of comparing the detected image and the similar image further comprises subtracting or superimposing the similar image and the detected image to generate a difference image. A medical image processing device, comprising: an image processing unit electrically connected to a database storing a plurality of recorded images, the image processing unit receiving the plurality of recorded images and a detection image from an image generating device , wherein the image processing unit detects a plurality of first distinguishing feature points on the detected image, and respectively detects a plurality of second distinguishing feature points on the plurality of recorded images, and converts the plurality of detected image The first distinguishing feature points are respectively compared with the second distinguishing feature points in the plurality of recorded images, so as to judge and select at least one record image that is similar to the detected image from the plurality of recorded images, wherein The image processing unit adjusts the outline, orientation or scale of the at least one recorded image similar to the detected image to generate a similar image; and a first image output unit electrically connected to the image processing unit for to output the similar image. The medical image processing device as described in claim 9, wherein the image processing unit subtracts or superimposes the detected image and the similar image to generate a difference image. The medical image processing device as described in item 10 of the scope of the application further comprises an input interface unit, which is electrically connected to the image processing unit and an input device to receive the input signal input by the input device, and pass the input signal through the input signal. Controls the mode in which the image processing unit processes images. The medical image processing device as described in item 9 of the scope of the patent application further comprises a network connection unit electrically connected to the database and the image processing unit, so as to obtain the plurality of recorded images from the database and transmit them to the image processing unit. The medical image processing device as described in claim 9, wherein the image processing unit detects the first and second distinguishing feature points through an algorithm. The medical image processing device according to claim 9, wherein when comparing the detected image with the recorded images, the image processing unit further defines a plurality of different first comparison regions for the detected image, so that Each of the first comparison regions contains some of the first distinguishing feature points, and then obtains a first recorded image and a second recorded image that are similar to the detected image from the recorded images, and compares the The first recorded image and the second recorded image respectively define a plurality of different second comparison areas corresponding to the plurality of first comparison areas, so that each of the second comparison areas includes some of the second differences Feature points. The medical image processing device as claimed in claim 14, wherein when the image processing unit adjusts the first recorded image and the second recorded image, the first recorded image and the second recorded image are further A weight value is defined for each second comparison area, wherein the second comparison area with more second distinctive feature points matching number has a larger weight value, and then the first recorded image and the second recorded image are The pixel feature in each second comparison area is calculated according to the weight value of the second comparison area to obtain the similar image. The medical image processing device as claimed in claim 15, wherein the image processing unit determines the weight value according to the distance between each pixel and a reference, wherein the reference is the center of the first or second recorded image, through the The horizontal or vertical axis of the center of the first or second recorded image. The medical image processing device according to claim 9, wherein when the image processing unit adjusts the at least one recorded image, it further processes a first recorded image and a second recorded image that are similar to the detected image Interpolation is used to generate the similar image. The medical image processing device according to claim 9, further comprising an image capture unit coupled to the image generation device and electrically connected to the image processing unit, the image capture unit receiving the image output from the image generation device A detection image is extracted from the detection image, and the detection image is sent to the image processing unit. The medical image processing device as described in item 18 of the scope of the application further comprises a second image output unit electrically connected to the image capture unit for the detection image, the recorded image, the detection image or the The similar image is sent back to the image generating device.

Images