CN114939988B - Ankle fracture detection and three-dimensional reconstruction method based on AI - Google Patents
Ankle fracture detection and three-dimensional reconstruction method based on AI Download PDFInfo
- Publication number
- CN114939988B CN114939988B CN202210550097.8A CN202210550097A CN114939988B CN 114939988 B CN114939988 B CN 114939988B CN 202210550097 A CN202210550097 A CN 202210550097A CN 114939988 B CN114939988 B CN 114939988B
- Authority
- CN
- China
- Prior art keywords
- ankle
- fracture
- picture
- network
- dimensional reconstruction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 208000027502 Ankle fracture Diseases 0.000 title claims abstract description 68
- 238000001514 detection method Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 24
- 208000010392 Bone Fractures Diseases 0.000 claims abstract description 54
- 206010017076 Fracture Diseases 0.000 claims abstract description 54
- 238000012549 training Methods 0.000 claims abstract description 24
- 210000004233 talus Anatomy 0.000 claims abstract description 23
- 210000003423 ankle Anatomy 0.000 claims abstract description 19
- 230000011218 segmentation Effects 0.000 claims abstract description 15
- 238000010146 3D printing Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000000605 extraction Methods 0.000 claims description 24
- 230000007246 mechanism Effects 0.000 claims description 12
- 210000000988 bone and bone Anatomy 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 230000008014 freezing Effects 0.000 claims description 4
- 238000007710 freezing Methods 0.000 claims description 4
- 238000007639 printing Methods 0.000 claims description 4
- 230000009467 reduction Effects 0.000 claims description 4
- 210000001519 tissue Anatomy 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 2
- 238000010257 thawing Methods 0.000 claims description 2
- 238000003745 diagnosis Methods 0.000 abstract description 3
- 230000005764 inhibitory process Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 208000027418 Wounds and injury Diseases 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000000399 orthopedic effect Effects 0.000 description 3
- 238000002790 cross-validation Methods 0.000 description 2
- 208000020084 Bone disease Diseases 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 210000001137 tarsal bone Anatomy 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
- B29C—SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING
- B29C64/00—Additive manufacturing, i.e. manufacturing of three-dimensional [3D] objects by additive deposition, additive agglomeration or additive layering, e.g. by 3D printing, stereolithography or selective laser sintering
- B29C64/20—Apparatus for additive manufacturing; Details thereof or accessories therefor
- B29C64/245—Platforms or substrates
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
- A61B6/505—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of bone
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
- A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B29—WORKING OF PLASTICS; WORKING OF SUBSTANCES IN A PLASTIC STATE IN GENERAL
- B29C—SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING
- B29C64/00—Additive manufacturing, i.e. manufacturing of three-dimensional [3D] objects by additive deposition, additive agglomeration or additive layering, e.g. by 3D printing, stereolithography or selective laser sintering
- B29C64/30—Auxiliary operations or equipment
- B29C64/386—Data acquisition or data processing for additive manufacturing
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B33—ADDITIVE MANUFACTURING TECHNOLOGY
- B33Y—ADDITIVE MANUFACTURING, i.e. MANUFACTURING OF THREE-DIMENSIONAL [3-D] OBJECTS BY ADDITIVE DEPOSITION, ADDITIVE AGGLOMERATION OR ADDITIVE LAYERING, e.g. BY 3-D PRINTING, STEREOLITHOGRAPHY OR SELECTIVE LASER SINTERING
- B33Y30/00—Apparatus for additive manufacturing; Details thereof or accessories therefor
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B33—ADDITIVE MANUFACTURING TECHNOLOGY
- B33Y—ADDITIVE MANUFACTURING, i.e. MANUFACTURING OF THREE-DIMENSIONAL [3-D] OBJECTS BY ADDITIVE DEPOSITION, ADDITIVE AGGLOMERATION OR ADDITIVE LAYERING, e.g. BY 3-D PRINTING, STEREOLITHOGRAPHY OR SELECTIVE LASER SINTERING
- B33Y50/00—Data acquisition or data processing for additive manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Materials Engineering (AREA)
- Chemical & Material Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Manufacturing & Machinery (AREA)
- Optics & Photonics (AREA)
- Physics & Mathematics (AREA)
- Pathology (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- High Energy & Nuclear Physics (AREA)
- Public Health (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Mechanical Engineering (AREA)
- Radiology & Medical Imaging (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Orthopedic Medicine & Surgery (AREA)
