Construction method and application of multi-nose structured three-dimensional face template data set

Abstract

The invention relates to a construction method and application of a multi-nose-form structured three-dimensional face template data set. The construction method comprises the following steps: (1) acquiring and processing three-dimensional face data of different nose forms; and (2) establishing a three-dimensional face template data set. The application comprises the following steps of: (1) collecting three-dimensional face data of a patient with nasal defect, and selecting a corresponding three-dimensional face template by referring to the nose form of the patient before the defect or the nose form wanted by the patient; (2) performing deformation registration on the three-dimensional face template to the three-dimensional face data of the patient by using a non-rigid registration algorithm to obtain data after deformation of the face template; and (3) obtaining nasal defect prosthetic morphological data based on the deformed data of the face template. According to the method, a structured three-dimensional face template data set with different nose forms can be constructed based on average data of three-dimensional face data of different nose forms of Chinese people; the three-dimensional face template data set further improves the universality and clinical suitability of constructing the nose prosthesis form by the template method.

Description

Technical Field

[0001] The present invention relates to a structured three-dimensional face template dataset with different nose morphologies, and particularly to a construction method and application of a structured three-dimensional face template dataset with multiple nose morphologies. Background Art

[0002] Facial defects caused by congenital defects, trauma, tumor resection and other reasons will have serious adverse effects on the patient's physiology, psychology and personal life. How to restore the integrity of the three-dimensional facial morphology for facial defect patients is a difficult point faced by the current oral aesthetic restoration field. Facial defects can be divided into the defects of each organ such as the ear, nose, eye and orbit and the defects of other facial tissues. For facial defect patients, facial prostheses are one of the important treatment methods to restore the facial integrity and aesthetics of patients. The traditional prosthesis manufacturing process is complex, involving more than a dozen steps such as impression taking, wax pattern making and prosthesis processing. Among them, wax pattern making particularly emphasizes the simulation effect of the prosthesis, requiring the prosthodontist and technician to not only have sufficient medical knowledge and experience, but also have high aesthetic qualities and sculpture skills, making the production of traditional prostheses highly dependent on experience, with a long production cycle and a high production failure rate, and it is difficult to be popularized and applied at the grass-roots level. With the development of digital technology, technologies such as computer-aided design (CAD) and computer-aided manufacturing (CAM) have been widely used in the design and manufacture of facial prostheses, improving the efficiency of facial prosthetic treatment and the comfort of patients.

[0003] 1. Research Development Trends and Current Situations at Home and Abroad

[0004] Currently, the digital design methods for facial prostheses are mainly divided into the "mirror method" and the "database method". For unilateral defects of paired organs such as ears and eyes, based on the natural symmetry of the face, an ideal prosthetic morphology can be obtained by mirroring the morphological features of the healthy side tissues. The nose is located in the center of the face and is not a paired facial organ. For nasal defects that cross the midline of the face, the nasal model database is an important basis for digitally designing the nasal prosthetic morphology. In previous studies, some researchers obtained three-dimensional nasal data of different morphologies through methods such as three-dimensional facial scanning or random generation of three-dimensional deformable human face models, and then created a nasal model database, and classified the data based on special three-dimensional morphologies, key feature dimensions, etc. For different nasal defect patients, by screening out the nasal data suitable for their facial morphological features in the database and combining certain data processing and adjustment, the digital design of the nasal prosthetic morphology can be completed. Compared with the traditional design method of nasal prostheses, the "database method" has the advantages of low experience dependence, high efficiency, strong interactivity, and good visualization effect. However, the construction of the nasal model database is affected by various factors such as ethics and cost. Currently, there are almost no publicly available databases, and there are significant morphological differences in the noses of different ethnic groups, resulting in limited generality of the database. In addition, the screening, processing, and adjustment of nasal data require certain manual participation, resulting in a low degree of automation of the "database method". The above deficiencies limit the application and popularization of the "database method" in oral clinical practice.

