Orthopedic 3D printing method and device based on AI

A 3D printing and orthopedic technology, applied in the field of 3D printing, can solve the problems of inability to build a 3D model of damaged human tissue, lack of it, etc.

Active Publication Date: 2022-04-01

西安博恩生物科技有限公司

18 Cites 0 Cited by

AI-Extracted Technical Summary

Problems solved by technology

However, when 3D printing technology is used to produce and manufacture artificial prostheses, there is a lack of a method and device that can perform 3D reconstruction according to the actual situation of bone defects, and it is impossible to quickly establish a 3D model...

Abstract

The invention relates to the technical field of 3D printing, in particular to an orthopaedic 3D printing method and device based on AI. The method comprises the following steps: acquiring tomography data of human skeleton tissues, and constructing a human skeleton virtual model; based on collected tomography data and diagnosis data, marking a bone defect part area in the human body bone virtual model; traversing a skeleton model database, and matching a three-dimensional model corresponding to the skeleton defect partial region; mapping the marked bone defect partial region to a corresponding three-dimensional model according to the diagnosis data to obtain a fitted bone defect model; and the obtained bone defect model is sliced and layered according to the vector direction of 3D printing, and a three-dimensional model of each slicing and layering angle is obtained for layered printing. According to the method, the tomography data are mapped into the created human skeleton virtual model, the damaged skeleton is matched with the three-dimensional model, and slicing and layered printing are carried out according to the skeleton defect model.

Application Domain

Additive manufacturing apparatusManufacturing data aquisition/processing

Technology Topic

TomographyNuclear medicine +13

Image

Examples

Experimental program(2)

Example Embodiment

[0071] Example 1

[0072] See figure 1 Distance figure 1 A flowchart of a AI-based orthopedic 3D printing method provided herein. One embodiment of the present application provides an AI-based orthopedic 3D printing method, including the following steps:

[0073] S1: Get the fault scan data of the human bone tissue and construct the human bone virtual model.

[0074] See figure 2 As shown, the acquisition method of the fault scan data is:

[0075] S101, for damaged some human bones that need to build a human prosthetic limb, electronic computer tomography;

[0076] S102, full-scale scanning around the same part of the human bone, obtaining the human bone stereo fault scan data;

[0077] S103, based on the human bone stereo tomography data, establish a virtual model with the proportion of the human bone.

[0078] Before constructing the human skeletal virtual model, the human bones are scanned by the X-ray bundle, gamma ray, ultrasonic waves, etc., with this screen as the basis for constructing the human bone virtual model, and excludes the interference of the non-bone structure.

[0079] Specifically, when scanning the image data is processed, the LBP image feature extraction algorithm is used to extract the skeletal profile in the tomographic scanning image data, and the human soft tissue characteristics are created, and the human skeleton virtual model containing the skeleton profile and the human soft tissue characteristics is created.

[0080] In this embodiment, see image 3 As shown, based on the LBP image feature extraction algorithm, the bone profile in the image data of the fault scan, the human soft tissue characteristics include:

[0081] S1021, acquire image data of the electronic computer fault scan;

[0082] S1022, based on the LBP image feature extraction algorithm, the bone profile in the image data of the fault scan, the human soft tissue characteristics are identified;

[0083] S1023, according to the skeleton profile, human soft tissue characteristics symmetrical comparison, differential feature data in image data;

[0084] S1024, based on differential feature data, the bone defect portion area is pre-taught in image data.

[0085] Specifically, when the LBP feature is extracted, the grayscale image of the scan window is divided into a small area of 16 × 16. For a central pixel point for one pixel in each small area, 8 pixels adjacent to the surroundings. Point comparison, and the pixel value of the center pixel point is contrast with the peripheral pixel value. When the pixel value of the center pixel point is greater than the pixel value of the surrounding pixel point, the center pixel point is marked as 1, otherwise the label is 0. This, 8 points in the 3 × 3 field were compared to 8-bit binary, gave the LBP value of each central pixel point. Then, according to the obtained LBP value, calculate the histogram of the frequency of 1,0 in each small area, and normalize the histogram, and finally connect the histogram of each small area to a feature. Vector, to obtain an LBP texture feature vector of the overall graph of the fault to scanned image data without the damaged bones, and then extract the skeletal profile of the tomographic scanned image data, the human soft tissue characteristics.

[0086] In particular, the constructed damaged section of the human bone virtual model is constructed of a three-dimensional virtual model based on CT scanning.

[0087] S2: Based on the acquisition of tomography data and diagnostic data, the bone defect portion is labeled in the human bone virtual model.