- Dentistry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an ankle fracture detection and three-dimensional reconstruction method based on AI, which comprises the steps of constructing an ankle fracture data set and an ankle bone data set, preprocessing an ankle fracture CT picture in the ankle fracture data set, training an improved Yolov4 model by utilizing the preprocessed ankle fracture data set, training an improved Unet network by utilizing the ankle bone data set, inputting the ankle fracture CT picture to be detected into the trained improved Yolov4 model, and carrying out fracture detection on the ankle fracture CT picture to be detected by the improved Yolov4 model; improving a Unet network to perform non-fracture region inhibition and semantic segmentation on the corresponding ankle CT picture; the three-dimensional reconstruction algorithm carries out three-dimensional reconstruction on the segmented ankle CT picture; the reconstructed model is printed using 3D printing. According to the invention, fracture detection can be carried out on the input ankle CT picture, and under the condition that fracture is detected, semantic segmentation, three-dimensional reconstruction and 3D printing are carried out on the input ankle CT picture, so that a doctor can make diagnosis better.
Description
Technical Field
The invention relates to the field of machine vision, in particular to an ankle fracture detection and three-dimensional reconstruction method based on AI.
Background
With the development of society, the life of people is longer and longer, population aging is more serious, the incidence rate of bone diseases and age correlation are extremely high, with the growth of age, the probability of fracture of human bodies is greatly improved, orthopedics patients are increased, the improvement type detection requirement for orthopedics is continuously improved, and the continuous improvement of the orthopedics industry is promoted.
Ankle fracture is mostly caused by direct or indirect external force, and a great deal of image data is required to be seen every day for an imaging doctor, so that the conditions of visual fatigue, missed diagnosis, misdiagnosis and the like are unavoidable. Because of the difference in cognitive level between doctors and patients, the doctors and the patients cannot effectively communicate, and even if doctors make detailed explanation, the doctors and the patients cannot really understand the doctors and the patients can easily generate medical disputes under the condition of not knowing the illness state and being excited. Therefore, the realization of fracture condition labeling can reduce the workload of image doctors, and the realization of three-dimensional visualization of fracture can strengthen the communication effect of doctors and patients. The existing fracture operation flow is mostly to cut tissues and then manufacture a guide plate, so that the operation wound is large and the exposure time of the wound is long.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides an ankle fracture detection and three-dimensional reconstruction method based on AI, which can detect fracture of an input ankle CT picture, and can perform semantic segmentation, three-dimensional reconstruction and 3D printing on the input ankle CT picture under the condition of detecting fracture so as to help doctors to make diagnosis better.
In order to solve the technical problems, the technical scheme of the invention is as follows: an ankle fracture detection and three-dimensional reconstruction method based on AI, comprising:
constructing an ankle fracture data set and an ankle bone data set; wherein the ankle fracture dataset comprises a plurality of ankle fracture CT pictures labeled with fracture locations, the ankle bone dataset comprises a plurality of ankle CT pictures labeled with bone regions;
preprocessing the ankle fracture CT picture in the ankle fracture data set to delete human tissue information and retain bone information to obtain a preprocessed ankle fracture data set;
training an improved YOLOV4 model by using the preprocessed ankle fracture data set, training an improved Unet network by using the ankle fracture data set, and storing a network model with the best training;
inputting an ankle bone CT picture to be detected into a trained improved YOLOV4 model, wherein the improved YOLOV4 model presents a fracture detection result, and transmitting a fracture image index and a fracture position to a trained improved Unet network under the condition that the ankle bone CT picture to be detected is judged to have fracture;
the trained improved Unet network finds out the corresponding ankle CT picture through the fracture image index and suppresses the non-fracture area of the corresponding ankle CT picture through the fracture position, then performs semantic segmentation on the non-fracture area and transmits the segmented picture index to a three-dimensional reconstruction algorithm;
the three-dimensional reconstruction algorithm finds out the corresponding segmented picture through the segmented picture index and carries out three-dimensional reconstruction on the segmented picture to obtain a reconstruction model, and an STL format file is exported;
printing the reconstructed model using 3D printing.