[0005] To improve the degree of automation of the "database method", some researchers proposed a three-dimensional morphological completion method for nasal defect prostheses (hereinafter referred to as the "template method") based on a structured three-dimensional human face template (general template) with the facial morphological features of an adult Chinese male and the open-source non-rigid registration program MeshMonk. The "template method" is based on the non-defective area of the three-dimensional facial data of nasal defects. Using the MeshMonk program, the three-dimensional human face template is deformed and registered to the nasal defect data. The spatial positions between the deformed three-dimensional human face template and the nasal defect data can be unified, and the complete three-dimensional facial morphology can be maintained, so as to obtain the nasal prosthetic morphological data within the projection range of the nasal defect boundary. The "template method" has a high degree of automation, as well as good feasibility and clinical applicability. However, the constructed nasal prosthetic morphology inherits to a certain extent from the nasal morphology of the three-dimensional human face template, resulting in a relatively single morphology and being unable to fully meet the prosthetic needs of nasal defect patients.

[0006] 2. Review and summary

[0007] As can be seen from the above review of previous literature, the digital design of nasal prostheses is of great significance for the prosthetic treatment of patients with nasal defects. However, the current commonly used "database method" has limited generality and low automation; the "template method" proposed in previous studies has improved the automation of constructing prosthetic forms, but the prosthetic forms constructed based on a single general template are relatively monotonous. Therefore, in response to the need to construct nasal prosthetic forms using the "template method", it is very necessary to establish a three-dimensional human face template dataset with different nasal forms.

[0008] 3. References

[0009] 1. Gaudin R, Raguse JD, Krause S, Mumm J, Motzkus Y, Ghanad I, et al. Quality of life and psychological evaluation of patients after anaplastology. J CranioMaxill Surg 2023;51:485-9.

[0010] 2. Dholam KP, Parkar SP, Dugad JA, Kharade PP, Shinde AA, Gurav SV. Development of a psychosocial perception scale and comparison of psychosocial perception of patients with extraoral defects before and after facial prosthesis. J Prosthet Dent 2022;128:1398-404.

[0011] 3. Rosen EB, Ahmed ZU, Huryn JM, Ganly I. Prosthetic rehabilitation of the geriatric oncologic rhinectomy patient utilizing a craniofacial implant-retained nasal prosthesis. Clin Case Rep 2020;8:278-82.

[0012] 4. Phasuk K, Haug SP. Maxillofacial Prosthetics. Oral Maxillofac Surg Clin North Am 2018; 30: 487 - 97.

[0013] 5. Nuseir A, Hatamleh MM, Alnazzawi A, Al - Rabab'Ah M, Kamel B, Jaradat E. Direct 3D Printing of Flexible Nasal Prosthesis: Optimized Digital Workflow from Scan to Fit. J Prosthodont 2019; 28: 10 - 4.

[0014] 6. Unkovskiy A, Spintzyk S, Brom J, Huettig F, Keutel C. Direct 3D printing of silicone facial prostheses: A preliminary experience in digital workflow. J Prosthet Dent 2018; 120: 303 - 8.

[0015] 7. Bi Y, Wei H. Fully digital workflow for the rehabilitation of a total nasal defect with a prosthesis. J Prosthet Dent 2024; 131: 167 - 71.

[0016] 8. Farook TH, Jamayet NB, Abdullah JY, Rajion ZA, Alam MK. A systematic review of the computerized tools and digital techniques applied to fabricate nasal, auricular, orbital and ocular prostheses for facial defect rehabilitation. J Stomatol Oral Maxillofac Surg 2020; 121: 268 - 77.

[0017] 9. Farook TH, Jamayet NB, Abdullah JY, Asif JA, Rajion ZA, Alam MK. Designing 3D prosthetic templates for maxillofacial defect rehabilitation: A comparative analysis of different virtual workflows. Comput Biol Med 2020;118:103646.

[0018] 10. You P, Liu YC, Silva RC. Fabrication of 3D Models for Microtia Reconstruction Using Smartphone-Based Technology. Annals of Otology, Rhinology & Laryngology 2021:1049759264.

[0019] 11. Bockey S, Berssenbrügge P, Dirksen D, Wermker K, Klein M, Runte C. Computer-aided design of facial prostheses by means of 3D-data acquisition and following symmetry analysis. J Cranio Maxill Surg 2018;46:1320 - 8.

