CN111311626A - Skull fracture automatic detection method based on CT image and electronic medium - Google Patents
Skull fracture automatic detection method based on CT image and electronic medium Download PDFInfo
- Publication number
- CN111311626A CN111311626A CN202010389715.6A CN202010389715A CN111311626A CN 111311626 A CN111311626 A CN 111311626A CN 202010389715 A CN202010389715 A CN 202010389715A CN 111311626 A CN111311626 A CN 111311626A
- Authority
- CN
- China
- Prior art keywords
- image
- fracture
- skull
- ellipse
- neural network
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30008—Bone
Abstract
The invention discloses a CT image-based skull fracture automatic detection method and an electronic medium, which are applied to the technical field of image processing. The invention carries out automatic fracture detection in an ellipse-expanded polar coordinate conversion mode, judges and outputs fracture areas through a profile gradient detection or a pre-trained neural network, improves the diagnosis efficiency and greatly reduces the rate of missed diagnosis.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a skull fracture automatic detection method based on CT images and an electronic medium.
Background
Skull fractures (skull fractures) represent 15% to 20% of craniocerebral injuries, with a probability of up to 80% occurring in severe craniocerebral injuries, and thus the presence of skull fractures is a risk signal for intracranial lesions. The fracture is classified into skull cap fracture and skull base fracture according to the occurring part, linear fracture and depressed fracture according to the form, open fracture and closed fracture according to whether the fracture is directly or indirectly communicated with the outside, and the like.
Computed Tomography (CT) is the primary method of detecting skull fractures. Accurate diagnosis requires experience accumulation, a plurality of slices need to be manually tracked in a conventional two-dimensional method, and a traditional three-dimensional visualization method needs to perform rotation observation on an image during diagnosis so as to avoid the influence of occlusion. The examination of CT images of the skull is time consuming and laborious and CT is prone to missed diagnosis of mild cranial lesions due to the particular structure and location of the skull.
The skull fracture can occur at any part of the skull, CT is a cross-sectional image, the density resolution is high, and the image has no reconnection, so that comminuted fracture and depressed fracture are more obvious than linear fracture. Due to the unique anatomical morphology of the skull, repeated observation of multiple CT cross sections is required to complete the diagnostic evaluation, which is time-consuming and labor-consuming. The diagnosis efficiency can be improved by a post-processing method, and the existing post-processing method comprises the following steps: 1. MPR, which recombines the image data obtained by volume scanning to obtain two-dimensional reconstruction images of coronal position, sagittal position and any oblique position, and can carefully and accurately observe the fracture condition from different directions, different angles and any planes to achieve the effect of tangentially observing the fracture; 2. VR, three-dimensionally reconstructing image data obtained by volume scanning, can rotate in multiple directions and multiple angles, can obtain clean, clear, well-layered and three-dimensional and visual skull images, and completely displays the anatomical relationship of the skull; however, the MPR method cannot observe the skull structure from the whole, VR has high requirements on operation accuracy and data accuracy, artifacts of three-dimensional reconstruction images, such as motion artifacts, are prone to misdiagnosis, and hidden lesions are not displayed clearly due to poor rotation angle, bone occlusion, unreasonable threshold selection and the like. Judgment of the fracture type requires a doctor to perform rich experience accumulation and skill training.
Disclosure of Invention
The technical purpose is as follows: aiming at the defect that the skull fracture area cannot be accurately detected by a CT image in the prior art, the invention discloses a skull fracture automatic detection method based on the CT image and an electronic medium, which are used for automatically detecting the skull fracture in an ellipse-expanded coordinate conversion mode, so that the diagnosis efficiency is improved, and the missed diagnosis rate is greatly reduced.
The technical scheme is as follows: in order to achieve the technical purpose, the invention adopts the following technical scheme.