Further, the preprocessing of the ankle fracture CT picture in the ankle fracture dataset includes:
and performing threshold segmentation on the ankle fracture CT picture in the ankle fracture dataset by adopting an adaptive threshold algorithm.
Further, the improved YOLOV4 model includes a feature extraction network, a feature enhancement network, and a network output layer; wherein,,
the characteristic extraction network adopts a DenseNet network;
the feature enhancement network is added with a CBAM attention mechanism and Convlstm space-time convolution;
the network output layer is divided into three scales according to the characteristics of ankle fracture: 20x20,40x40,60x60.
Further, the training of the modified YOLOV4 model with the pre-treated ankle fracture dataset comprises:
freezing the backbond of the improved YOLOV4 model to train only the predictive network;
then, carrying out unfreezing training on the whole network on the backbone of the improved YOLOV4 model, selecting Adam algorithm to optimize the improved YOLOV4, and selecting step learning rate reduction mode to adjust the learning rate of each epoch; wherein, the initial learning rate is 0.001.
Further, the improved Unet network comprises a feature extraction module, an attention mechanism module and an enhanced feature extraction module; wherein,,
the network in the characteristic extraction module adopts a MobileNet_V3 network;
the attention mechanism adopts an ECA module;
the size of each layer of the enhanced feature extraction module is the same as that of the feature extraction module so as to realize direct stacking.
Further, the training of the modified une network using the ankle bone dataset includes:
firstly, freezing a backup of an improved Unet network, and only training a prediction network;
then, thawing and training the whole network of a backbone of the improved Unet network, selecting an SGD algorithm to optimize the improved UNet, and selecting a cos learning rate reduction mode to adjust the learning rate of each epoch; wherein the initial learning rate is 0.01 and the momentum parameter is set to 0.9.
Further, the suppression of non-fractured regions corresponding to CT pictures of ankle bones by fracture positions comprises
The picture is divided into left and right parts by utilizing the approximate symmetry of the ankle CT picture, the fracture position is used for determining which part is positioned by the fracture, and the gray level of the other part is set to be zero.
Further, the three-dimensional reconstruction algorithm uses a MachingCube improvement algorithm.
Further, before the deriving the STL, the method further includes:
and performing grid smoothing on the reconstruction model.
Further, the 3D printing adopts original proportion printing.
After the technical scheme is adopted, a large amount of clinical data is used for training, so that the robustness of a model is ensured, the fracture picture can be rapidly and accurately detected, semantic segmentation, three-dimensional reconstruction and 3D printing are performed on the fracture picture, the doctor-patient communication obstacle is reduced, the operation implementation efficiency is improved, the medical disputes are reduced, and the operation wound and operation time can be reduced by prefabricating the guide plate.
Drawings
FIG. 1 is a flow chart of an embodiment of an AI-based ankle fracture detection and three-dimensional reconstruction method of the invention;
fig. 2 is a schematic diagram of an operation window of an embodiment of the AI-based ankle fracture detection and three-dimensional reconstruction method according to the present invention.
FIG. 3 is a schematic diagram showing fracture detection results of an embodiment of the AI-based ankle fracture detection and three-dimensional reconstruction method of the invention;
FIG. 4 is a schematic diagram showing the bone segmentation results of an embodiment of the AI-based ankle fracture detection and three-dimensional reconstruction method of the invention;
fig. 5 is a schematic diagram showing three-dimensional reconstruction results of an embodiment of the AI-based ankle fracture detection and three-dimensional reconstruction method according to the present invention.
Detailed Description
In order that the invention may be more readily understood, a more particular description of the invention will be rendered by reference to specific embodiments that are illustrated in the appended drawings.
