Myeloma focus semi-automatic threshold segmentation method and system based on WB-DWI

Abstract

The invention relates to the technical field of medical image processing, in particular to a myeloma lesion semi-automatic threshold segmentation method and system based on WB-DWI, and the method comprises the steps: taking out an original image, determining a window width and a window level, setting the window width and the window level as the window width and the window level suitable for a radiologist to observe a lesion, and carrying out the data normalization according to the given window width and the window level; calculating a background threshold value and removing part of the background; calculating a focus threshold and obtaining a suspicious focus; physicians perform quality control to delete a focus area and exclude a T2 penetration effect area. A background threshold is a relative threshold instead of an absolute threshold, so that generalization ability and robustness are improved; data normalization based on window width and window level is added, the lesion threshold value can be more accurate after execution, and lesion can be detected more sensitively; by excluding the specified coordinate range in the Z-axis direction and deleting the area above the C4 centrum, the workload of doctors is reduced to a certain extent.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a semi-automatic threshold segmentation method and system for myeloma lesions based on WB-DWI. Background Art

[0002] Multiple myeloma (MM) is the second most common malignant tumor in hematology, characterized by the abnormal proliferation of malignant plasma cells throughout the bone marrow. Whole body diffusion weighted imaging (WB-DWI) can detect bone marrow infiltration at an early stage, which is of great value for the detection and follow-up of MM lesions, making it possible to non-invasively and objectively quantify MM tumor burden. However, unlike most other tumors, MM lesions are multiple and heterogeneous. How to accurately and efficiently segment MM lesions on WB-DWI images is a technical bottleneck in this field.

[0003] There are few methods for segmenting MM lesions based on WB-DWI reported in the literature. Among them, the threshold-based semi-automatic segmentation method uses the high signal essential characteristics of the lesion on the WB-DWI image, combined with physician quality control, and has the advantages of being simple and easy to implement and not dependent on computing power. It has been applied to clinical value research. The algorithm is summarized as follows: First, the b value is calculated based on the original DWI = 999s / mm 2 The DWI images were obtained by maximizing the contrast between the lesions and the background tissues observed by naked eyes, and part of the background was removed with a background threshold of 97. Then, for the remaining areas, the classic Otsu algorithm was used to determine the lesion threshold to obtain suspicious lesions. Finally, for suspicious lesions, the physician quality control manually deleted physiological high signals such as brain, lymph nodes, and intestines, and excluded those with ADC values ​​>2.0×10 -3 mm 2 / s area to eliminate the influence of T2 penetration effect.

[0004] However, this method still has the following shortcomings: ① The background threshold of 97 is an absolute threshold, and it is difficult to ensure repeatability in other data sets; ② Data normalization based on window width and window position is not performed, and we found through experiments that this process may optimize the threshold (detect lesions more sensitively); ③ The manual deletion of physiological high signals by the physician is labor-intensive.

[0005] In this context, taking the above-mentioned literature methods as an important reference, a semi-automatic threshold segmentation method with strong universality, reasonable threshold determination, and physician-friendly is proposed, which is expected to help physicians segment MM lesions on WB-DWI images more accurately and efficiently, and provide a basis for the further popularization and promotion of related clinical value research. Summary of the invention

[0006] In view of the above problems existing in the prior art, the present invention is proposed.

[0007] Therefore, the present invention provides a semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI, aiming to make up for the limitations of the prior art. After testing and verification, the method has good robustness, high accuracy and is simple and easy to implement, and can effectively solve the problems in the above-mentioned background technology.

[0008] To solve the above technical problems, the present invention provides the following technical solutions, a semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI, comprising: taking out the original image and determining the window width and window position, setting the window width and window position suitable for radiologists to observe lesions, and normalizing the data according to the given window width and window position; based on the normalized image, calculating the background threshold and removing part of the background; based on the normalized image and the area after removing part of the background, calculating the lesion threshold and obtaining suspicious lesions; based on the original image, the suspicious lesion area, and the ADC map, the physician quality controls and deletes the lesion area and excludes the T2 penetration effect area.