A skull fracture automatic detection method based on CT images comprises the following steps:
s1, carrying out CT scanning on the patient to obtain a head CT image;
s2, thin-layer reconstruction is carried out on the head CT image, skull segmentation is carried out, a plurality of skull images respectively comprising a skull and a facial skull are obtained, and each skull image comprises a z-axis position coordinate of the skull image;
s3, performing two-dimensional plane ellipse expansion on the skull image, converting a rectangular coordinate system of the skull image into a polar coordinate system, outputting an ellipse expansion image, performing automatic fracture detection on the ellipse expansion image on the polar coordinate system, and marking a two-dimensional polar coordinate area for outputting a fracture if a fracture state is detected; if the fracture state is not detected, judging that the skull image has no abnormal condition;
and S4, summarizing the two-dimensional polar coordinate regions of the fracture of all the ellipse expansion images, converting the two-dimensional polar coordinate regions into two-dimensional rectangular coordinate regions through coordinate conversion, and marking and outputting the fracture three-dimensional coordinate regions under the rectangular coordinate system by combining the z-axis position coordinates of the skull images corresponding to the ellipse expansion images.
Preferably, the two-dimensional plane expansion is performed on the skull image in S3, and a specific process of converting the rectangular coordinate system of the skull image into the polar coordinate system is as follows: converting the rectangular coordinate system of the skull image into a polar coordinate system by an ellipse expansion method, performing approximate ellipse processing on the skull image, performing coordinate transformation by taking the mass center of the image as the origin and the proportion of the major axis and the minor axis of the ellipse as the same, and outputting an ellipse expansion image, wherein the coordinate transformation formula is as follows:
wherein
An image is expanded for an ellipse of the outputted skull bone image,
for the input image of the skull bone,
is a rectangular coordinate and is a coordinate of the rectangular coordinate,
is a polar coordinate, the transverse axis of the polar coordinate system is an angle, the longitudinal axis is a length, the transformed ellipse major axis is 1, and the transformed ellipse minor axis is
。
Preferably, in S3, the method for automatically detecting the fracture of the elliptical expansion image in the polar coordinate system is contour gradient detection, and the contour gradient detection specifically includes the steps of:
s31, firstly carrying out binarization processing on the elliptical expansion image in the polar coordinate system, and then carrying out contour extraction;
s32, gradient calculation and detection are carried out on the contour extracted from the ellipse expansion image, and the calculation formula of the gradient is as follows:
wherein
Expanding the polar coordinates of the contour in the image for the ellipse;
is the gradient value at the contour polar coordinates;
when the gradient value of the contour polar coordinate point is greater than the threshold value, judging that the fracture exists, summarizing all the contour polar coordinate points with the gradient values greater than the threshold value, and positioning and outputting a two-dimensional polar coordinate area of the fracture; otherwise, the skull image is judged to have no abnormal condition.
Preferably, the contour extraction method in S31 is a region-based extraction method or an edge-based extraction method.
Preferably, the detection method for automatically detecting the fracture of the elliptical expansion image in the polar coordinate system in S3 is a neural network detection, and the neural network detection specifically includes the steps of:
s301, pre-training the neural network: collecting a plurality of head CT images of patients with skull fracture, manually framing a fracture area mask on the head CT images, performing thin layer reconstruction and skull segmentation, and outputting skull images; carrying out ellipse expansion on the skull image, outputting an ellipse expansion image, generating a plurality of training samples from the ellipse expansion image containing the fracture area mask, inputting the training samples into a neural network, and then pre-training the neural network, wherein the ellipse expansion image still comprises an artificially marked fracture area mask;
s302, inputting the ellipse expansion image into a pre-trained neural network, automatically detecting the fracture of the ellipse expansion image through the neural network, and if the fracture state is detected, marking a two-dimensional polar coordinate area of the output fracture by the pre-trained neural network; if no fracture state is detected, the skull image is judged to have no abnormal condition.
Preferably, the specific method for labeling the two-dimensional polar coordinate region of the output fracture by the neural network after pre-training in step S302 includes:
calculating a Hessian matrix of each pixel point in the ellipse expansion image, wherein the Hessian matrix calculation formula is as follows:
wherein
Second order of the elliptically expanded image at the pixel points, respectivelyA derivative;
calculating eigenvalues for each Hessian matrix
And a feature vector corresponding to each feature value, wherein
；
Extracting all pixel points satisfying the linear structure, and marking all coincidences
And
the positions of the pixel points are collected, and a linear structure area is output, wherein the linear structure area is a two-dimensional polar coordinate area of the fracture.