As shown in fig. 1, an AI-based ankle fracture detection and three-dimensional reconstruction method includes:
step S1: constructing an ankle fracture data set and an ankle bone data set; wherein the ankle fracture dataset comprises a plurality of ankle fracture CT pictures labeled with fracture locations, the ankle bone dataset comprises a plurality of ankle CT pictures labeled with bone regions;
step S2: preprocessing the ankle fracture CT picture in the ankle fracture data set to delete human tissue information and retain bone information to obtain a preprocessed ankle fracture data set;
in an embodiment, the preprocessing the ankle fracture CT picture in the ankle fracture dataset includes:
performing threshold segmentation on the ankle fracture CT picture in the ankle fracture dataset by adopting an adaptive threshold algorithm; the correlation algorithm has been embedded within the YOLOV4 algorithm;
step S3: training an improved YOLOV4 model by using the preprocessed ankle fracture data set, training an improved Unet network by using the ankle fracture data set, and storing a network model with the best training;
in one embodiment, to make full use of the existing dataset, K-fold cross-validation is used during training, with the mosaics data enhancement used during each cross-validation.
In an embodiment, the improved YOLOV4 model includes a feature extraction network, a feature enhancement network, and a network output layer, the feature extraction network uses a DenseNet network to replace an original feature extraction network; the DenseNet network can realize feature reuse, and can reduce model parameter quantity and improve model detection efficiency; the characteristic enhancement network adds a CBAM attention mechanism and ConvLstm space-time convolution on the basis of the original structure; the CBAM attention mechanism combines the spatial attention mechanism and the channel attention mechanism to obtain a better effect; the Convlstm space-time convolution is image space-time convolution, has the image convolution capability and time sequence modeling capability, and can better extract useful features of the previous picture; the network output layer uses finer grid division aiming at ankle fracture characteristics three dimensions, and the three dimensions are respectively as follows: 20x20,40x40,60x60;
in an embodiment, the improved Unet network includes a feature extraction module, an attention mechanism module and an enhanced feature extraction module, wherein the feature extraction module uses a mobilenet_v3 network to replace the original network, and the mobilenet_v3 can reduce the parameter quantity of a model and improve the detection speed of the model; the attention mechanism module is placed between the feature extraction module and the enhanced feature extraction module by adopting an ECA module, namely a channel attention module, and the ECA module has few parameters and good effects. The size of each layer of the enhanced feature extraction module is the same as that of the feature extraction module, and direct stacking can be realized.
Step S4: inputting an ankle bone CT picture to be detected into a trained improved YOLOV4 model, starting detection by the improved YOLOV4 model, judging whether the ankle bone CT picture to be detected has a fracture, ending operation if not, transmitting a fracture image index and a fracture position to an improved UNet network if the fracture is detected, finding out a corresponding ankle bone CT picture through the fracture image index by the improved UNet network and inhibiting a non-fracture area of the corresponding ankle bone CT picture through the fracture position, then carrying out semantic segmentation on the ankle bone CT picture and transmitting the segmented picture index to a three-dimensional reconstruction algorithm;
in an embodiment, an adaptive thresholding algorithm is used to thresholding the fracture data set of the ankle, and the relevant algorithm is already embedded in the YOLOV4 algorithm, so after the CT image of the ankle to be detected is input to the modified YOLOV4, the CT image of the ankle to be detected is preprocessed, and after detection, a detection frame is drawn on the CT image of the ankle.
In one embodiment, the three-dimensional reconstruction algorithm is: the Machingcube improved algorithm is used, the Machingcube improved algorithm can directly reconstruct the segmented picture in three dimensions by using the improved UNet, the reconstruction result only contains skeleton information and does not contain other contents which are slightly different from a skeleton threshold, in addition, the algorithm only reconstructs the fracture part of the ankle bone CT picture, the operand is reduced, and the reconstruction speed is improved;
step S5: the three-dimensional reconstruction algorithm finds out the corresponding segmented picture through the segmented picture index and carries out three-dimensional reconstruction on the segmented picture to obtain a reconstruction model;
step S6: grid smoothing the reconstruction model to derive STL format file;
in particular, the mesh smoothing may reduce the number of reconstructed model patches to reduce stl model size;
step S7: the reconstruction model is printed by adopting 3D printing 1:1, then a fracture guide plate can be prefabricated on the model, and the guide plate is directly used during operation to reduce operation time.