[0088] It is necessary to specify that see Figure 4 As shown, the LBP image feature extraction algorithm, the method of identifying the skeletal profile in the image data of the fault scan, including:

[0089] S201, a grayscale image of a full-scale scanned human bone stereo tomogram of the obtained human bone, obtains a grayscale image of the scan window;

[0090] S202, divide the grayscale image of the scan window into several small regions, compare the pixel values of each center pixel point in the small area to the pixel value of the adjacent pixel point, to obtain the LBP value of each center pixel point;

[0091] S203, calculates the histogram of each small area according to the obtained LBP value, and normalizes the histogram;

[0092] S204, a histogram of each small area is connected to a feature vector, forming an LBP texture feature vector of the entire fault scan image data to obtain a skeletal profile of the fault scanned image data, and human soft tissue characteristics.

[0093] In particular, it is, the bone video data of the fault scanned image data, is surrounded by the damaged segment by CT. The video input bone recognition model is taken to identify the bone characteristics of the damaged bones to quickly match the corresponding bone three-dimensional model from the skeleton model database. Among them, the skeleton model database includes three-dimensional models of different types of bones in advance, such as leg bone models, fingertone model, spinal model, hip model, hip joint model, knee model, etc.

[0094] In this embodiment, see Figure 5 As shown, a method of matching a three-dimensional model corresponding to the skeleton defect portion area, including:

[0095] S211, the human bone virtual model of the marker bone defect portion is acquired, and the human bone virtual model is used to rotate sampling processing to obtain continuous picture frame data;

[0096] S212, input the picture frame data into the pre-trained bone recognition model, output bone characteristics of the human bone virtual model of the part of the bone defect;

[0097] S213, traversing the skeleton model database, query the matching bone three-dimensional model according to the bone characteristics of the bones.

[0098] Among them, the training method of the skeleton recognition model includes:

[0099] Get training sample set, the training sample set includes the bone characteristics of the collected skeletal image and the skeleton image;

[0100] The obtained training sample is entered into the bone recognition model to be trained for training, and the bone characteristics corresponding to the skeletal image are output. When training is completed, the bone recognition model is obtained.

[0101] Among them, the method of bone characteristics corresponding to the skeletal image is:

[0102] Get collected skeleton images;

[0103] Bone characteristics in the bone image based on HAAR image feature extraction algorithm;

[0104] Among them, the HAAR image feature extraction algorithm identifies the bone characteristics in the bone image, including:

[0105] Gray the obtained bone image;

[0106] The Gamma Correction Law is used to standardize the input skeletal image, and calculate the gradient of each pixel value of the bone image;

[0107] The bone image after calculating the gradient is divided into several cells, calculate the gradient histogram of each cell;

[0108] The cell combination forms a communication section, and the interval is normalized, and all overlapping interulations in the bone image are characterized to obtain bone characteristics in the bone image.

[0109]Specifically, when the bone characteristics in the bone image are identified by the HAAR image feature extraction algorithm, the directional gradient histogram characteristics are divided into small communication regions, and the gradient of each pixel point in the connecting area or the edge of the edge is collected. The figure, finally combines these histograms to constitute a bone feature descriptor.

[0110] When the GAMMA correction method is standardized for color space of the input bone image, the contrast of the bone image can be adjusted, reducing the impact of the shadow and lighting changes of the bone image, and can suppress noise interference, calculate the gradient of each cell. The histogram can capture outline information while further weakening the interference of the light.

[0111] S3: Traverse the skeleton model database, match the three-dimensional model corresponding to the partial area of the skeleton defect.

[0112] In order to distinguish the damaged bones in the image data of the bones, the bone marking box is set in the skeleton image data, where, see Image 6 As shown, the method of labeling the bone label frame is:

[0113] S301, obtain images that contain bone information;

[0114] S302, identify the edge feature points of each bone in the image, and fit the edge feature points of each bone to obtain an edge feature frame of each bone;

[0115] S303, based on the edge characteristic box to obtain a bone annotation box in the collected image.

[0116] Such as image 3 As shown, the damaged skeleton image is acquired by CT, and the damaged bones of the bone will be collected by CT, and the scattering point is collected, the scattering point is used as the edge feature point of each bone in the image, and the edge characteristic frame of fitting the bone is formed. Using the edge characteristic frame to obtain a bone annotation box.

[0117] At the same time, it is also possible to determine the center point of the damaged bone according to several retroactive points, the center point of the damaged bone is the bone geometry of the scanned obtained.

[0118] S4, in accordance with the diagnostic data, the diagnostic data is mapped to the corresponding three-dimensional model to obtain a fitted bone defect model.

[0119] S5, the obtained bone defect model is slid in a slice in the vector direction of 3D, resulting in a three-dimensional model of each slice hierarchy angle for a layered printing.

[0120] When printing, the skeleton defect model is printed in accordance with the vector direction of the print, for example, from bottom to, after completing the printing of a slice, continue the printing of the layers until the entire print operation is completed.

[0121] It should be understood that although described above is described in a certain order, these steps are not necessarily executed in the above order. Unless otherwise specified herein, the implementation of these steps does not have a strict order, which can be performed in other order. Moreover, a portion of the steps of the present embodiment may include a plurality of steps or more stages, and these steps or phases are not necessarily performed at the same time, but can be performed at different times, these steps or phases are not executed. It is inevitably performed in turn, but can be performed or alternately performed with at least a portion of the steps or stages in other steps or other steps.

Example Embodiment

[0122] Example 2

[0123] See Figure 7 As shown, one embodiment of the present application provides an AI-based orthopedic 3D printing apparatus, including a bone virtual model build module 100, a defect area marking module 200, a three-dimensional model matching module 300, a bone defect model fitting module 400, and 3D Print Module 500. in:

[0124] The skeletal virtual model build module 100 for constructing a human bone virtual model based on the acquired tomographic scan image data.

[0125] When the human bone virtual model is constructed, by obtaining the image data of the electronic computer tomography, based on the LBP image feature extraction algorithm, the bone profile of the image data of the fault scan, the human soft tissue characteristics, according to the skeleton profile, human soft tissue characteristics Comparison, differential feature data in image data is obtained.

[0126] The damaged section of the damaged section of the damaged section is based on the three-dimensional virtual model based on CT scan, and is constructed for equal proportions.

[0127] The defect area marking module 200 is used to mark the bone defect portion area in the human bone virtual model based on the acquired fault scan data and diagnostic data.

[0128] When matching the three-dimensional model corresponding to the skeleton defect portion, the human bone virtual model of the marker bone defect portion is acquired, and the human bone virtual model is rotatable to obtain continuous picture frame data; input picture frame data The pre-trained bone identification model outputs the bone characteristics of the human bone virtual model of the part of the bone defect; traversing the bone model database, and query the matching bone three-dimensional model according to the bone characteristics of the bone.

[0129] The three-dimensional model matching module 300 is used to traverse the skeleton model database, match the three-dimensional model corresponding to the partial region of the bone defect.

[0130] In the present embodiment, the three-dimensional model matching module 300 maps the obtained damaged bone on the actual position of the damaged portion end surface data to the human bone virtual model, and maps the matching three-dimensional model to the human bone virtual model 3D Print the structure of the bone damaged portion.

[0131] Among them, in order to distinguish the damaged bones in the image data of the bones, the bone marker box is provided in the skeletal image data, and the edge feature points of each bone in the image are identified, and the edge feature points of each bone are fitted. Get the edge feature frame of each bone, based on the edge characteristic frame, to obtain a bone label box in the collected image.

[0132] The skeleton defect model fitting module 400; for maping the skeleton defect portion of the labeled data to the corresponding three-dimensional model, the diagnostic data is mapped to the corresponding three-dimensional model.

[0133] In the present embodiment, the process of generating a moving trajectory image corresponding to each of the damaged bones is to match the three-dimensional model corresponding to the bone defect portion region according to the image data acquired; the image data of the collected bone is processed, obtained Damaged bone skeletal marking box; three-dimensional model and bone label sign corresponding to the damaged bones, and calculate the center of damage to the damaged bones; according to the actual position of the damaged bones according to the bone position CT data, and calculate the damaged bone Central point coordinates; map the central point coordinates of continuous picture frame data to any picture frame data image to generate a moving trajectory image corresponding to each damaged bone.

[0134] The 3D print module 500 is configured to slice the obtained skeleton defect model in a vector direction of 3D, to obtain a three-dimensional model of each slice hierarchy, in this embodiment, based on Ai's orthopedics The 3D printing device is performed using an AI-based orthopedic 3D printing method based on the foregoing embodiment. Therefore, in this embodiment, the operation of the AI-based orthopedic 3D printing device is not described in detail.

[0135] In summary, the AI-based orthopedic 3D printing method and apparatus is provided, which fully utilizes the fault scan data, creating the human bone virtual model consistent with the damaged section, and recognizes the behavior of the bone model database. Query the matching three-dimensional model, mapping the fault scan data into the created human bone virtual model, matches the damaged bones, and prints the slice layered printing according to the bone defect model, and can establish a three-dimensional model according to human damaged tissue. 3D print provides accurate data and provides precise positioning for 3D printing operations.

PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.