[0020] 12. Sun MH, Yen CH, Tsai YJ, Liao YL, Wu SY. Fabrication of a facial prosthesis for a 13-year-old child by using a point-and-shoot three-dimensional scanner and CAD / CAM technology. Taiwan J Ophthalmol 2022;12:219 - 22.

[0021] 13. Ballo AM, Nguyen CT, Lee V. Digital Workflow of Auricular Rehabilitation: A Technical Report Using an Intraoral Scanner. J Prosthodont 2019; 28: 596 - 600.

[0022] 14. Unkovskiy A, Roehler A, Huettig F, Geis - Gerstorfer J, Brom J, Keutel C, et al. Simplifying the digital workflow of facial prostheses manufacturing using a three - dimensional (3D) database: setup, development, and aspects of virtual data validation for reproduction. J Prosthodont Res 2019; 63: 313 - 20.

[0023] 15. Palousek D, Rosicky J, Koutny D. Use of digital technologies for nasal prosthesis manufacturing. Prosthet Orthot Int 2014; 38: 171 - 5.

[0024] 16. Reitemeier B, Gotzel B, Schone C, Stockmann F, Muller R, Lexmann J, et al. Creation and utilization of a digital database for nasal prosthesis models. Onkologie 2013; 36: 7 - 11.

[0025] 17. Fantini M, De Crescenzio F, Ciocca L. Design and Rapid Manufacturing of anatomical prosthesis for facial rehabilitation. International journal on interactive design and manufacturing 2013; 7: 51 - 62.

[0026] 18. Jablonski RY, Malhotra T, Coward TJ, Shaw D, Bojke C, Pavitt SH, et al. Digital database for nasal prosthesis design with a 3D morphable face model approach. The Journal of prosthetic dentistry 2024; 131: 1271 - 5.

[0027] 19. Frias V, Hsu L. Three - dimensionally printed facial prosthesis with a silicone veneer technique: A clinical report. J Prosthet Dent 2023.

[0028] 20. Wen A, Zhu Y, Zheng S, Xiao N, Gao Z, Fu X, et al. Preliminary study on the method of automatically determining facial landmarks based on three - dimensional face template. Chin J Stomatol 2022; 57: 358 - 65.

[0029] 21. White JD, Ortega - Castrillon A, Matthews H, Zaidi AA, Ekrami O, Snyders J, et al. MeshMonk: Open - source large - scale intensive 3D phenotyping. Sci Rep 2019; 9: 6085.

[0030] 22. Wen A, Wang Y, Ye H, Gao Z, Zhu Y, Qin Q, et al. Preliminary study on three-dimensional morphological reconstruction method for external nose defect based on three-dimensional face template. Chin J Stomatol 2023;58:414-21.

[0031] 23. Dong Y. Establishment and Applications of the Three-dimensional Nasal Morphological Database in Han Ethnicity. Fourth Military Medical University, Xi’an, 2010.

[0032] 24. Liu X, Zhao Y, Wu G, Wang Y, Qu L, Shen L. Construction of 3-D standard external nasal morphological database for nasal prostheses. Journal of Practical Stomatology 2006;22:835-7.

[0033] 25. Li F. Medical aesthetic anatomy. Beijing: People's Medical Publishing House; 1999. p.541.

[0034] 26. Gower JC. Procrustes analysis. International Encyclopedia of the Social & Behavioral Sciences 2001;39:12141-3.

[0035] 27. Ruse MK, Calhoun M, Davis BK. Prosthetic Nasal Reconstruction. Facial Plast Surg Cl 2024;32:327-37.

[0036] 28. Amberg B, Romdhani S, Vetter T. Optimal Step Nonrigid ICP Algorithms for Surface Registration. In: 2007, IEEE, 2007: 1-8.

[0037] 29. Feng Y, Wu F, Shao X, Wang Y, Zhou X. Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network. In: Ferrari, Hebert, Sminchisescu, Weiss eds. Cham;2018, Cham: Springer International Publishing, 2018: 557-74.