The invention uses a large amount of clinical data for training, thereby ensuring the robustness of the model, being capable of rapidly and accurately detecting the fracture picture, carrying out semantic segmentation, three-dimensional reconstruction and 3D printing on the fracture picture, reducing the communication barriers between doctors and patients, improving the operation implementation efficiency, reducing the medical disputes, and prefabricating the guide plate to reduce the operation wound and the operation time.
The present invention also provides an operation method for the operation window of the above embodiment, where the schematic diagram of the operation window is shown in fig. 2, and the steps include:
leading in an ankle bone CT picture to be detected through a file leading-in button;
realizing fracture detection of the CT picture of the ankle bone to be detected through a start detection button and transmitting a detection result comprising fracture image indexes and fracture positions to an improved UNet network; the detection result is shown in fig. 3;
the segmentation of the restrained picture is realized through a start segmentation button, and the index of the segmented image is transmitted to a three-dimensional reconstruction algorithm; the segmentation result is shown in fig. 4;
realizing three-dimensional reconstruction of the segmented picture through a start reconstruction button; the reconstruction effect is shown in fig. 5;
and exporting the STL format file after model smoothing is realized through an export STL button.
According to the invention, the algorithms are combined through the operation interface, and a user can easily realize the algorithm function by using the interface button.
The technical problems, technical solutions and advantageous effects solved by the present invention have been further described in detail in the above-described embodiments, and it should be understood that the above-described embodiments are only illustrative of the present invention and are not intended to limit the present invention, and any modifications, equivalent substitutions, improvements, etc. within the spirit and principle of the present invention should be included in the scope of protection of the present invention.
Claims (8)
1. An ankle fracture detection and three-dimensional reconstruction method based on AI is characterized in that,
comprising the following steps:
constructing an ankle fracture data set and an ankle bone data set; wherein the ankle fracture dataset comprises a plurality of ankle fracture CT pictures labeled with fracture locations, the ankle bone dataset comprises a plurality of ankle CT pictures labeled with bone regions;
preprocessing the ankle fracture CT picture in the ankle fracture data set to delete human tissue information and retain bone information to obtain a preprocessed ankle fracture data set;
training an improved YOLOV4 model by using the preprocessed ankle fracture data set, training an improved Unet network by using the ankle fracture data set, and storing a network model with the best training;
inputting an ankle bone CT picture to be detected into a trained improved YOLOV4 model, wherein the improved YOLOV4 model presents a fracture detection result, and transmitting a fracture image index and a fracture position to a trained improved Unet network under the condition that the ankle bone CT picture to be detected is judged to have fracture;
the improved Unet network finds out the corresponding ankle CT picture through the fracture image index, suppresses the non-fracture area of the corresponding ankle CT picture through the fracture position, performs semantic segmentation on the non-fracture area, and transmits the segmented picture index to a three-dimensional reconstruction algorithm;
the three-dimensional reconstruction algorithm finds out the corresponding segmented picture through the segmented picture index and carries out three-dimensional reconstruction on the segmented picture to obtain a reconstruction model, and an STL format file is exported;
printing the reconstructed model with 3D printing; wherein,,
the improved YOLOV4 model comprises a feature extraction network, a feature enhancement network and a network output layer; wherein,,
the characteristic extraction network adopts a DenseNet network;
the feature enhancement network is added with a CBAM attention mechanism and Convlstm space-time convolution;
the network output layer is divided into three scales according to the characteristics of ankle fracture: 20x20,40x40,60x60;
the improved Unet network comprises a feature extraction module, an attention mechanism module and an enhanced feature extraction module; wherein,,
the network in the characteristic extraction module adopts a MobileNet_V3 network;
the attention mechanism adopts an ECA module;
the size of each layer of the enhanced feature extraction module is the same as that of the feature extraction module so as to realize direct stacking.