Preferably, the neural network is an AI neural network, Resnet or U-Net.
Preferably, the specific process of neural network pre-training in S301 includes:
building a target detection model or an image segmentation model based on a neural network in the neural network;
inputting training samples into a target detection model or an image segmentation model, and outputting all fracture areas to be selected;
and calculating a loss function of the neural network according to the fracture area to be selected and the truly marked fracture area, and finishing the training of the target detection model or the image segmentation model when the loss function is converged, namely realizing the pre-training of the neural network.
Preferably, in step S2, the skull segmentation method is a threshold-based segmentation method, a region growing-based segmentation method, or a neural network-based segmentation method.
The invention discloses an electronic medium, which comprises a memory and a processor, wherein the memory is connected with the processor, the memory stores at least one instruction which can be executed by the processor, and when the at least one instruction is executed by the processor, the automatic skull fracture detection method based on CT images is realized.
Has the advantages that:
the invention carries out automatic fracture detection in a polar coordinate conversion mode of ellipse expansion, judges and outputs a two-dimensional polar coordinate region of the fracture through a neural network after contour gradient detection or pre-training in an ellipse expansion image, and marks and outputs a three-dimensional coordinate region of the fracture under a rectangular coordinate system by combining a z-axis position coordinate of a skull image corresponding to the ellipse expansion image.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a flow chart of automatic fracture detection using a contour gradient detection method;
FIG. 3 is a schematic diagram of coordinate transformation of the ellipse expansion method of FIG. 2;
FIG. 4 is a flow chart of image results using an ellipse expansion method;
FIG. 5 is a flow chart of automatic fracture detection using neural network detection;
fig. 6 is a flow chart of the pre-training of the neural network of fig. 5.
Detailed Description
The invention will be further explained and explained with reference to the drawings.
The skull fracture can be divided into linear fracture, depressed fracture, comminuted fracture, through open fracture and skull suture separation, the latter four fractures are special in shape and obvious in abnormal anatomical structure, and can be clearly diagnosed in a conventional standard algorithm image. For slight linear fracture, the invention discloses a skull fracture automatic detection method based on CT images and an electronic medium, and examination and output are carried out on the skull fracture automatic detection method and the electronic medium.
As shown in fig. 1, a method for automatically detecting skull fracture based on CT image includes:
s1, carrying out CT scanning on the patient to obtain a head CT image;
s2, thin-layer reconstruction is carried out on the head CT image, skull segmentation is carried out, a plurality of skull images respectively comprising a skull and a facial skull are obtained, and each skull image comprises a z-axis position coordinate of the skull image; the CT thin layer reconstruction has more information display than the thin layer CT scanning, and the thin layer reconstruction is carried out in the CT image according to the parameters of the thin layer reconstruction.
The skull segmentation method is a threshold-based segmentation method, a region growing-based segmentation method or a neural network-based segmentation method, and the neural network in the neural network-based segmentation method is U-Net, SegNet, DeepLab and the like. Taking a threshold-based segmentation method as an example, the specific segmentation process is as follows: the tissue area with the CT value HU of the CT image larger than 1300 and smaller than 5000 is extracted, the tissue with the CT value HU of the CT image larger than 1300 and smaller than 5000 is a skull, the tissue with the HU smaller than 1300 is soft tissue and air, and the tissue with the HU larger than 5000 is metal substance. The CT value is a measure of the density of a local tissue or organ in the human body, and is usually called Hounsfield Unit (HU);
s3, performing two-dimensional plane ellipse expansion on the skull image, converting a rectangular coordinate system of the skull image into a polar coordinate system, outputting an ellipse expansion image, performing automatic fracture detection on the ellipse expansion image on the polar coordinate system, and marking a two-dimensional polar coordinate area for outputting a fracture if a fracture state is detected; if the fracture state is not detected, judging that the skull image has no abnormal condition;
and S4, summarizing the two-dimensional polar coordinate regions of the fracture of all the ellipse expansion images, converting the two-dimensional polar coordinate region under a polar coordinate system into a two-dimensional rectangular coordinate region under a rectangular coordinate system through coordinate conversion, and marking and outputting the fracture three-dimensional coordinate region under the rectangular coordinate system by combining the z-axis position coordinate of the skull image corresponding to the ellipse expansion images.