[0038] 30. Lium O, Kwon YB, Danelakis A, Theoharis T. Robust 3D Face Reconstruction Using One / Two Facial Images. J Imaging 2021;7. Summary of the Invention

[0039] (1) Technical problems to be solved

[0040] The object of the present invention is to provide a construction method and application of a structured three-dimensional human face template dataset with multiple nose morphologies, which can construct a structured three-dimensional human face template dataset with different nose morphologies based on the average data of three-dimensional facial data of different nose morphologies of Chinese people; the three-dimensional human face template dataset of the present invention further improves the generality and clinical suitability of constructing nasal prosthesis morphologies by the "template method"; each piece of template data in the three-dimensional human face template dataset of the present invention has been structurally processed and has a consistent number of vertices and data range.

[0041] (2) Technical solutions

[0042] A construction method of a structured three-dimensional human face template dataset with multiple nose morphologies according to the present invention includes the following steps:

[0043] (1) Collection and processing of three-dimensional facial data of different nose morphologies:

[0044] Collect the three-dimensional facial data of subjects with different nasal forms, and process it in Geomagic Wrap 2021 software to ensure that there are no defects in the facial form of the data. Crop the redundant data, and keep the unified data range from the hairline at the top, to the ears and mandibular angles on both the left and right sides, and along the lower edge of the mandible to below the mental region; then, the same operator marks 32 three-dimensional facial anatomical landmark points of the hairline point, glabella point, and nasion point for each piece of data, and obtains the three-dimensional coordinates of the landmark points;

[0045] (2) Establishment of three-dimensional human face template data:

[0046] The construction of the three-dimensional human face template is illustrated by taking the construction of the three-dimensional human face template of male saddle nose as an example:

[0047] Step 1: Screen out the reference data that uses the PA algorithm to perform size normalization and alignment on all male saddle nose three-dimensional facial data. The yellow points set on the reference data are the positions of the 32 facial anatomical landmark points used for running the PA algorithm;

[0048] Step 2: Construct the remaining three-dimensional facial data of the male saddle nose three-dimensional human face template, with the yellow points the same as before;

[0049] Step 3: Normalize and align all the data together, with different colors representing different male saddle nose three-dimensional facial data;

[0050] Step 4: The average three-dimensional facial data of the male saddle nose three-dimensional facial data;

[0051] Step 5: Complete the construction of the male saddle nose three-dimensional human face template;

[0052] (3) Establishment of the three-dimensional human face template dataset:

[0053] Referring to the Chinese adult head and face size data of GB / T 2428-1998, the average morphological facial length of Chinese adult men / women, the straight-line distance between the nasion point and the submental point: are 119 mm and 109 mm respectively; the average intertragal width, the straight-line distance between the bilateral tragus points: are both 140 mm; select 1 piece of data with the morphological facial length and intertragal width closest to the above measurement values from 10 pieces of male saddle nose data as the reference data for the size scaling and alignment of the remaining data; based on the landmark point coordinates and the PA algorithm, calculate the scaling coefficient and transformation matrix of the remaining data relative to the reference data in MATLAB software, and complete the size normalization and alignment of 10 pieces of male saddle nose data;

[0054] Import the data of 10 cases of male saddle nose processed by the PA algorithm into the Geomagic Wrap 2021 software, use the "average value" function to calculate the average data of the 10 cases of data, and obtain the average data of male saddle nose; move the center point of the average data of male saddle nose to the coordinate origin, and adjust its spatial pose in the Geomagic Wrap 2021 software to the natural head position; delete the facial data on one side based on the median sagittal plane, and convert the remaining half-side facial data into point cloud data; after sampling the point cloud data, define a group of vertices closest to the median sagittal plane in the data as "midline points", and the remaining vertices as "unilateral points"; adjust the X coordinates of all "midline points" to 0, and based on the YZ plane where X = 0, obtain the opposite points of "unilateral points" through mirroring; combine the midline points and bilateral points into a point cloud data, and package it into triangular mesh data to obtain a three-dimensional face template of male saddle nose;

[0055] Using the above process, three-dimensional face templates of men / women with different nose shapes are constructed, and a three-dimensional face template database can be obtained.

[0056] The application of a method for constructing a structured three-dimensional face template dataset with multiple nose shapes according to the present invention includes the steps:

[0057] ① Collect the three-dimensional facial data of nasal defect patients, and select the corresponding three-dimensional face template with reference to the nasal shape before the patient's defect or the nasal shape the patient wants;

[0058] ② Use the non-rigid registration algorithm to deform and register the three-dimensional face template to the three-dimensional facial data of the patient to obtain the data after the face template is deformed;

[0059] ③ Obtain the nasal defect prosthesis restoration shape data based on the data after the face template is deformed.