2. The AI-based ankle fracture detection and three-dimensional reconstruction method according to claim 1, wherein,
the preprocessing of the ankle fracture CT picture in the ankle fracture dataset includes:
and performing threshold segmentation on the ankle fracture CT picture in the ankle fracture dataset by adopting an adaptive threshold algorithm.
3. The AI-based ankle fracture detection and three-dimensional reconstruction method according to claim 1, wherein,
the training of the modified YOLOV4 model with the pre-treated ankle fracture dataset comprises:
freezing the backbond of the improved YOLOV4 model to train only the predictive network;
then, carrying out unfreezing training on the whole network on the backbone of the improved YOLOV4 model, selecting Adam algorithm to optimize the improved YOLOV4, and selecting step learning rate reduction mode to adjust the learning rate of each epoch; wherein, the initial learning rate is 0.001.
4. The AI-based ankle fracture detection and three-dimensional reconstruction method according to claim 1, wherein,
the training of the modified une network using the ankle bone dataset comprises:
firstly, freezing a backup of an improved Unet network, and only training a prediction network;
then, thawing and training the whole network of a backbone of the improved Unet network, selecting an SGD algorithm to optimize the improved UNet, and selecting a cos learning rate reduction mode to adjust the learning rate of each epoch; wherein the initial learning rate is 0.01 and the momentum parameter is set to 0.9.
5. The AI-based ankle fracture detection and three-dimensional reconstruction method according to claim 1, wherein,
the method for suppressing non-fracture region corresponding to ankle bone CT picture by fracture position comprises
The picture is divided into left and right parts by utilizing the approximate symmetry of the ankle CT picture, the fracture position is used for determining which part is positioned by the fracture, and the gray level of the other part is set to be zero.
6. The AI-based ankle fracture detection and three-dimensional reconstruction method according to claim 1, wherein,
the three-dimensional reconstruction algorithm uses a Machingcube modified algorithm.
7. The AI-based ankle fracture detection and three-dimensional reconstruction method according to claim 1, wherein,
before the STL format file is exported, the method further comprises:
and performing grid smoothing on the reconstruction model.
8. The AI-based ankle fracture detection and three-dimensional reconstruction method according to claim 1, wherein,
and the 3D printing adopts original proportion printing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210550097.8A CN114939988B (en) | 2022-05-20 | 2022-05-20 | Ankle fracture detection and three-dimensional reconstruction method based on AI |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210550097.8A CN114939988B (en) | 2022-05-20 | 2022-05-20 | Ankle fracture detection and three-dimensional reconstruction method based on AI |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114939988A CN114939988A (en) | 2022-08-26 |
| CN114939988B true CN114939988B (en) | 2023-09-22 |
Family
ID=82908562
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210550097.8A Active CN114939988B (en) | 2022-05-20 | 2022-05-20 | Ankle fracture detection and three-dimensional reconstruction method based on AI |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114939988B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109859233A (en) * | 2018-12-28 | 2019-06-07 | 上海联影智能医疗科技有限公司 | The training method and system of image procossing, image processing model |
| CN111260650A (en) * | 2018-11-15 | 2020-06-09 | 刘华清 | Spine CT sequence image segmentation method and system |
| US20200261231A1 (en) * | 2019-02-19 | 2020-08-20 | The Chinese University Of Hong Kong | Semi-automated imaging reconstruction for orbital fracture repair |
| AU2020101836A4 (en) * | 2020-08-14 | 2020-09-24 | Xi'an university of posts and telecommunications | A method for generating femoral x-ray films based on deep learning and digital reconstruction of radiological image |
| CN113409319A (en) * | 2021-08-17 | 2021-09-17 | 点内（上海）生物科技有限公司 | Rib fracture detection model training system, method, detection system and detection method |
| CN113658142A (en) * | 2021-08-19 | 2021-11-16 | 江苏金马扬名信息技术股份有限公司 | Hip joint femur near-end segmentation method based on improved U-Net neural network |