[0009] As a preferred solution of the semi-automatic threshold segmentation method of myeloma lesions based on WB-DWI of the present invention, wherein: the original image is a high b-value DWI image;

[0010] The data normalization is expressed as follows: let the original image be I, I be a three-dimensional matrix of a×b×c, a, b, c represent the number of elements of I in the X, Y, and Z directions respectively, let the window width be WW, the window position be WC, the values ​​of WW and WC are determined by the radiologist, and the standard is that when WW=ww and WC=wc are obtained, the radiologist can observe the lesion on the original image I, that is, the lesion is clearly displayed and the contrast between the lesion and the background is large, then ww and wc are set to WW and WC;

[0011] The upper and lower bounds are expressed as,

[0012] p=WC+0.5×WW

[0013] q=WC-0.5×WW

[0014] Among them, p represents the upper limit value, and q represents the lower limit value;

[0015] Grayscale values ​​greater than the upper bound are mapped to 1, grayscale values ​​less than the lower bound are mapped to 0, and grayscale values ​​between the lower bound and the upper bound are linearly mapped to between 0 and 1;

[0016] The normalized image is J, which is a three-dimensional matrix of a×b×c. The calculation method of J is to traverse each element X=x, Y=y, Z=z of the matrix I. For elements greater than p, the value of the corresponding position of J is assigned to 1; for elements less than q, the value of the corresponding position of J is assigned to 0; for elements in the range of [q,p], the value of the corresponding position of J is assigned to the value of I linearly mapped from [q,p] to [0,1]. The calculation process is expressed as,

[0017]

[0018] Where I(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in matrix I, and J(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in matrix J.

[0019] As a preferred embodiment of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI of the present invention, the calculation of the background threshold includes taking the grayscale average value of the normalized image as the background threshold, th1 is represented as the background threshold, and th1 is calculated as the average value of all elements of the matrix J. The calculation process is represented as follows:

[0020]

[0021] The removing of part of the background comprises extracting a region in the normalized image whose gray value is greater than a background threshold value to obtain mask1;

[0022] When the mask 1 required is M1, M1 is a three-dimensional matrix of a×b×c. The calculation method of M1 is to traverse each element of the matrix J. For elements greater than th1, the value of the corresponding position of M1 is assigned to 1; for elements less than or equal to th1, the value of the corresponding position of M1 is assigned to 0. The calculation process is expressed as:

[0023]

[0024] Among them, M1(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M1, and J(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix J.

[0025] As a preferred embodiment of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI described in the present invention, the calculation of the lesion threshold includes extracting the values ​​of all M1=1 element positions in the matrix J to obtain a vector V, and using the classic Otsu algorithm to calculate the threshold of the vector V, and the obtained threshold is th2.

[0026] As a preferred embodiment of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI of the present invention, the suspicious lesions are represented by extracting the area with gray value greater than the lesion threshold in the normalized image to obtain mask 2;

[0027] Assume that the mask2 to be sought is M2, which is a three-dimensional matrix of a×b×c. The calculation method of M2 is to traverse each element of the matrix J. For elements greater than th2, the value of the corresponding position of M2 is assigned to 1; for elements less than or equal to th2, the value of the corresponding position of M2 is assigned to 0. The calculation process is expressed as:

[0028]

[0029] Among them, M2(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in the matrix M2, and J(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in the matrix J.

[0030] As a preferred embodiment of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI of the present invention, wherein: the physician quality control deletion of the lesion area includes deleting the area above the C4 vertebra, the physician observes the original image and determines the coordinate range of the area above the C4 vertebra in the Z-axis direction, and excludes the area within the specified coordinate range in the Z-axis direction from mask2 to obtain mask3;

[0031] Suppose the mask3 is M3, which is a three-dimensional matrix of a×b×c. Suppose the doctor determines the coordinate range of the area above the C4 vertebral body in the Z-axis direction to be [z1,z2]. Then the calculation method of M3 is:

[0032]

[0033] Wherein, M3(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M3, and M2(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M2;

[0034] The remaining physiological high signal was deleted. The physician used the ROI delineation tool to superimpose mask3 and the original image for observation, and manually deleted the physiological high signal area in mask3 that the physician considered, to obtain mask4.