[0060] (III) Beneficial effects

[0061] The advantages of the present invention are:

[0062] (1) A structured three-dimensional face template dataset with different nose shapes is constructed based on the average data of three-dimensional facial data of different nose shapes of Chinese people;

[0063] (2) The three-dimensional face template dataset further improves the generality and clinical suitability of constructing nasal prosthesis restoration shapes by the "template method";

[0064] (3) Each template data in the three-dimensional face template dataset has been structurally processed and has a consistent number of vertices and data range. Description of the drawings

[0065] Figure 1 is a flow chart of constructing a three-dimensional face template according to the present invention;

[0066] Figure 1 Among them: 1. Reference data for size normalization and alignment using the PA algorithm, with yellow dots representing the positions of 32 facial anatomical landmark points; 2. The remaining 3D facial data for constructing the 3D human face template, with yellow dots as before; 3. Schematic diagram of all data size-normalized and aligned together, with different colors representing different 3D facial data; 4. Average 3D facial data; 5. 3D human face template;

[0067] Figure 2 It is a schematic diagram of 12 3D human face templates constructed by the present invention;

[0068] Figure 2 Among them: 6. Saddle nose of male; 7. Lion nose of male; 8. Bulbous nose of male; 9. Hump nose of male; 10. Aquiline nose of male; 11. Ideal nose of male; 12. Saddle nose of female; 13. Lion nose of female; 14. Bulbous nose of female; 15. Hump nose of female; 16. Aquiline nose of female; 17. Ideal nose of female. Detailed implementation manners

[0069] The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0070] A method for constructing a structured 3D human face template dataset with multiple nose morphologies according to the present invention includes the following steps:

[0071] (1) Collection and processing of 3D facial data with different nose morphologies:

[0072] Collect 3D facial data of subjects with different nose morphologies, and process them in Geomagic Wrap 2021 software to ensure that there are no defects in the facial morphology of the data, crop the redundant data, and keep the data range uniformly from the hairline to the ears and mandibular angles on both the left and right sides, and along the lower edge of the mandible to below the chin; then, the same operator marks 32 3D facial anatomical landmark points of the hairline point, glabella point, and nasion point for each piece of data, and obtains the 3D coordinates of the landmark points;

[0073] (2) Establishment of 3D human face template data:

[0074] Taking the construction of the 3D human face template of the saddle nose of male as an example to illustrate the construction of the 3D human face template:

[0075] Step 1: Screen out the reference data for size normalization and alignment of all 3D facial data of the saddle nose of male using the PA algorithm. The yellow dots set on the reference data are the positions of 32 facial anatomical landmark points used for running the PA algorithm;

[0076] Step 2: Construct the remaining 3D facial data of the 3D human face template of the saddle nose of male, with yellow dots as before;

[0077] Step 3: Normalize and align all the data sizes together, with different colors representing different 3D facial data of male saddle nose;

[0078] Step 4: The average 3D facial data of male saddle nose 3D facial data;

[0079] Step 5: Complete the construction of the male saddle nose 3D face template;

[0080] (3) Establishment of 3D face template data:

[0081] Referring to the head and face size data of Chinese adults in GB / T 2428-1998, the average morphological face length of Chinese adult men / women, the straight-line distance between nasion and submental point: are 119 mm and 109 mm respectively; the average intertragal width, the straight-line distance between bilateral tragus points: are both 140 mm; select 1 piece of data with the morphological face length and intertragal width closest to the above measurement values from 10 pieces of male saddle nose data as the reference data for the scaling and alignment of the remaining data sizes; based on the landmark coordinates and PA algorithm, calculate the scaling coefficient and transformation matrix of the remaining data relative to the reference data in MATLAB software, and complete the size normalization and alignment of 10 pieces of male saddle nose data;

[0082] Import the 10 pieces of male saddle nose data processed by the PA algorithm into Geomagic Wrap 2021 software, use the "average value" function to calculate the average data of the 10 pieces of data, and obtain the average data of male saddle nose; move the center point of the average data of male saddle nose to the coordinate origin, and adjust its spatial pose in Geomagic Wrap 2021 software to the natural head position; delete the facial data on one side based on the median sagittal plane, and convert the remaining half-side facial data into point cloud data; after sampling the point cloud data, define a group of vertices closest to the median sagittal plane in the data as "midline points", and the remaining vertices as "unilateral points"; adjust the X coordinates of all "midline points" to 0, and based on the YZ plane of X = 0, obtain the opposite-side points of "unilateral points" through mirroring; combine the midline points and bilateral points into a point cloud data, and package it into triangular mesh data to obtain the male saddle nose 3D face template;

[0083] Using the above process, construct 3D face templates of men / women with different nose shapes to obtain a 3D face template database.

[0084] The application of a method for constructing a structured 3D face template dataset with multiple nose shapes according to the present invention includes the steps:

[0085] ① Collect the 3D facial data of nasal defect patients, and select the corresponding 3D face template with reference to the nasal shape before the patient's defect or the nasal shape the patient wants;

[0086] ② Use a non-rigid registration algorithm to deform and register the three-dimensional facial template onto the patient's three-dimensional facial data to obtain the deformed data of the facial template.

[0087] ③ Obtain the nasal defect prosthetic morphological data based on the deformed data of the facial template.

[0088] The "scaling factor and transformation matrix" mentioned in the present invention are calculated by the Procrustes Analysis (PA) algorithm based on the landmark pairs with a one-to-one correspondence between different data. Based on the transformation matrix, the data can be translated and rotated, and based on the scaling factor, the overall size of the data can be scaled. Under the action of the "scaling factor and transformation matrix", the sum of the squares of the distances between the landmark pairs can be minimized, thereby realizing the size normalization and overlapping alignment between the data. Currently, the PA algorithm can be directly called as a function in MATLAB software. The specific calculation process includes centroid coincidence of the data, singular value decomposition, etc. The specific algorithm principle and formula are detailed in the following references.

[0089] Gower J C, Dijksterhuis G B. Procrustes problems[M]. Oxford University Press, 2004.

[0090] Ross A. Procrustes analysis[J]. Course report, Department of Computer Science and Engineering, University of South Carolina, 2004, 26: 1-8.

[0091] The description of the relevant software programs in the present invention is as follows:

[0092] MeshMonk: It is a non-rigid registration algorithm program reported in the literature and can realize the deformation matching between three-dimensional point cloud data. In the field of stomatology, it has been applied to the symmetry analysis, morphological analysis of three-dimensional craniofacial point cloud data, and automatic determination of anatomical landmarks, etc. Related literature: White J D, Ortega-Castrillón A, Matthews H, et al. MeshMonk: Open-source large-scale intensive 3D phenotyping[J]. Scientific Reports, 2019, 9(1): 6085.

[0093] Geomagic Wrap 2021: It is a commercial reverse engineering software with powerful functions such as three-dimensional point cloud data processing, editing, and analysis, and has been widely used in the field of stomatology.

[0094] Procrustes Analysis (PA) algorithm: It is a morphological analysis method. For three-dimensional point cloud data, based on the corresponding landmark point pairs on the data, it can calculate the "scaling coefficient and transformation matrix" that minimizes the sum of squared distances between landmark point pairs, thereby realizing size normalization and alignment between data. Related literature: ① Gower J C, Dijksterhuis G B. Procrustes problems[M]. Oxford University Press, 2004. ② Ross A. Procrustes analysis[J]. Course report, Department of Computer Science and Engineering, University of South Carolina, 2004, 26: 1-8.

[0095] MATLAB: It is a commercial professional computing software with powerful functions such as numerical calculation, data analysis, and data visualization, and has been widely used in many fields such as scientific computing, engineering design, data analysis, and algorithm development.

[0096] Non-rigid registration algorithm: It is a general term for a class of algorithms. For three-dimensional point cloud data, such algorithms calculate the dense point correspondence relationships between different data, and then apply different spatial transformation matrices to each vertex on one of the three-dimensional point cloud data (usually a structured three-dimensional point cloud template data) to achieve deformation matching between data. Related literature: Amberg B, Romdhani S, Vetter T. Optimal Step Nonrigid ICP Algorithms for Surface Registration[C]. IEEE, 2007.

[0097] As described above, the present invention can be more fully realized. The above are only relatively reasonable implementation examples of the present invention. The protection scope of the present invention includes but is not limited to this. Any non-substantive variant changes based on the technical solution of the present invention by those skilled in the art are included within the scope of the present invention.

Claims

1. A construction method of a structured three-dimensional human face template dataset with a multi-nose shape, characterized in that It includes the following steps: (1) Acquisition and processing of three-dimensional facial data of different nasal shapes: Acquire the three-dimensional facial data of subjects with different nasal shapes and process it in Geomagic Wrap 2021 software to ensure that there are no defects in the facial shape of the data. Crop the redundant data, and keep the data range uniformly from the hairline to the ears and mandibular angles on both the left and right sides, along the lower edge of the mandible to below the chin. Then, the same operator marks 32 three-dimensional facial anatomical landmark points of the hairline point, glabella point, and nasion point for each piece of data, and obtains the three-dimensional coordinates of the landmark points; (2) Establishment of three-dimensional human face template data: The construction of the three-dimensional human face template is illustrated by taking the construction of the three-dimensional human face template of male saddle nose as an example: Step 1: Screen out the reference data for size normalization and alignment of all male saddle nose three-dimensional facial data using the PA algorithm. The yellow dots set on the reference data are the positions of the 32 facial anatomical landmark points used for running the PA algorithm; Step 2: Construct the remaining three-dimensional facial data of the male saddle nose three-dimensional human face template, with the yellow dots the same as before; Step 3: Normalize and align all the data together, with different colors representing different male saddle nose three-dimensional facial data; Step 4: The average three-dimensional facial data of the male saddle nose three-dimensional facial data; Step 5: Complete the construction of the male saddle nose three-dimensional human face template; (3) Establishment of a three-dimensional human face template dataset: Referring to the Chinese adult head and face size data of GB / T 2428-1998, the average morphological facial length of Chinese adult men / women, and the straight-line distance between the nasion and the mental point: are 119 mm and 109 mm respectively; the average intertragal width, the straight-line distance between the bilateral tragus points: are both 140 mm; select 1 piece of data with the morphological facial length and intertragal width closest to the above measurement values from 10 pieces of male saddle nose data as the reference data for the size scaling and alignment of the remaining data; based on the landmark point coordinates and the PA algorithm, calculate the scaling coefficient and transformation matrix of the remaining data relative to the reference data in MATLAB software, and complete the size normalization and alignment of 10 pieces of male saddle nose data; Import the 10 pieces of male saddle nose data processed by the PA algorithm into Geomagic Wrap 2021 software, use the "Average" function to calculate the average data of the 10 pieces of data, and obtain the average data of male saddle nose; move the center point of the average data of male saddle nose to the coordinate origin, and adjust its spatial pose in Geomagic Wrap 2021 software to the natural head position; based on the median sagittal plane, delete one side of the facial data, and convert the remaining half-side facial data into point cloud data; after sampling the point cloud data, define a group of vertices closest to the median sagittal plane in the data as "midline points", and the remaining vertices as "unilateral points"; adjust the X coordinates of all "midline points" to 0, and based on the YZ plane of X = 0, obtain the opposite side points of "unilateral points" through mirroring; combine the midline points and bilateral points into a point cloud data, and package it into triangular mesh data to obtain the male saddle nose three-dimensional human face template; Using the above process, three-dimensional male / female face templates with different nose shapes are constructed, and a three-dimensional face template dataset can be obtained.

2. Application of a method for constructing a structured three-dimensional face template dataset with multiple nose shapes according to claim 1, characterized in that: ① Collect the three-dimensional facial data of patients with nasal defects, and select the corresponding three-dimensional face template by referring to the nasal shape before the defect of the patient or the nasal shape desired by the patient; ② Use a non-rigid registration algorithm to deform and register the three-dimensional face template onto the three-dimensional facial data of the patient to obtain the data after the face template is deformed; ③ Obtain the nasal defect prosthesis restoration shape data based on the data after the face template is deformed.

Images