| CN114187278A (en) * | 2021-12-14 | 2022-03-15 | 山东众阳健康科技集团有限公司 | Rib tiny fracture detection method and system |
| CN114261095A (en) * | 2022-03-03 | 2022-04-01 | 西安博恩生物科技有限公司 | AI-based orthopedic 3D printing method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210370993A1 (en) * | 2020-05-27 | 2021-12-02 | University Of South Carolina | Computer vision based real-time pixel-level railroad track components detection system |
-
2022
- 2022-05-20 CN CN202210550097.8A patent/CN114939988B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111260650A (en) * | 2018-11-15 | 2020-06-09 | 刘华清 | Spine CT sequence image segmentation method and system |
| CN109859233A (en) * | 2018-12-28 | 2019-06-07 | 上海联影智能医疗科技有限公司 | The training method and system of image procossing, image processing model |
| US20200261231A1 (en) * | 2019-02-19 | 2020-08-20 | The Chinese University Of Hong Kong | Semi-automated imaging reconstruction for orbital fracture repair |
| AU2020101836A4 (en) * | 2020-08-14 | 2020-09-24 | Xi'an university of posts and telecommunications | A method for generating femoral x-ray films based on deep learning and digital reconstruction of radiological image |
| CN113409319A (en) * | 2021-08-17 | 2021-09-17 | 点内（上海）生物科技有限公司 | Rib fracture detection model training system, method, detection system and detection method |
| CN113658142A (en) * | 2021-08-19 | 2021-11-16 | 江苏金马扬名信息技术股份有限公司 | Hip joint femur near-end segmentation method based on improved U-Net neural network |
| CN114187278A (en) * | 2021-12-14 | 2022-03-15 | 山东众阳健康科技集团有限公司 | Rib tiny fracture detection method and system |
| CN114261095A (en) * | 2022-03-03 | 2022-04-01 | 西安博恩生物科技有限公司 | AI-based orthopedic 3D printing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114939988A (en) | 2022-08-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108596884B (en) | Esophagus cancer segmentation method in chest CT image | |
| CN106372390B (en) | A kind of self-service healthy cloud service system of prevention lung cancer based on depth convolutional neural networks | |
| CN107730489A (en) | Wireless capsule endoscope small intestine disease variant computer assisted detection system and detection method | |
| CN111353539A (en) | Cervical OCT image classification method and system based on double-path attention convolutional neural network | |
| CN110443268B (en) | Liver CT image benign and malignant classification method based on deep learning | |
| CN110910404B (en) | Anti-noise data breast ultrasonic nodule segmentation method | |
| CN114841320A (en) | Organ automatic segmentation method based on laryngoscope medical image | |
| GB2605391A (en) | Medical Image Analysis Using Neural Networks | |
| CN115719334A (en) | Medical image evaluation method, device, equipment and medium based on artificial intelligence | |
| Pradhan et al. | Lung cancer detection using 3D convolutional neural networks | |
| CN1846616A (en) | Computer aided method of predicting mammary cancer risk | |
| CN116342927A (en) | Breast cancer image classification system and method based on ultrasonic radiography video | |
| JP2024143991A (en) | Image segmentation method and system in a multitask learning network | |
| CN114939988B (en) | Ankle fracture detection and three-dimensional reconstruction method based on AI | |
| CN117058307A (en) | Method, system, equipment and storage medium for generating heart three-dimensional nuclear magnetic resonance image | |
| CN111513743B (en) | Fracture detection method and device | |
| Boyanapalli et al. | Ovarian cancer detection in computed tomography images using ensembled deep optimized learning classifier | |
| CN117952944A (en) | Multitasking recursive convolutional network for rib fracture time prediction | |
| Nijaguna et al. | Novel Ultrasonic Image Reconstruction and Deep Learning Model-Based Categorization for Accurate Liver Disease Diagnosis | |
| CN112967254A (en) | Lung disease identification and detection method based on chest CT image | |
| Bar et al. | Improved ich classification using task-dependent learning | |
| CN116385317A (en) | Low-dose CT image recovery method based on self-adaptive convolution and transducer mixed structure | |
| Wang et al. | X-ray images detection of COVID-19 based on deepwise separable DenseNet | |
| CN112862787B (en) | CTA image data processing method, device and storage medium | |
| Aruna et al. | Cloud based decision support system for diagnosis of breast cancer using digital mammograms |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |