A 3D bone segmentation method based on improved region growing

Abstract

A three-dimensional bone segmentation method based on improved region growing includes: first, reading in three-dimensional CT bone data; second, using an adaptive threshold segmentation method to separate the bone area from the background area to obtain all bone areas; then using morphological operations to remove the adhesion between the broken bones to obtain the core area of ​​each broken bone; finally, using these core areas as seed points for region growth, and performing region growth on multiple areas at the same time, thereby ensuring the accuracy of bone segmentation while ensuring the efficiency of segmentation, that is, ensuring that each pixel point belongs to a unique area and is the area closest to the seed point. According to the above steps, the pixel points contained in all the broken bones of the three-dimensional bone image and the broken bone number to which each pixel point belongs can be determined. The present invention can be applied to three-dimensional bone segmentation to help complete disease diagnosis and treatment, lesion analysis, etc., and help doctors better complete diagnosis and understand the patient's condition.

Description

Technical field:

[0001] The present invention belongs to the field of medical image processing and relates to a three-dimensional bone segmentation method for medical CT bone images, which is particularly used to help doctors perform lesion analysis, disease diagnosis and treatment, etc. Background technology:

[0002] With the development of orthopedic surgery, the demand for medical image processing technologies such as bone segmentation is increasing, including in the fields of maxillofacial surgery, spinal surgery, and joint replacement. More and more doctors are using medical image processing technology to assist in diagnosis and treatment, improving the safety and effectiveness of surgery.

[0003] At the same time, the rapid development and popularization of computer technology has provided more possibilities for the development of medical image processing technology. Among the existing medical image processing technologies, bone segmentation technology is one of the important branches. However, there are some common problems with existing bone segmentation technologies. The existing mainstream methods for bone segmentation are divided into traditional methods and deep learning methods. Among them, traditional methods mainly include threshold segmentation, region growing, boundary tracking, etc. These methods have certain limitations. For example, threshold segmentation is easily affected by noise and external interference. The region growing method requires manual specification of growth points, and the growth criteria are difficult to determine, resulting in unstable results. Deep learning methods also have problems such as difficulty in data labeling, low training efficiency, and high computational cost. Therefore, the existing methods may have high costs, low efficiency, insufficient accuracy, and high reliance on manual interaction, which makes it difficult to meet the needs of doctors for bone segmentation technology.

[0004] Therefore, improving the accuracy and stability of bone segmentation, improving segmentation efficiency and reducing the need for manual intervention have become one of the important topics in the field of medical image processing. The present invention is based on a three-dimensional broken bone segmentation method based on improved region growing. Unlike traditional methods and deep learning methods, it directly starts from the entire CT image and performs segmentation processing in combination with the relationship between the front and back images. It can improve the accuracy and stability of segmentation, and can generate multiple broken bones in parallel, greatly improving the processing efficiency. Especially in terms of the algorithm, the present invention optimizes region growing and automatically generates seed points without manual specification, which reduces the requirements for manual intervention and improves the autonomous intelligence of the algorithm. Therefore, the present invention has the characteristics of high accuracy, strong stability, high processing efficiency, and intelligence. It has high application value and promotion prospects, and can provide a more efficient and accurate processing method for bone segmentation work in multiple medical fields such as orthopedics and dentistry. Thereby better improving the doctor's diagnostic accuracy and treatment success rate, promoting the development of the field of medical image processing, and having important innovative and practical value. Summary of the invention:

[0005] The present invention aims to solve the shortcomings of previous bone segmentation tasks, such as low efficiency, low accuracy, high computational cost, and the need for manual intervention, and proposes a three-dimensional bone segmentation method based on improved region growing.

[0006] Specifically, the present invention provides a three-dimensional bone segmentation method based on improved region growing, comprising the following steps:

[0007] S1: Input a CT bone image path;

[0008] S2: Read this set of images and convert them into a three-dimensional array. Each value in the array represents the HU value of the pixel at that position.

[0009] S3: Perform threshold segmentation on the entire three-dimensional array according to a certain threshold, and convert the three-dimensional array after threshold segmentation into a binary three-dimensional array of either 0 or 1;

[0010] S31: For each position in the array, if the value is greater than or equal to the threshold, the value of the position in the array is assigned to 1;

[0011] S32: For each position in the array, if the value is less than the threshold, the value of the position in the array is assigned to 0;

[0012] S4: Traverse the newly generated three-dimensional array and perform a morphological corrosion operation using certain structural elements to remove the adhesions between the broken bones. After the operation, the value of the three-dimensional array is also a binary three-dimensional array of either 0 or 1.

[0013] S5: Obtain the preprocessed valid three-dimensional array through the three-dimensional array read in S2 and the newly generated three-dimensional array in S4;

[0014] S51: traverse the original three-dimensional array in S2. If the value of the corresponding position in the newly generated three-dimensional array in S4 is 1, the corresponding position in the effective three-dimensional array is retained as the value of the S2 position.

[0015] S52: traverse the original three-dimensional array in S2. If the value of the corresponding position in the newly generated three-dimensional array in S4 is 0, the value of the corresponding position in the valid three-dimensional array is set to a minimum value to indicate that the position does not contain a bone.

[0016] S6: performing a region growing operation on the valid three-dimensional array generated in S5, and finally growing multiple connected regions, that is, obtaining the corresponding regions of multiple broken bones;

[0017] S61: Color the array. That is, declare a color array colors with the same dimension and size as the three-dimensional array read in S2. Use the union-find algorithm on the binary array S4 to obtain multiple connected regions. All positions in each connected region are numbered 1, 2, 3, etc. in the color array colors.

[0018] S62: Add the corresponding position in the array S4 to the queue, and perform region growing in the six directions of the three-dimensional space at the same time according to a certain threshold rule. A certain seed point must have the same color number as the new point grown from it. If a growth position has been grown by other seed points (that is, it has been colored), it cannot be grown at this position;

[0019] S63: The newly grown points are added to the queue, and the S62 operation is repeated until the queue is empty;

[0020] S64: Finally, in the color array colors, positions with the same color number represent the same broken bone, and positions with different color numbers represent different broken bones;

[0021] S7: Obtain the pixel position of each broken bone and the broken bone number corresponding to each pixel position.

[0022] Preferably, the certain threshold in step S3 is 300HU.

[0023] Preferably, the size of the structuring element in step S4 is [2, 2].

[0024] Preferably, in step S52, the representative parameter of the minimum value is -10000.

[0025] Preferably, in step S62, the certain threshold rule is that the difference between adjacent elements is less than 50.

[0026] The present invention combines threshold segmentation and region growing methods, acts on a set of three-dimensional CT images, and is used to extract the bone region of the three-dimensional image. After bone recognition, different broken bone regions are distinguished, and finally multiple separated broken bones are identified, realizing a three-dimensional bone segmentation method with high efficiency and low error rate. First, the three-dimensional CT bone data is read in; secondly, an adaptive threshold segmentation method is used to separate the bone region from the background region to obtain all the bone regions; then, morphological operations are used to remove the adhesion between the broken bones to obtain the core region of each broken bone; finally, these core regions are used as seed points for region growth, and region growth is performed on multiple regions at the same time. In this way, while ensuring the efficiency of segmentation, the accuracy of bone segmentation is guaranteed, that is, each pixel point belongs to a unique region and is the region closest to the seed point. According to the above steps, the pixel points contained in all the broken bones in the three-dimensional bone image and the broken bone number to which each pixel point belongs can be determined. This method can be applied to three-dimensional bone segmentation to help complete disease diagnosis and treatment, lesion analysis, etc., help doctors better complete diagnosis and understand the patient's condition, and provide possibilities for computer-assisted preoperative planning, disease diagnosis and treatment.

[0027] In summary, the present invention has created a 3D bone segmentation method based on improved region growing, which has the following beneficial effects:

[0028] (1) Using CT images as a whole to segment broken bones can better consider the relationship between adjacent images, providing a new idea for bone segmentation. (2) The region growing method is improved. First, a relatively rough threshold segmentation method + morphological corrosion operation is used to obtain the core seed growth point, and then the growth rule of region growing is used to obtain each complete broken bone. This idea of ​​coarse first and then fine, from point to surface, avoids the awkward step of manually selecting growth points, while ensuring the accuracy and efficiency of the algorithm. This optimization greatly improves the scope of application and limitations of the region growing method. Description of the drawings:

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] Figure 1 It is a flow chart of the method of the present invention.

[0031] Figure 2 It is a flow chart of the method of the present invention for threshold segmentation.

[0032] Figure 3 It is a flow chart of the region growing algorithm of the method of the present invention. Specific implementation methods:

[0033] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0034] This example uses a CT bone image as an example. A 3D bone segmentation method based on improved region growing includes the following specific steps:

[0035] S1: Input a CT bone image path;

[0036] S2: Read this set of images and convert them into a three-dimensional array. Each value in the array represents the HU value of the pixel at that position.

[0037] S3: Perform threshold segmentation on the entire three-dimensional array according to a certain threshold, and convert the three-dimensional array after threshold segmentation into a binary three-dimensional array of either 0 or 1;

[0038] S31: For each position in the array, if the value is greater than or equal to the threshold, the value of the position in the array is assigned to 1;

[0039] S32: For each position in the array, if the value is less than the threshold, the value of the position in the array is assigned to 0;

[0040] S4: Traverse the newly generated three-dimensional array and perform a morphological corrosion operation using certain structural elements to remove the adhesions between the broken bones. After the operation, the value of the three-dimensional array is also a binary three-dimensional array of either 0 or 1.

[0041] S5: Obtain the preprocessed valid three-dimensional array through the three-dimensional array read in S2 and the newly generated three-dimensional array in S4;

[0042] S51: traverse the original three-dimensional array in S2. If the value of the corresponding position in the newly generated three-dimensional array in S4 is 1, the corresponding position in the effective three-dimensional array is retained as the value of the S2 position.

[0043] S52: traverse the original three-dimensional array in S2. If the value of the corresponding position in the newly generated three-dimensional array in S4 is 0, the value of the corresponding position in the valid three-dimensional array is set to a minimum value to indicate that the position does not contain a bone.

[0044] S6: performing a region growing operation on the valid three-dimensional array generated in S5, and finally growing multiple connected regions, that is, obtaining the corresponding regions of multiple broken bones;

[0045] S61: Color the array. That is, declare a color array colors with the same dimension and size as the three-dimensional array read in S2. Use the union-find algorithm on the binary array S4 to obtain multiple connected regions. All positions in each connected region are numbered 1, 2, 3, etc. in the color array colors.

[0046] S62: Add the corresponding position in the array S4 to the queue, and perform region growing in the six directions of the three-dimensional space at the same time according to a certain threshold rule. A certain seed point must have the same color number as the new point grown from it. If a growth position has been grown by other seed points (that is, it has been colored), it cannot be grown at this position;

[0047] S63: The newly grown points are added to the queue, and the S62 operation is repeated until the queue is empty;

[0048] S64: Finally, in the color array colors, positions with the same color number represent the same broken bone, and positions with different color numbers represent different broken bones;

[0049] S7: Obtain the pixel position of each broken bone and the broken bone number corresponding to each pixel position.

[0050] Figure 1 This is an overall flow chart of the broken bone segmentation method based on three-dimensional improved region growing provided by the present invention.

[0051] Figure 2 This is a flowchart for threshold segmentation in the overall flowchart of the broken bone segmentation method based on three-dimensional improved region growing provided by the present invention;

[0052] S3: Perform threshold segmentation on the entire three-dimensional array according to a certain threshold, and convert the three-dimensional array after threshold segmentation into a binary three-dimensional array of either 0 or 1;

[0053] S31: For each position in the array, if the value is greater than or equal to the threshold, the value of the position in the array is assigned to 1;

[0054] S32: For the value of each position in the array, if the value is less than the threshold, the value of the position in the array is assigned to 0.

[0055] Figure 3 This is a flowchart of the region growing algorithm in the overall flowchart of the broken bone segmentation method based on three-dimensional improved region growing provided by the present invention:

[0056] S6: performing a region growing operation on the valid three-dimensional array generated in S5, and finally growing multiple connected regions, that is, obtaining the corresponding regions of multiple broken bones;

[0057] S61: Color the array. That is, declare a color array colors with the same dimension and size as the three-dimensional array read in S2. Use the union-find algorithm on the binary array S4 to obtain multiple connected regions. All positions in each connected region are numbered 1, 2, 3, etc. in the color array colors.

[0058] S62: Add the corresponding position in the array S4 to the queue, and perform region growing in the six directions of the three-dimensional space at the same time according to a certain threshold rule. A certain seed point must have the same color number as the new point grown from it. If a growth position has been grown by other seed points (that is, it has been colored), it cannot be grown at this position;

[0059] S63: The newly grown points are added to the queue, and the S62 operation is repeated until the queue is empty;

[0060] S64: Finally, in the color array colors, positions with the same color number represent the same broken bone, and positions with different color numbers represent different broken bones.

[0061] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The scope of protection of the present invention should not be regarded as limited to the specific forms described in the embodiments. The scope of protection of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A three-dimensional improved region growing method for bone segmentation, characterized in that: The following steps are involved: S1: Input a 3D CT bone image path; S2: Read this set of images and convert them into a three-dimensional array. Each value in the array represents the HU value of the pixel at that position. S3: Perform threshold segmentation on the entire three-dimensional array according to the threshold, and convert the three-dimensional array after threshold segmentation into a binary three-dimensional array of either 0 or 1; S31: For each position in the array, if the value is greater than or equal to the threshold, the value of the position in the array is assigned to 1; S32: For each position in the array, if the value is less than the threshold, the value of the position in the array is assigned to 0; S4: traverse the newly generated three-dimensional array and perform a morphological corrosion operation on it using the structured element to remove the adhesion between the broken bones. After the operation, the value of the three-dimensional array is also a binary three-dimensional array of either 0 or 1. S5: Obtain a preprocessed valid three-dimensional array through the three-dimensional array read in step S2 and the three-dimensional array newly generated in step S4; S51: traverse the original three-dimensional array in step S2. If the value of the corresponding position in the newly generated three-dimensional array in step S4 is 1, the corresponding position in the effective three-dimensional array is retained as the value of the position in step S2. S52: traverse the original three-dimensional array in step S2. If the value of the corresponding position in the newly generated three-dimensional array in step S4 is 0, the value of the corresponding position in the valid three-dimensional array is set to a minimum value to indicate that the position does not contain a bone. S6: performing a region growing operation on the valid three-dimensional array generated in step S5, and finally growing multiple connected regions, that is, obtaining the corresponding regions of multiple broken bones; S61: Color the array, that is, declare a color array colors, with the same dimension and size as the three-dimensional array read in step S2, use the union-find algorithm on the binary array in step S4 to obtain multiple connected regions, and number all positions in each connected region in the color array colors in sequence as 1, 2, 3, etc.; S62: Add the corresponding position in the array in step S4 to the queue, and perform region growing in the six directions of the three-dimensional space simultaneously according to the threshold rule. A certain seed point must have the same color number as the new point grown from it. If a growth position has been grown by other seed points, that is, it has been colored, then growth cannot be performed at this position; S63: The newly grown points are added to the queue, and step S62 is repeated until the queue is empty. S64: Finally, in the color array colors, positions with the same color number represent the same broken bone, and positions with different color numbers represent different broken bones; S7: Obtain the pixel position of each broken bone and the broken bone number corresponding to each pixel position.

2. The 3D bone segmentation method based on improved region growing according to claim 1, characterized in that: The threshold in step S3 is 300 HU.

3. The 3D bone segmentation method based on improved region growing according to claim 1, characterized in that: The size of the structuring element in step S4 is [2, 2].

4. The 3D bone segmentation method based on improved region growing according to claim 1, characterized in that: In step S52, the representative parameter of the minimum value is -10000.

5. The 3D bone segmentation method based on improved region growing according to claim 1, characterized in that: In step S62, the threshold rule is that the difference between adjacent elements is less than 50.

Images