As shown in fig. 2, in step S3, the two-dimensional plane expansion of the skull image is performed, and the specific process of converting the rectangular coordinate system of the skull image into the polar coordinate system is as follows: converting the rectangular coordinate system of the skull image into a polar coordinate system by an ellipse expansion method, performing approximate ellipse processing on the skull image, performing coordinate transformation by taking the mass center of the image as the origin and the proportion of the major axis and the minor axis of the ellipse as the same, and outputting an ellipse expansion image, wherein the coordinate transformation formula is as follows:
wherein
An image is expanded for an ellipse of the outputted skull bone image,
for the input image of the skull bone,
is a rectangular coordinate and is a coordinate of the rectangular coordinate,
is a polar coordinate, the transverse axis of the polar coordinate system is an angle, the longitudinal axis is a length, the transformed ellipse major axis is 1, and the transformed ellipse minor axis is
。
Example one
In the embodiment, a contour gradient detection method is adopted to further detect the skull fracture area, as shown in the attached figure 2.
The contour gradient detection comprises the following specific steps:
s31, firstly carrying out binarization processing on the elliptical expansion image in the polar coordinate system, and then carrying out contour extraction, wherein the contour extraction precision can be improved by carrying out binarization processing on 0 and 1; the contour extraction method is an extraction method based on a region, an extraction method based on an edge and the like;
and S32, performing gradient calculation and detection on the contour extracted from the elliptical expansion image, as shown in the attached figures 3 and 4. Wherein the gradient is calculated by the formula:
wherein
Expanding the polar coordinates of the contour in the image for the ellipse;
when the gradient value at the contour polar coordinate point is greater than a threshold value, judging that the fracture exists, summarizing all contour polar coordinate points with gradient values greater than the threshold value, and positioning and outputting the fracture position; otherwise, the skull image is judged to have no abnormal condition. The image of the normal skull top bone structure should be a smooth continuous image, the gradient of the outline is small, when the fracture occurs, the continuity of the skull is damaged, the gradient of the fracture area is increased, and the threshold value can be 10 HU/mm.
Example two
In the embodiment, a neural network detection method is adopted to further detect the skull fracture area, as shown in fig. 5.
The neural network detection comprises the following specific steps:
s301, pre-training the neural network: collecting a plurality of head CT images of patients with skull fracture, manually framing a fracture area mask on the head CT images, performing thin layer reconstruction and skull segmentation, and outputting skull images; carrying out ellipse expansion on the skull image, outputting an ellipse expansion image, generating a plurality of training samples from the ellipse expansion image containing the fracture area mask, inputting the training samples into a neural network, and then pre-training the neural network, wherein the ellipse expansion image still comprises an artificially marked fracture area mask;
s302, inputting the ellipse expansion image into a pre-trained neural network, automatically detecting the fracture of the ellipse expansion image through the neural network, and if the fracture state is detected, marking a two-dimensional polar coordinate area of the output fracture by the pre-trained neural network; if the fracture state is not detected, the skull image is judged to have no abnormal condition, and the neural network is AI neural network, Resnet or U-Net. Where Resnet refers to a Residual Network, i.e., a Residual Network, which may not be a convolutional neural Network, but may also be implemented with a full connection layer.
As shown in fig. 6, the specific process of the neural network pre-training in step S301 includes:
1) constructing a sample: collecting CT images of the head of a skull fracture patient to the required number, framing the image to select a fracture area mask, performing thin-layer reconstruction and skull segmentation on the CT image of the head, and outputting a skull image; and performing two-dimensional plane ellipse expansion on the skull image, converting a rectangular coordinate system of the skull image into a polar coordinate system through ellipse expansion, converting a fracture area mask marked in the rectangular coordinate system into the polar coordinate system according to a formula (1) to obtain a plurality of ellipse expansion images, wherein the ellipse expansion images are used as training samples.
2) Building a target detection model or an image segmentation model based on a neural network; training the constructed model by using the constructed training sample, wherein the trained model can automatically detect the fracture of the input ellipse expansion image and output all fracture areas to be selected and the types of the fracture areas, wherein the fracture areas are linear fracture, depressed fracture, comminuted fracture, through open fracture and skull suture separation; and calculating a loss function of the neural network according to the fracture area to be selected and the truly marked fracture area in the model training process, and finishing the training of the target detection model or the image segmentation model when the loss function is converged, namely realizing the pre-training of the neural network.
When the target detection model is of a U-Net structure, the U-Net structure comprises a plurality of convolution layers, a pooling layer, an up-sampling layer and a down-sampling layer, and finally a fracture area mask corresponding to an image is output; the loss function of the U-Net network here is:
wherein the content of the first and second substances,
in order to actually mark the data,
is the prediction data.
The specific method for labeling the two-dimensional polar coordinate region of the output fracture by the neural network after pre-training in the step S302 includes:
calculating a Hessian matrix of each pixel point in the ellipse expansion image, wherein the Hessian matrix calculation formula is as follows:
wherein
Second derivatives of the elliptically expanded image at the pixel points, respectively;
calculating eigenvalues for each Hessian matrix
And a feature vector corresponding to each feature value, wherein
；
Extracting all pixel points satisfying the linear structure, and marking all coincidences
And
the positions of the pixel points are collected, and a linear structure area is output, wherein the linear structure area is a two-dimensional polar coordinate area of the fracture.
Through the two automatic fracture detection methods of the first embodiment and the second embodiment, the fracture is automatically detected in an ellipse expansion polar coordinate conversion mode, a two-dimensional polar coordinate area of the fracture is judged and output through a neural network after contour gradient detection or pre-training in an ellipse expansion image, and a three-dimensional fracture coordinate area under a rectangular coordinate system is marked and output by combining a z-axis position coordinate of a skull image corresponding to the ellipse expansion image.
The invention also discloses an electronic medium which comprises a memory and a processor, wherein the memory is connected with the processor, the memory stores at least one instruction which can be executed by the processor, and when the at least one instruction is executed by the processor, the skull fracture automatic detection method based on the CT image is realized. The memory can be various types of memory, such as random access memory, read only memory, flash memory, and the like. The processor may be various types of processors, such as a central processing unit, a microprocessor, a digital signal processor, or an image processor.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.
Claims (10)
1. A skull fracture automatic detection method based on CT images is characterized by comprising the following steps:
s1, carrying out CT scanning on the patient to obtain a head CT image;
s2, thin-layer reconstruction is carried out on the head CT image, skull segmentation is carried out, a plurality of skull images respectively comprising a skull and a facial skull are obtained, and each skull image comprises a z-axis position coordinate of the skull image;
s3, performing two-dimensional plane ellipse expansion on the skull image, converting a rectangular coordinate system of the skull image into a polar coordinate system, outputting an ellipse expansion image, performing automatic fracture detection on the ellipse expansion image on the polar coordinate system, and marking a two-dimensional polar coordinate area for outputting a fracture if a fracture state is detected; if the fracture state is not detected, judging that the skull image has no abnormal condition;
and S4, summarizing the two-dimensional polar coordinate regions of the fracture of all the ellipse expansion images, converting the two-dimensional polar coordinate regions into two-dimensional rectangular coordinate regions through coordinate conversion, and marking and outputting the fracture three-dimensional coordinate regions under the rectangular coordinate system by combining the z-axis position coordinates of the skull images corresponding to the ellipse expansion images.
2. The automatic skull fracture detection method based on CT image according to claim 1, wherein the two-dimensional plane expansion of the skull image in S3 is performed, and the specific process of converting the rectangular coordinate system of the skull image into the polar coordinate system is as follows: converting the rectangular coordinate system of the skull image into a polar coordinate system by an ellipse expansion method, performing approximate ellipse processing on the skull image, performing coordinate transformation by taking the mass center of the image as the origin and the proportion of the major axis and the minor axis of the ellipse as the same, and outputting an ellipse expansion image, wherein the coordinate transformation formula is as follows:
wherein
An image is expanded for an ellipse of the outputted skull bone image,
for the input image of the skull bone,
is a rectangular coordinate and is a coordinate of the rectangular coordinate,
is a polar coordinate, the transverse axis of the polar coordinate system is an angle, the longitudinal axis is a length, the transformed ellipse major axis is 1, and the transformed ellipse minor axis is
。
3. The automatic skull fracture detection method based on CT image according to claim 1, wherein the detection method for automatically detecting the fracture of the ellipse unfolded image in the polar coordinate system in S3 is contour gradient detection, and the contour gradient detection specifically comprises the following steps:
s31, firstly carrying out binarization processing on the elliptical expansion image in the polar coordinate system, and then carrying out contour extraction;
s32, gradient calculation and detection are carried out on the contour extracted from the ellipse expansion image, and the calculation formula of the gradient is as follows:
wherein
Expanding the polar coordinates of the contour in the image for the ellipse;
is the gradient value at the contour polar coordinates;
when the gradient value of the contour polar coordinate point is greater than the threshold value, judging that the fracture exists, summarizing all the contour polar coordinate points with the gradient values greater than the threshold value, and positioning and outputting a two-dimensional polar coordinate area of the fracture; otherwise, the skull image is judged to have no abnormal condition.
4. The automatic skull fracture detection method based on CT image as claimed in claim 3, wherein the contour extraction method in S31 is a region-based extraction method or an edge-based extraction method.
5. The automatic skull fracture detection method based on the CT image according to claim 1, wherein the detection method for automatically detecting the fracture of the ellipse unfolded image in the polar coordinate system in S3 is neural network detection, and the neural network detection specifically comprises the following steps:
s301, pre-training the neural network: collecting a plurality of head CT images of patients with skull fracture, manually framing a fracture area mask on the head CT images, performing thin layer reconstruction and skull segmentation, and outputting skull images; carrying out ellipse expansion on the skull image, outputting an ellipse expansion image, generating a plurality of training samples from the ellipse expansion image containing the fracture area mask, inputting the training samples into a neural network, and then pre-training the neural network, wherein the ellipse expansion image still comprises an artificially marked fracture area mask;
s302, inputting the ellipse expansion image into a pre-trained neural network, automatically detecting the fracture of the ellipse expansion image through the neural network, and if the fracture state is detected, marking a two-dimensional polar coordinate area of the output fracture by the pre-trained neural network; if no fracture state is detected, the skull image is judged to have no abnormal condition.
6. The automatic skull fracture detection method based on CT image as claimed in claim 5, wherein the specific method for labeling the two-dimensional polar coordinate region of the output fracture by the neural network after pre-training in step S302 comprises:
calculating a Hessian matrix of each pixel point in the ellipse expansion image, wherein the Hessian matrix calculation formula is as follows:
wherein
Second derivatives of the elliptically expanded image at the pixel points, respectively;
calculating eigenvalues for each Hessian matrix
And a feature vector corresponding to each feature value, wherein
；
Extracting all pixel points satisfying the linear structure, and marking all coincidences
And
the positions of the pixel points are collected, and a linear structure area is output, wherein the linear structure area is a two-dimensional polar coordinate area of the fracture.
7. The automatic skull fracture detection method based on CT image according to claim 5, wherein the neural network is AI neural network, Resnet or U-Net.
8. The automatic skull fracture detection method based on CT image as recited in claim 5, wherein the specific process of neural network pre-training in S301 comprises:
building a target detection model or an image segmentation model based on a neural network in the neural network;
inputting training samples into a target detection model or an image segmentation model, and outputting all fracture areas to be selected;
and calculating a loss function of the neural network according to the fracture area to be selected and the actually marked fracture area, and finishing the training of the target detection model or the image segmentation model when the loss function is converged.
9. The method for automatically detecting skull fracture based on CT image according to claim 1, wherein in step S2, the skull segmentation method is a threshold-based segmentation method, a region growing-based segmentation method or a neural network-based segmentation method.
10. An electronic medium, comprising: comprises a memory and a processor, wherein the memory is connected with the processor, the memory stores at least one instruction which can be executed by the processor, and when the at least one instruction is executed by the processor, the method for automatically detecting the skull fracture based on the CT image realizes the method for automatically detecting the skull fracture based on the CT image according to any one of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010389715.6A CN111311626A (en) | 2020-05-11 | 2020-05-11 | Skull fracture automatic detection method based on CT image and electronic medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010389715.6A CN111311626A (en) | 2020-05-11 | 2020-05-11 | Skull fracture automatic detection method based on CT image and electronic medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111311626A true CN111311626A (en) | 2020-06-19 |
Family
ID=71147646
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010389715.6A Pending CN111311626A (en) | 2020-05-11 | 2020-05-11 | Skull fracture automatic detection method based on CT image and electronic medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111311626A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112150473A (en) * | 2020-09-24 | 2020-12-29 | 北京羽医甘蓝信息技术有限公司 | Three-dimensional jaw bone image segmentation modeling method and device based on CT and terminal equipment |
| CN112330701A (en) * | 2020-11-26 | 2021-02-05 | 山东师范大学 | Tissue pathology image cell nucleus segmentation method and system based on polar coordinate representation |
| CN112837226A (en) * | 2021-01-15 | 2021-05-25 | 深圳市铱硙医疗科技有限公司 | Morphology-based method, system, terminal and medium for extracting sagittal plane in brain |
| CN114881992A (en) * | 2022-05-24 | 2022-08-09 | 北京安德医智科技有限公司 | Skull fracture detection method and device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102737250A (en) * | 2011-01-14 | 2012-10-17 | 西门子公司 | Method and system for automatic detection of spinal bone lesions in 3d medical image data |
| CN106504229A (en) * | 2016-09-30 | 2017-03-15 | 上海联影医疗科技有限公司 | The detection method of characteristic point in image |
| CN107154038A (en) * | 2016-04-22 | 2017-09-12 | 孔德兴 | A kind of visual fracture of rib aided diagnosis method of rib |
| CN110458799A (en) * | 2019-06-24 | 2019-11-15 | 上海皓桦科技股份有限公司 | Fracture of rib automatic testing method based on rib cage expanded view |
| CN111091605A (en) * | 2020-03-19 | 2020-05-01 | 南京安科医疗科技有限公司 | Rib visualization method, identification method and computer-readable storage medium |
-
2020
- 2020-05-11 CN CN202010389715.6A patent/CN111311626A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102737250A (en) * | 2011-01-14 | 2012-10-17 | 西门子公司 | Method and system for automatic detection of spinal bone lesions in 3d medical image data |
| CN107154038A (en) * | 2016-04-22 | 2017-09-12 | 孔德兴 | A kind of visual fracture of rib aided diagnosis method of rib |
| CN106504229A (en) * | 2016-09-30 | 2017-03-15 | 上海联影医疗科技有限公司 | The detection method of characteristic point in image |
| CN110458799A (en) * | 2019-06-24 | 2019-11-15 | 上海皓桦科技股份有限公司 | Fracture of rib automatic testing method based on rib cage expanded view |