[0035] Suppose the mask4 to be sought is M4, which is a three-dimensional matrix of a×b×c. Suppose the physiological high signal area considered by the physician after observation is S, then the calculation method of M4 is:

[0036]

[0037] Among them, M4(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M4, and M3(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M3.

[0038] As a preferred embodiment of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI of the present invention, the exclusion of T2 penetration effect areas includes taking out the ADC map and excluding the ADC values ​​in mask4 greater than 2.0×10 -3 mm 2 / s area, and obtain mask5, which is the final lesion area;

[0039] Let the mask5 be M5, M5 is a three-dimensional matrix of a×b×c, and let the ADC image be I ADC , I ADC For a three-dimensional matrix of a×b×c, the calculation method of M5 is:

[0040]

[0041] Wherein, M5(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in matrix M5, M4(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in matrix M4, I ADC (x,y,z represent the matrix I ADC The value of the element at position X=x, Y=y, Z=z.

[0042] As a preferred solution of the semi-automatic threshold segmentation system for myeloma lesions based on WB-DWI of the present invention, it includes: an image extraction module, a data normalization module, a background removal module, a lesion detection module, and a physician quality control module;

[0043] The image extraction module extracts the original image and determines the window width and window position, and sets the window width and window position suitable for radiologists to observe lesions;

[0044] The data normalization module performs data normalization according to a given window width and window level;

[0045] The background removal module calculates the background threshold and removes part of the background;

[0046] The lesion detection module calculates the lesion threshold and obtains suspicious lesions;

[0047] The physician quality control module deletes the lesion area and excludes the T2 penetration effect area.

[0048] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and wherein when the processor executes the computer program, the processor implements the steps of any one of the methods for semi-automatic threshold segmentation of myeloma lesions based on WB-DWI.

[0049] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of any one of the methods for semi-automatic threshold segmentation of myeloma lesions based on WB-DWI are implemented.

[0050] The beneficial effects of the present invention are as follows: the background threshold adopts a relative threshold rather than an absolute threshold, which improves generalization ability and robustness; data normalization based on window width and window position is added, and the lesion threshold after execution may be more accurate, which helps to detect lesions more sensitively; by excluding the specified coordinate range in the Z-axis direction and deleting the area above the C4 vertebra, the workload of doctors is reduced to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0052] Figure 1 A schematic flow chart of a semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI provided in an embodiment of the present invention.

[0053] Figure 2 The original image of high b-value DWI of a MM patient (i.e., the input image in step 1) used in the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI provided in the embodiment of the present invention shows a total of 6 coronal sections (6 columns).

[0054] Figure 3 This is a rendering of mask 1 obtained by the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI provided in an embodiment of the present invention.

[0055] Figure 4 This is a mask 2 effect diagram obtained by the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI provided in an embodiment of the present invention.

[0056] Figure 5 This is a mask 3 effect diagram obtained by the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI provided in an embodiment of the present invention.

[0057] Figure 6This is a mask 4 rendering obtained by the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI provided in an embodiment of the present invention.

[0058] Figure 7 This is a mask 5 rendering obtained by the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI provided in an embodiment of the present invention.

[0059] Figure 8 A comparison diagram of the results of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI provided in an embodiment of the present invention and whether data normalization is performed according to a given window width and window level is shown. A total of 2 cross-sections (2 rows) of the horizontal axis are shown, the left column is the original image, the middle column is the segmentation result of direct normalization (i.e., the grayscale value of each voxel in the original image is mapped to between 0 and 1 according to the interval range from the minimum to the maximum grayscale of the image), and the right column is the segmentation result after data normalization according to a given window width and window level; referring to the original image, it can be seen that the segmentation result after data normalization according to the given window width and window level is more sensitive and more consistent with the lesion area of ​​the original image. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0061] Example 1

[0062] Reference Figure 1-Figure 8 , which is the first embodiment of the present invention, provides a semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI, comprising:

[0063] The clinical value of assessing the tumor burden of multiple myeloma (MM) based on whole-body diffusion-weighted imaging (WB-DWI) is significant, and lesion segmentation is an important part of it. However, due to the particularity of multiple and heterogeneous MM lesions, how to accurately and efficiently segment MM lesions on WB-DWI images is a technical bottleneck in solving this clinical problem.

[0064] At present, in the literature of this field, the threshold-based semi-automatic segmentation method is a relatively mature technology, but it still has some shortcomings, including: the background threshold is an absolute threshold, which is difficult to ensure repeatability; data normalization based on window width and window position is not performed, and this process may optimize the threshold; the workload of doctors manually deleting physiological high signals is large. To this end, the present invention takes the literature method as an important reference and proposes a semi-automatic threshold segmentation method with strong universality, reasonable threshold determination, and doctor-friendly. The innovation of this method lies in: the background threshold adopts a relative threshold, which improves the generalization ability; the data normalization based on window width and window position is added, and the lesion threshold determination may be more accurate; by excluding the specified area, the workload of doctors is reduced. After testing and verification, this method has good robustness, high accuracy and simplicity, and can effectively make up for the limitations of previous methods.

[0065] The present invention can help physicians to segment MM lesions on WB-DWI images more accurately and efficiently, and provides a basis for further popularization and promotion of related clinical value research.

[0066] Step 1: normalize the data according to the given window width and window level; the specific process is:

[0067] Input a high b-value DWI original image of a MM patient, such as Figure 2 As shown. Determine the window width and window level, and set them to the window width and window level that the radiologist considers suitable for observing the lesion; in this embodiment, the window width is set to 100, and the window level is set to 25. Calculate the upper and lower bounds, and the algorithm is: upper bound = window level + 0.5 × window width, lower bound = window level - 0.5 × window width; in this embodiment, the upper bound = 25 + 0.5 × 100 = 75, and the lower bound = 25 - 0.5 × 100 = -25. Map the grayscale value of each voxel of the original image to between 0 and 1 (i.e., the normalized image) from the lower bound to the upper bound, and the algorithm is: the grayscale value greater than the upper bound is mapped to 1, the grayscale value less than the lower bound is mapped to 0, and the grayscale value between the lower bound and the upper bound is linearly mapped to between 0 and 1; after the original image of this embodiment undergoes this process, a normalized image is obtained, and the grayscale value is a floating point number between 0 and 1.

[0068] The high b-value DWI image (hereinafter referred to as the original image) was taken out and the window width and window position were determined and set to the window width and window position that the radiologist considered suitable for observing the lesion;

[0069] Specifically, let the original image be I, I be a three-dimensional matrix of a×b×c, a, b, c represent the number of elements of I in the X, Y, and Z directions, respectively. Let the window width be WW, the window level be WC, the values ​​of WW and WC are freely determined by the radiologist, and the standard is that when a certain value (WW=ww, WC=wc) is obtained, the radiologist believes that the lesion can be well observed on the original image I, that is, the lesion is clearly displayed and the contrast between the lesion and the background is large, then ww and wc are set to WW and WC.

[0070] ② Calculate the upper and lower bounds. The algorithm is: upper bound = window position + 0.5 × window width, lower bound = window position - 0.5 × window width;

[0071] Specifically, let the upper bound be p and the lower bound be q, then p=WC+0.5×WW, q=WC-0.5×WW.

[0072] ③ Map the grayscale value of each voxel in the original image to between 0 and 1 (i.e., the normalized image) according to the interval range from the lower limit to the upper limit. The algorithm is: the grayscale value greater than the upper limit is mapped to 1, and the grayscale value less than the lower limit is mapped to 0. The grayscale values ​​between the lower limit and the upper limit are linearly mapped to between 0 and 1.

[0073] Specifically, let the normalized image be J, where J is a three-dimensional matrix of a×b×c. The calculation method of J is: traverse each element of the matrix I (X=x, Y=y, Z=z), for elements greater than p, the value of the corresponding position of J is assigned to 1; for elements less than q, the value of the corresponding position of J is assigned to 0; for elements in the range of [q,p], the value of the corresponding position of J is assigned to the value of I linearly mapped from [q,p] to [0,1]. The calculation process can be expressed as:

[0074]

[0075] Among them, I(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in matrix I, and J(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in matrix J.

[0076] Step 2: Calculate the background threshold and remove part of the background. The specific process is as follows:

[0077] Calculate the background threshold, the algorithm is: take the grayscale average value of the normalized image as the background threshold; in this embodiment, the background threshold = the grayscale average value of the normalized image = 0.2929. Remove part of the background, the algorithm is: extract the area in the normalized image with a grayscale value greater than the background threshold (obtain mask 1); in this embodiment, extract the area in the normalized image with a grayscale value greater than 0.2929, and the obtained mask 1 is as follows Figure 3 shown.

[0078] ① Calculate the background threshold. The algorithm is: take the grayscale average of the normalized image as the background threshold;

[0079] Specifically, let the background threshold be th1, then th1 is calculated as the average value of all elements of matrix J. The calculation process can be expressed as:

[0080]

[0081] ② Remove part of the background. The algorithm is: extract the area in the normalized image whose gray value is greater than the background threshold (obtain mask 1). [Note: Mask is an important concept in the field of image processing. It is used to specify a certain area of ​​the image. It is usually a binary image of the same size as the original image, in which the selected area is marked as 1 and the rest of the area is marked as 0.]

[0082] Specifically, let the mask 1 to be sought be M1, and M1 is a three-dimensional matrix of a×b×c. The calculation method of M1 is: traverse each element of the matrix J (X=x, Y=y, Z=z), and for the elements greater than th1, the value of the corresponding position of M1 is assigned to 1; the value of the corresponding position of M1 in the remaining area is assigned to 0. The calculation process can be expressed as:

[0083]

[0084] Among them, M1(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M1, and J(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix J.

[0085] Step 3: Calculate the lesion threshold and obtain suspicious lesions; the specific process is as follows:

[0086] The algorithm for calculating the lesion threshold is: extract the grayscale value within the range of mask 1 of the normalized image, and use the classic Otsu algorithm to automatically determine the threshold within this range; after calculation in this embodiment, the lesion threshold = 0.6431. The algorithm for obtaining suspicious lesions is: extract the area in the normalized image with a grayscale value greater than the lesion threshold (obtain mask 2); this embodiment extracts the area in the normalized image with a grayscale value greater than 0.6431, and the obtained mask 2 is as follows Figure 4 shown.

[0087] ① Calculate the lesion threshold. The algorithm is: extract the gray value within the range of mask 1 of the normalized image, and use the classic Otsu algorithm to automatically determine the threshold within this range;

[0088] Specifically, let the lesion threshold be th2, then the calculation method of th2 is: extract the values ​​of all M1=1 element positions in the matrix J to obtain the vector V, and use the classic Otsu algorithm to find the threshold of the vector V, and the threshold obtained is th2. The Otsu algorithm, also known as the "maximum inter-class variance method", is a commonly used image binarization method used to convert grayscale images into binary images. It was proposed by Japanese scholar Otsu in 1979.

[0089] ② Obtain suspicious lesions. The algorithm is: extract the area in the normalized image whose grayscale value is greater than the lesion threshold (obtain mask 2).

[0090] Specifically, let the mask 2 to be sought be M2, and M2 is a three-dimensional matrix of a×b×c. The calculation method of M2 is: traverse each element of the matrix J (X=x, Y=y, Z=z), and for the elements greater than th2, the value of the corresponding position of M2 is assigned to 1; the value of the corresponding position of M2 in the remaining area is assigned to 0. The calculation process can be expressed as:

[0091]

[0092] Among them, M2(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in the matrix M2, and J(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in the matrix J.

[0093] Step 4: Physician quality control deletes the lesion area; the specific process is as follows:

[0094] The area above the C4 vertebra (physiological high signals such as the brain and head and neck glands) is deleted. The doctor observes the original image and determines the coordinate range of the area above the C4 vertebra in the Z-axis direction, and then excludes the area within the specified coordinate range in the Z-axis direction from mask 2 (obtaining mask 3); in this embodiment, the doctor observes that the coordinate range in the Z-axis direction to be excluded is 312-400, and the mask 3 obtained after execution is as follows: Figure 5 shown.

[0095] The remaining physiological high signals (intestines, kidneys, lymph nodes, etc.) are deleted. The physician uses a ROI delineation tool (such as ITK-SNAP) to overlay mask 3 with the original image for observation, and manually deletes the physiological high signal areas in mask 3 that the physician considers (obtaining mask 4). In this embodiment, the physician observes and edits in ITK-SNAP, and the obtained mask 4 is as follows: Figure 6 shown.

[0096] ① Delete the area above the C4 vertebra (physiological high signals such as the brain and head and neck glands). The physician observes the original image and determines the coordinate range of the area above the C4 vertebra in the Z-axis direction, and then excludes the area within the specified coordinate range in the Z-axis direction from Mask 2 (obtaining Mask 3);

[0097] Specifically, let the mask 3 to be sought be M3, and M3 is a three-dimensional matrix of a×b×c. Let the doctor determine the coordinate range of the area above the C4 vertebra in the Z-axis direction as [z1,z2]. Then the calculation method of M3 is:

[0098]

[0099] Among them, M3(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M3, and M2(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M2.

[0100] ② Delete the remaining physiological high signals (intestines, kidneys, lymph nodes, etc.). The physician uses a ROI delineation tool (such as ITK-SNAP) to superimpose mask 3 and the original image for observation, and manually deletes the physiological high signal areas in mask 3 that the physician considers (obtaining mask 4).

[0101] Specifically, let the mask 4 to be sought be M4, which is a three-dimensional matrix of a×b×c. Let the physiological high signal area (a set of coordinate points) considered by the physician after observation be S. Then the calculation method of M4 is:

[0102]

[0103] Among them, M4(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M4, and M3(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M3.

[0104] Step 5: Eliminate the T2 penetration effect area; the specific process is as follows:

[0105] Take out the ADC map and exclude the ADC values ​​in mask 4 that are greater than 2.0×10 -3 mm 2 / s area (obtaining mask 5), i.e., the final lesion area; after executing this embodiment, mask 5 is obtained as follows Figure 7 shown.

[0106] The advantages of the present invention over the prior art are illustrated below through this embodiment: ① The background threshold uses the grayscale average value of the normalized image (0.2929 in this embodiment), which is a relative threshold, and can avoid the influence of equipment factors on the image grayscale, thereby improving robustness and repeatability; ② Data normalization based on window width and window position is added, which is in line with the perspective of radiologists, and the lesion threshold determination may be more accurate, which helps to detect lesions more sensitively; to illustrate the significance of this step, we output and compare the different results of direct normalization (i.e., mapping the grayscale value of each voxel of the original image to between 0 and 1 according to the interval range from the minimum to the maximum grayscale of the image) and data normalization according to a given window width and window position, as shown in FIG. Figure 8 As shown, referring to the original image ( Figure 8 Left column), we can see the segmentation result of data normalization according to the given window width and window level ( Figure 8 The right column) is a more direct normalized segmentation result ( Figure 8 ③ By excluding the specified coordinate range in the Z-axis direction (312-400 in this embodiment), the area above the C4 vertebra is deleted, which saves the doctor the workload of manually deleting the brain and head and neck glands.

