A method for establishing a skull anatomy database

Abstract

The application discloses a skull anatomical structure database establishment method, characterized in that the database establishment comprises the following steps: step one, reconstructing and printing normal human and case skull models; step two, reconstructing brain tissue, calculating corresponding brain volume, recording classification matching and archiving; step three, performing three-dimensional multi-planar reconstruction and surface reconstruction based on a bone threshold range on a case skull to-be-cut replacement or bone defect part; and step four, constituting a complete database containing normal human and case two types of skull anatomical structures. The skull anatomical structure database establishment method respectively establishes directories for storage classification, performs processing identification and matching of different samples, is convenient for artificial intelligence learning, performs three-dimensional multi-planar reconstruction and surface reconstruction, can cooperate with a convolutional neural network algorithm of artificial intelligence, extracts a patient part with the largest characteristic attribute to facilitate design of a subsequent convolutional layer and deep calculation, and makes recognition efficiency and effect better.

Description

Technical Field

[0001] This invention relates to the field of skull anatomical structure database technology, specifically a method for establishing a skull anatomical structure database. Background Technology

[0002] Skull defects caused by trauma, tumors, infections, and other factors are a common and challenging clinical problem. Skull repair can maintain the tightness of the skull and stabilize physiological intracranial pressure, thereby alleviating skull defect-related syndromes. However, significant individual differences among patients greatly increase the difficulty of repair. Currently, the conventional process for preparing a skull repair product mainly includes clinical data collection, manual 3D modeling, finished product fabrication and sterilization for surgical application. The entire cycle typically takes about two weeks, and for imported skull products, the preparation and delivery cycle can be as long as several months. This traditional, time-consuming process cannot meet the timely needs of clinical surgery.

[0003] In this context, the overall research focus has shifted to a comprehensive cranioplasty solution that integrates three key elements: 1. a hospital-enterprise integrated data transmission and management platform that links clinical imaging data, 3D reconstruction data processing, and production, enabling rapid response to individual patients' precise treatment; 2. an automated or semi-automated AI-powered 3D cranioplasty system based on a human skull database and incorporating AI convolutional neural network algorithms and 3D image reconstruction technology; and 3. a personalized cranioplasty product whose specific structural dimensions are determined based on 3D reconstruction model analysis.

[0004] This invention is based on a research plan for a personalized 3D-printed skull repair scheme that is completely different from the traditional approach of using artificial intelligence to analyze and transmit data. It provides a method for establishing a skull anatomical structure database during the construction of an artificial intelligence-based 3D modeling system. Summary of the Invention

[0005] The purpose of this invention is to provide a method for establishing a database of skull anatomical structures, laying the foundation for subsequent establishment of convolutional layers and optimization of a semi-automatic or fully automatic skull 3D model reconstruction system using convolutional neural network algorithms.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for establishing a skull anatomical structure database, characterized in that the database establishment includes the following steps:

[0007] Step 1: Reconstruct and print skull models of normal individuals and cases;

[0008] Step 2: Reconstruct brain tissue, calculate corresponding brain volume, and record, classify, match, and archive the data;

[0009] Wherein, the first database is established based on the normal skull model, and the second database is established based on the case skull model;

[0010] Meanwhile, the third database is established based on the selection of the closest brain volume data file in the first database in the second database.

[0011] Step three: based on the three-dimensional multi-planar reconstruction and surface reconstruction within the bone threshold range, the case skull to be cut or replaced or the bone defect part is particularly processed, the reconstruction file is combined with the case record storage to form the fourth database according to the specific part naming file group.

[0012] Step four: the first, second, third and fourth databases constitute a database containing the anatomical structure of the normal and case two types of skull.

[0013] Preferably, the model reconstruction in step one is: based on the CT and MRI imaging data of normal and case skulls, a plurality of skull digital models with different shapes and sizes are reconstructed, and 3D printing is performed to obtain a batch of normal skull models and a batch of case skull models.

[0014] Preferably, the database establishment in step two includes the following steps:

[0015] Step 1: image acquisition, preprocessing, feature extraction of normal sample, establishment of first database;

[0016] Step 2: image acquisition, preprocessing, feature extraction of case sample, establishment of second database;

[0017] Step 3: matching and identification of normal sample and case sample, establishment of third database.

[0018] Preferably, the establishment of the first database in step 1 is: reconstructing a brain tissue model suitable for the normal skull model, calculating the corresponding brain volume, recording and archiving the single or multiple normal skull models suitable for different brain volumes, marking the skull points and recording the anatomical three planes of all cases by artificial or AI, and establishing the first database.

[0019] Preferably, the establishment of the second database in step 2 is: based on digital simulation repair, the case skull to be cut or replaced or the bone defect part is repaired and designed, which is converted into a processed three-dimensional reconstruction model, and recorded and archived, and the skull points and anatomical three planes of all cases are marked by artificial or AI, and the second database is established.

[0020] Preferably, the third database in step 3 is established by reconstructing brain tissue matched with the skull model of the case, and searching the normal skull model group corresponding to the calculated brain volume of the normal skull model according to the obtained brain volume, so as to select the normal skull model with the best matching with the skull of the case for classification and archiving, and establish the third database.

[0021] Preferably, the skull point markers include, but are not limited to, the X-ray cephalometric marker points of the nasion point, the precomb point, the sella point, the posterior margin point of the sphenoid wing, the ear point, and the skull base point.

[0022] Preferably, the anatomical three surfaces are the axial surface (transverse surface), coronal surface, and sagittal surface of the skull.

[0023] Preferably, the matching method between the skull model of the case and the normal skull model in step 3 is as follows:

[0024] S1: Preliminary screening according to the gender, age, and head shape of the patient of the skull model of the case;

[0025] S2: Comparing the positioning coincidence degree of the marked skull points of the three-dimensional reconstruction model of the skull model of the case and the three-dimensional reconstruction model of the normal skull model in the three-dimensional reconstruction space coordinates, and comparing the superposition difference size of the recorded anatomical three surface images of the three-dimensional reconstruction model of the skull model of the case and the three-dimensional reconstruction model of the normal skull model;

[0026] The one with the minimum difference and the highest coincidence degree is the most suitable model.

[0027] Compared with the prior art, the skull anatomical structure database establishment method has the advantages that the directories are established and stored separately, which is convenient for artificial intelligence learning, facilitates the processing, recognition, and matching of different samples, and performs three-dimensional multi-planar reconstruction and surface reconstruction, can cooperate with the convolutional neural network algorithm of artificial intelligence to extract the patient part with the largest characteristic attribute for subsequent deepening calculation of the convolutional layer, so that the recognition efficiency and effect are better. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 It is a system architecture diagram of the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0030] Referring to Figure 1 The present application provides a technical solution: a skull anatomical structure database establishment method, the database establishment comprising the following steps:

[0031] Step 1: Reconstruct and print normal and case skull models; based on CT and MRI and other imaging data of normal and case skulls, reconstruct a plurality of skull digital models of different shapes and sizes, and perform 3D printing to obtain a batch of normal skull models and a batch of case skull models.

[0032] Step 2: Reconstruct brain tissue, calculate corresponding brain volume, and record classification matching and archiving;

[0033] Among them, the normal sample image collection, preprocessing, feature extraction, the establishment of the first database, the establishment of the first database is: reconstructing the brain tissue model matched with the normal skull model, calculating the corresponding brain volume, recording and archiving the single or multiple normal skull model groups matched under different brain volumes, marking the skull points and recording the anatomical three-plane of all cases by artificial or AI, and establishing the first database;

[0034] The second database is established by collecting, preprocessing and extracting features of the case sample images, and the establishment of the second database is: based on digital simulation repair, the repair guide is designed to fill in the case skull to be cut and replaced or the bone defect part, which is converted into a processed three-dimensional reconstruction model and recorded and archived, and the skull points and anatomical three-plane of all cases are marked by artificial or AI, and the second database is established;

[0035] At the same time, the matching and identification of normal samples and case samples are based on the selection of the closest brain volume data file in the first database in the second database, and the third database is jointly established, and the establishment of the third database is: reconstructing the brain tissue matched with the case skull model, and selecting the normal skull model with the best matching with the case skull from the normal skull model group corresponding to the brain volume calculated according to the brain volume obtained by reverse searching the normal skull model, so as to classify and archive the normal skull model with the best matching with the case skull, and establish the third database;

[0036] Among them, the skull point marking includes but is not limited to marking the X-ray cephalometric landmark points such as the nasion point, the precomb point, the sella turcica point, the posterior margin point of the sphenoid wing, the ear point, and the skull base point, and the anatomical three planes are the cross-sectional images of the skull axial plane (transverse plane), coronal plane and sagittal plane.

[0037] At the same time, the matching method between the case skull model and the normal skull model is:

[0038] S1: According to the gender, age and head shape of the case skull model patient, preliminary screening is performed;

[0039] S2: The positioning coincidence degree of the marked skull points of the three-dimensional reconstruction model of the comparative case skull model and the marked skull points of the three-dimensional reconstruction model of the normal skull model in the three-dimensional reconstruction space coordinates, and the difference size of the recorded anatomic three-plane images of the three-dimensional reconstruction model of the comparative case skull model and the three-dimensional reconstruction model of the normal skull model;

[0040] The one with the highest minimum coincidence degree is the most suitable model

[0041] Step three: Based on the three-dimensional multi-planar reconstruction and surface reconstruction within the bone threshold range, the case skull to be cut or replaced or the bone defect part is specially recorded, and the reconstruction file is combined with the case to form a fourth database;

[0042] Step four: The first, second, third and fourth databases form a database containing the anatomic structures of normal and case skulls, which lays a foundation for the implementation of the convolutional neural network algorithm optimization to build a semi-automatic or fully-automatic skull three-dimensional model reconstruction system.

[0043] Embodiment:

[0044] Firstly, based on the CT and MRI imaging data of normal and case skulls, a number of skull digital models of different shapes and sizes are reconstructed, and 3D printing is performed to obtain a batch of normal skull models: class A, a batch of case skull models: class B; three-dimensional reconstruction files are used for subsequent engineering processing and provide data file basis for the construction of subsequent three-dimensional modeling system;

[0045] Reconstruct the brain tissue suitable for class A and calculate the corresponding brain volume. Record and archive the single or multiple normal skull models suitable for different brain volumes, such as three examples A1, A2 and A3 under the volume of α, recorded as α-A1, A2 and A3. Mark the skull points and record the anatomic three-plane of all examples by artificial or AI to establish the first database. Collect, preprocess and extract the features of normal sample images to facilitate artificial intelligence learning;

[0046] Based on digital simulation repair, the case skull to be cut or replaced or the bone defect part is repaired and designed into a processed three-dimensional reconstruction model (B1→B1*), and is recorded and archived (such as B1-B1*). Mark the skull points and record the anatomic three-plane of all processed examples by artificial or AI to establish the second database. Collect, preprocess and extract the features of case sample images to facilitate artificial intelligence learning;

[0047] Reconstruct the brain tissue matched with B, and according to the obtained brain volume, inversely search the normal skull model group corresponding to the brain volume calculated by A, so as to select the normal skull model best matched with the skull of the case to classify and archive, such as B1, the corresponding brain volume calculated is a, the normal skull example A1 best matched with the case is selected from A1, A2 and A3, and classified as B1-a-A1, a third database is established, and the matching and identification of the normal person and the case sample are matched, so as to facilitate the learning of artificial intelligence;

[0048] Among them, the case newly added after the initial data is classified as A or B after the first step of reconstruction processing, if it is a normal skull (A), the calculation group is added to the first database, and if it is a case skull (B), the calculation group is added to the second or third database

[0049] Meanwhile, taking B1 matched to A1 as an example, specifically, the method of selecting the normal skull model best matched with the skull of the case is: first, screening according to the gender, age and basic shape of the head of the B1 patient; second, comparing the positioning coincidence degree of the skull points marked by B1* and the skull points marked by A1 in the three-dimensional reconstruction space coordinates, and comparing the superposition difference size of the anatomical three-plane images recorded by B1* and the plane images of A1. The smallest difference and the highest coincidence degree are the best match;

[0050] The case skull (B, not processed, not B*) is specially subjected to three-dimensional multi-planar reconstruction and surface reconstruction based on the bone threshold range, and the specific part is named as a file group, such as “coronal suture”, the reconstruction file is combined with the case record to store and form a fourth database, such as B1-coronal suture; the reason for specially performing three-dimensional multi-planar reconstruction and surface reconstruction on the patient part is that the convolutional neural network algorithm used in the artificial intelligence of the patent extracts the patient part with the largest characteristic attribute, which facilitates the design of subsequent convolutional layers and deepens the calculation, so that the recognition efficiency and effect are better;

[0051] The first, second, third and fourth databases constitute a complete database containing the anatomical structure of the skull of the normal person and the case, which lays a foundation for the implementation of the convolutional neural network algorithm optimization to construct a semi-automatic or fully-automatic skull three-dimensional model reconstruction system.

[0052] Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to part of the technical features, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of establishing a skull anatomy database, characterized by, The database establishment comprises the following steps: Step 1: reconstruct and print normal human and case skull models; Step 2: reconstruct brain tissue, calculate corresponding brain volume, and record classification matching and archiving; Among them, the first database is established based on the normal human skull model, and the second database is established based on the case skull model; At the same time, the first database closest to the brain volume data file is selected from the second database to jointly establish the third database; The establishment of the third database is: reconstructing the brain tissue matched with the case skull model, and according to the obtained brain volume, the normal human skull model group corresponding to the brain volume calculated by the normal human skull model is reversely searched, so that the normal skull model with the best matching with the case skull is selected for classification and archiving, and the third database is established Step 3: Based on the three-dimensional multi-planar reconstruction and surface reconstruction within the bone threshold range, the case skull to be cut or replaced or the bone defect part is particularly processed, and the reconstructed file is combined with the case to form the fourth database; Step 4: The first, second, third and fourth databases constitute a database containing the anatomical structures of normal human and case skulls.

2. The method of claim 1, wherein: In step 1, the model reconstruction is based on CT and MRI imaging data of normal human and case skulls to reconstruct a plurality of skull digital models with different shapes and sizes, and 3D printing is performed to obtain a batch of normal human skull models and a batch of case skull models.

3. The method of claim 1, wherein: The establishment of the database in step 2 comprises the following steps: Step 1: image acquisition, preprocessing, and feature extraction of normal human samples to establish a first database; Step 2: image acquisition, preprocessing, and feature extraction of case samples to establish a second database; Step 3: matching and identification of normal human samples and case samples to establish a third database.

4. The method of claim 3, wherein: In step 1, the establishment of the first database is: reconstructing a brain tissue model matched with the normal human skull model, calculating the corresponding brain volume, recording and archiving the single or multiple normal human skull model groups matched under different brain volumes, marking the skull points and recording the anatomical three-plane of all cases by manual or AI, and establishing the first database.

5. The method of claim 3, wherein: In step 2, the establishment of the second database is: based on digital simulation repair, the case skull to be cut or replaced or the bone defect part is repaired and designed, and the processed three-dimensional reconstruction model is converted and recorded, and the skull points and anatomical three-plane of all cases are marked by manual or AI, and the second database is established.

6. A method of establishing a skull anatomy database according to any one of claims 4-5, characterized in that: The skull point marking includes but is not limited to marking the X-ray cephalometric landmark points such as the nasion point, the precomb point, the sella turcica point, the posterior edge point of the sphenoid wing, the ear point, and the skull base point.

7. A method of establishing a skull anatomy database according to any one of claims 4-5, characterized in that: The anatomical three-plane is the cross-sectional recording of the axial plane (transverse plane), coronal plane, and sagittal plane of the skull.

8. The method of claim 3, wherein: In step 3, the matching method between the case skull model and the normal skull model is: S1: According to the gender, age and head shape of the case skull model patient, preliminary screening is performed; S2: The positioning coincidence degree of the marked skull points of the three-dimensional reconstruction model of the comparative case skull model and the three-dimensional reconstruction model of the normal skull model in the three-dimensional reconstruction space coordinates, and the size of the difference between the recorded anatomical three-plane images of the three-dimensional reconstruction model of the comparative case skull model and the plane images of the three-dimensional reconstruction model of the normal skull model; Wherein the most suitable model is the one with the minimum difference and the highest coincidence degree.

Images