| CN111091605A (en) * | 2020-03-19 | 2020-05-01 | 南京安科医疗科技有限公司 | Rib visualization method, identification method and computer-readable storage medium |
Non-Patent Citations (3)
| Title |
|---|
| AYUMI YAMADA等: "Preliminary study on the automated skull fracture detection in CT images using black-hat transform", 《HTTP://WWW.FJT.INFO.GIFU-U.AC.JP/PUBLICATION/890.PDF》 * |
| R.AISHWARIYA等: "Computer-Aided Fracture Detection Of X-Ray Images", 《HTTP://WWW.IOSRJOURNALS.ORG/IOSR-JCE/PAPERS/ICRTEM/VOLUME-2/IOSRCS116.PDF?ID=7557》 * |
| 石明国主编: "《现代医学影像技术学》", 30 April 2007 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112150473A (en) * | 2020-09-24 | 2020-12-29 | 北京羽医甘蓝信息技术有限公司 | Three-dimensional jaw bone image segmentation modeling method and device based on CT and terminal equipment |
| CN112330701A (en) * | 2020-11-26 | 2021-02-05 | 山东师范大学 | Tissue pathology image cell nucleus segmentation method and system based on polar coordinate representation |
| CN112837226A (en) * | 2021-01-15 | 2021-05-25 | 深圳市铱硙医疗科技有限公司 | Morphology-based method, system, terminal and medium for extracting sagittal plane in brain |
| CN112837226B (en) * | 2021-01-15 | 2023-11-07 | 深圳市铱硙医疗科技有限公司 | Morphology-based mid-brain sagittal plane extraction method, system, terminal and medium |
| CN114881992A (en) * | 2022-05-24 | 2022-08-09 | 北京安德医智科技有限公司 | Skull fracture detection method and device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107622492B (en) | Lung fissure segmentation method and system | |
| US8605978B2 (en) | Medical image processing apparatus and method, and computer readable recording medium on which is recorded program for the same | |
| JP4931027B2 (en) | Medical image diagnosis support apparatus and method, and program | |
| JP6877868B2 (en) | Image processing equipment, image processing method and image processing program | |
| US7970196B2 (en) | Automatic detection and quantification of plaque in the coronary arteries of subjects from CT scans | |
| CN111311626A (en) | Skull fracture automatic detection method based on CT image and electronic medium | |
| US8483467B2 (en) | Medical image diagnosis assisting apparatus and method, and computer readable recording medium on which is recorded program for the same | |
| US8634628B2 (en) | Medical image processing apparatus, method and program | |
| US8649843B2 (en) | Automated calcium scoring of the aorta | |
| CN114129240B (en) | Method, system and device for generating guide information and electronic equipment | |
| CN110796636A (en) | CT image bone condition detection method and device based on convolutional neural network | |
| CN111374712B (en) | Ultrasonic imaging method and ultrasonic imaging equipment | |
| JP2008212396A (en) | Image processor and its program | |
| CN111340825A (en) | Method and system for generating mediastinal lymph node segmentation model | |
| EP3971830B1 (en) | Pneumonia sign segmentation method and apparatus, medium and electronic device | |
| CN111260669A (en) | Lung lobe segmentation method and device based on CT image | |
| CN113470060B (en) | Coronary artery multi-angle curved surface reconstruction visualization method based on CT image | |
| WO2022164374A1 (en) | Automated measurement of morphometric and geometric parameters of large vessels in computed tomography pulmonary angiography | |
| JP5486616B2 (en) | MEDICAL IMAGE DIAGNOSIS SUPPORT DEVICE, ITS OPERATION METHOD, AND PROGRAM | |
| EP1668592A1 (en) | Method and system for ground glass nodule (ggn) segmentation with shape analysis | |
| CN111415341A (en) | Pneumonia stage evaluation method, pneumonia stage evaluation device, pneumonia stage evaluation medium and electronic equipment | |
| Tong et al. | Computer-aided lung nodule detection based on CT images | |
| JP2003265463A (en) | Image diagnosis support system and image diagnosis support program | |
| Kiraly et al. | A novel visualization method for the ribs within chest volume data | |
| US11963670B2 (en) | Systems and methods for identifying biopsy location coordinates |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200619 |
|
| RJ01 | Rejection of invention patent application after publication |