[0107] Exclude the T2 penetration effect area, specifically: take out the ADC map and exclude the ADC value greater than 2.0×10 in mask 4 -3 mm 2 / s area (obtaining mask 5), that is, the final lesion area. The principle of this process is explained as follows: the diffusion-restricted area is the final desired lesion area, but the high signal on the DWI image does not necessarily mean that the diffusion is restricted. It may also be the T2 penetration effect (DWI high signal without diffusion restriction), and its ADC value does not decrease. Therefore, it is often used to remove the area with too high ADC value to exclude the T2 penetration effect. The 2.0×10 -3 mm 2 / s is the threshold reported in the literature.

[0108] Specifically, let the mask 5 to be sought be M5, where M5 is a three-dimensional matrix of a×b×c. Let the ADC graph be I ADC , I ADC is a three-dimensional matrix of a×b×c. Then the calculation method of M5 is:

[0109]

[0110] Wherein, M5(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in matrix M5, M4(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in matrix M4, I ADC (x,y,z) represents the matrix I ADCThe value of the element at position X=x, Y=y, Z=z.

[0111] The above is a schematic scheme of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI of the present embodiment. It should be noted that the technical scheme of the system of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI and the technical scheme of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI above belong to the same concept, and the details of the technical scheme of the system of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI in the present embodiment that are not described in detail can all be referred to the description of the technical scheme of the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI above.

[0112] The semi-automatic threshold segmentation system for myeloma lesions based on WB-DWI in this embodiment includes an image extraction module, a data normalization module, a background removal module, a lesion detection module, and a physician quality control module;

[0113] The image extraction module extracts the original image and determines the window width and window position, and sets the window width and window position suitable for radiologists to observe lesions;

[0114] The data normalization module performs data normalization according to a given window width and window level;

[0115] The background removal module calculates the background threshold and removes part of the background;

[0116] The lesion detection module calculates the lesion threshold and obtains suspicious lesions;

[0117] The physician quality control module deletes the lesion area and excludes the T2 penetration effect area.

[0118] This embodiment further provides a computing device, which is applicable to the case of a semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI, including:

[0119] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0120] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0121] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0122] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. A semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI, characterized by: include, The original image is taken out and the window width and window position are determined, and the window width and window position are set to be suitable for radiologists to observe the lesions, and the data is normalized according to the given window width and window position; Based on the normalized image, calculate the background threshold and remove part of the background; Based on the normalized image and the area after removing some background, the lesion threshold is calculated and the suspicious lesions are obtained; Based on the original image, suspicious lesion area, and ADC map, the physician quality control deleted the lesion area and excluded the T2 penetration effect area.

2. The semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI according to claim 1, characterized in that: The original image is a high b-value DWI image; The data normalization is represented as follows: the original image is I, I is a three-dimensional matrix of a×b×c, a, b, c represent the number of elements of I in the X, Y, and Z directions respectively, the window width is WW, and the window position is WC; The upper and lower bounds are expressed as, p=WC+0.5×WW q=WC-0.5×WW Among them, p represents the upper limit value, and q represents the lower limit value; Grayscale values ​​greater than the upper bound are mapped to 1, grayscale values ​​less than the lower bound are mapped to 0, and grayscale values ​​between the lower bound and the upper bound are linearly mapped to between 0 and 1; The normalized image is J, which is a three-dimensional matrix of a×b×c. The calculation method of J is to traverse each element X=x, Y=y, Z=z of the matrix I. For elements greater than p, the value of the corresponding position of J is assigned to 1; for elements less than q, the value of the corresponding position of J is assigned to 0; for elements in the range of [q,p], the value of the corresponding position of J is assigned to the value of I linearly mapped from [q,p] to [0,1]. The calculation process is expressed as, Where I(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in matrix I, and J(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in matrix J.

3. The semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI as claimed in claim 2, characterized in that: The calculation of the background threshold includes taking the grayscale average value of the normalized image as the background threshold, th1 is represented as the background threshold, then th1 is calculated as the average value of all elements of the matrix J, and the calculation process is represented as follows: The removing of part of the background comprises extracting a region in the normalized image whose gray value is greater than a background threshold value to obtain mask1; Assume that the mask 1 to be sought is M1, which is a three-dimensional matrix of a×b×c. The calculation method of M1 is to traverse each element of the matrix J. For elements greater than th1, the value of the corresponding position of M1 is assigned to 1; for elements less than or equal to th1, the value of the corresponding position of M1 is assigned to 0. The calculation process is expressed as: Among them, M1(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M1, and J(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix J.

4. The semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI as claimed in claim 3, characterized in that: The calculation of the lesion threshold includes extracting the values ​​of all M1=1 element positions in the matrix J to obtain a vector V, and using the classic Otsu algorithm to calculate the threshold of the vector V, and the obtained threshold is th2.

5. The semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI according to claim 4, characterized in that: The suspicious lesion is obtained by extracting the area whose gray value is greater than the lesion threshold in the normalized image to obtain mask2; Assume that the mask2 to be sought is M2, which is a three-dimensional matrix of a×b×c. The calculation method of M2 is to traverse each element of the matrix J. For elements greater than th2, the value of the corresponding position of M2 is assigned to 1; for elements less than or equal to th2, the value of the corresponding position of M2 is assigned to 0. The calculation process is expressed as: Among them, M2(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in the matrix M2, and J(x,y,z) represents the value of the element at position X=x,Y=y,Z=z in the matrix J.

6. The semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI according to claim 5, characterized in that: The physician quality control deletion of the lesion area includes deleting the area above the C4 vertebra, wherein the physician observes the original image and determines the coordinate range of the area above the C4 vertebra in the Z-axis direction, and excludes the area within the specified coordinate range in the Z-axis direction from mask2 to obtain mask3; Suppose the mask3 is M3, which is a three-dimensional matrix of a×b×c. Suppose the doctor determines the coordinate range of the area above the C4 vertebral body in the Z-axis direction to be [z1,z2]. Then the calculation method of M3 is: Wherein, M3(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M3, and M2(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M2; The remaining physiological high signal was deleted. The physician used the ROI delineation tool to superimpose mask3 and the original image for observation, and manually deleted the physiological high signal area in mask3 that the physician considered, to obtain mask4. Assume that the mask4 to be sought is M4, which is a three-dimensional matrix of a×b×c. The physiological high signal area considered by the physician after observation is S, and the calculation method of M4 is: Among them, M4(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M4, and M3(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in the matrix M3.

7. The semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI according to claim 6, characterized in that: The method of excluding the T2 penetration effect area includes taking out the ADC map and excluding the ADC value in mask4 greater than 2.0×10 -3 mm 2 / s area, and obtain mask5, which is the final lesion area; Let the mask5 be M5, M5 is a three-dimensional matrix of a×b×c, and let the ADC image be I ADC , I ADC For a three-dimensional matrix of a×b×c, the calculation method of M5 is: Wherein, M5(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in matrix M5, M4(x, y, z) represents the value of the element at position X=x, Y=y, Z=z in matrix M4, I ADC (x,y,z) represents the matrix I ADC The value of the element at position X=x, Y=y, Z=z.

8. A system based on the semi-automatic threshold segmentation method for myeloma lesions based on WB-DWI according to any one of claims 1 to 7, characterized in that: Including image extraction module, data normalization module, background removal module, lesion detection module, and physician quality control module; The image extraction module extracts the original image and determines the window width and window position, and sets the window width and window position suitable for radiologists to observe lesions; The data normalization module performs data normalization according to a given window width and window level; The background removal module calculates the background threshold and removes part of the background; The lesion detection module calculates the lesion threshold and obtains suspicious lesions; The physician quality control module deletes the lesion area and excludes the T2 penetration effect area.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images