Spine cobb angle recognition method and system based on neural network and medium

Abstract

The invention belongs to the technical field of Cobb angle detection, and discloses a spinal column cobb angle identification method and system based on a neural network and a medium, and the method comprises the steps: obtaining spinal column test sample data of a patient, carrying out the alignment processing of the spinal column test sample data, and obtaining the processed spinal column test sample data; a training set is constructed based on the processed spine test sample data, posture data and distance data in the processed spine test sample data serve as training data of the training set, and cobb angle labeling data in the processed spine test sample data serve as label data of the training set; a preset neural network is trained based on the training set, a spine cobb angle recognition model is obtained, and the spine cobb angle recognition model is used for outputting the spine cobb angle of the patient when the actually measured spine data of the patient serve as input. According to the invention, the operation difficulty is reduced, so that non-professionals can also carry out accurate measurement and evaluation.

Description

Technical Field

[0001] The present invention belongs to the technical field of Cobb angle detection, and in particular relates to a spinal Cobb angle recognition method, system and medium based on a neural network. Background Art

[0002] The Cobb angle is a measurement method used to quantify the degree of scoliosis. It is determined on a scoliosis X-ray by measuring the angle formed by the two vertebrae with the most severe spinal curvature.

[0003] The measurement of the Cobb angle is very important for monitoring the progression of scoliosis and evaluating the treatment effect. The normal spine should maintain a certain physiological curvature in both the sagittal and coronal planes, but the scoliosis angle in the coronal plane should be very small, usually not exceeding 10 degrees. The size of the Cobb angle can help doctors judge the severity of scoliosis.

[0004] The traditional method of measuring the Cobb angle relies on X-rays, but this method carries the risk of radiation exposure, places a heavy workload on doctors, and wastes medical resources. Summary of the Invention

[0005] The purpose of the present invention is to provide a spinal Cobb angle identification method, system and medium based on a neural network to solve the problem that the traditional Cobb angle measurement method relies on X-rays, but this method has the risk of radiation exposure, heavy workload for doctors, and waste of medical resources.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for identifying the spinal cobb angle based on a neural network, the method comprising: Acquiring the patient's spinal test sample data, wherein the spinal test sample data includes: posture data, distance data, and Cobb angle annotation data, wherein the posture data and distance data are collected by a three-dimensional electronic spinal measuring instrument; Performing alignment processing on the spinal test sample data to obtain processed spinal test sample data; Constructing a training set based on the processed spinal test sample data, wherein the posture data and distance data in the processed spinal test sample data are used as the training data of the training set, and the Cobb angle annotation data in the processed spinal test sample data are used as the label data of the training set; The preset neural network is trained based on the training set to obtain a spinal Cobb angle recognition model. The spinal Cobb angle recognition model is used to output the patient's spinal Cobb angle when the patient's spinal measured data is used as input.

[0007] Preferably, the spinal column test sample data is aligned to obtain the processed spinal column test sample data, including: Statistical length of the spine test sample data; Determine whether the data length of the spinal test sample data is greater than the preset expected data length. If so, perform data elimination on the spinal test sample data based on the equal-interval elimination method to obtain the eliminated spinal test sample data, and use the eliminated spinal test sample data as the processed spinal test sample data.

[0008] Preferably, the spinal test sample data is eliminated based on the equal-interval elimination method to obtain the eliminated spinal test sample data, including: Calculate the difference between the data length of the spinal test sample data and the preset expected data length; The number and spacing of data to be eliminated by constructing the equal spacing elimination method based on the length difference; The spinal test sample data is eliminated based on the number and spacing of data elimination to obtain the eliminated spinal test sample data.

[0009] Preferably, the method further comprises: If the data length of the spine test sample data is not greater than the preset expected data length, determining whether the data length of the spine test sample data is less than the preset expected data length; If so, the spinal test sample data is expanded based on the adjacent point linear interpolation method to obtain expanded spinal test sample data; the expanded spinal test sample data is used as the processed spinal test sample data.

[0010] Preferably, when the data length of the spinal test sample data is greater than the preset expected data length, the function expression of the adjacent point linear interpolation method is: ; Where, Indicates the supplementary data inserted at the i-th point in the preset expected data, Represents the i-th data in the spine test sample data, Represents the i+1th data in the spine test sample data, Indicates the current interval index, Indicates the next interval index.

[0011] Preferably, the preset neural network is an artificial neural network, and the model parameters of the artificial neural network include: number of model iterations, model tolerance and model learning rate; wherein, the number of model iterations is 1500 to 2500 times, the model tolerance is ±2.0°, and the model learning rate is 0.00005 to 0.002.

[0012] Preferably, the method further comprises: Acquiring the patient's spine measured data, wherein the spine measured data includes: measured posture and measured distance; The measured posture and the measured distance are input into the spinal cobb angle recognition model to obtain the patient's spinal cobb angle.

[0013] In a second aspect, the present invention provides a spinal Cobb angle recognition system based on a neural network, for implementing the above-mentioned spinal Cobb angle recognition method based on a neural network, the system comprising: A data acquisition module is used to acquire the patient's spinal test sample data, wherein the spinal test sample data includes: posture data, distance data and Cobb angle annotation data, wherein the posture data and distance data are collected by a three-dimensional electronic spine measuring instrument; A data processing module is used to align the spinal test sample data to obtain processed spinal test sample data; A training set construction module is used to construct a training set based on the processed spinal test sample data, wherein the posture data and distance data in the processed spinal test sample data are used as the training data of the training set, and the Cobb angle annotation data in the processed spinal test sample data are used as the label data of the training set; The model training module is used to train a preset neural network based on a training set to obtain a spinal Cobb angle recognition model. The spinal Cobb angle recognition model is used to output the patient's spinal Cobb angle when the patient's actual spinal data is used as input.

[0014] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned neural network-based spinal Cobb angle recognition method when executing the computer program.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned neural network-based spinal Cobb angle identification method.

[0016] Beneficial effects: 1. When acquiring the patient's spinal test sample data, the present invention uses a three-dimensional electronic spine measuring instrument to collect posture data and distance data, thereby minimizing the impact on the patient and reducing the patient's radiation exposure; 2. Compared with traditional manual measurement and evaluation methods, this invention can significantly improve the accuracy and efficiency of diagnosis and reduce human errors; 3. The present invention simplifies the measurement process and reduces the difficulty of operation, so that non-professionals can also perform accurate measurement and evaluation; 4. The identification method of the present invention is not only applicable to medical institutions such as hospitals, but can also be used in schools, communities and other places for large-scale scoliosis screening; at the same time, it is also suitable for patients with scoliosis of different ages and degrees, and has wide practicality and applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings: Figure 1 This is a flowchart of a spinal cobb angle identification method based on a neural network provided by one embodiment of the present invention; Figure 2 This is a block diagram of a spinal cobb angle recognition system based on a neural network provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the present invention will be briefly introduced below in conjunction with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0019] Example 1 Figure 1 FIG. 1 is a flowchart of a spinal cobb angle recognition method based on a neural network provided by an embodiment of the present invention. Figure 1 As shown, this embodiment provides a spinal Cobb angle recognition method based on a neural network. The method is applied to a server and executed on the server. The method includes: Step S10: Obtain the patient's spinal test sample data, which includes: posture data, distance data and Cobb angle annotation data, wherein the posture data and distance data are collected by a three-dimensional electronic spine measuring instrument; and during the medical treatment of each patient, an EOS full-length spine film will be generated. The EOS full-length spine film is an X-ray of the entire spine, that is, it is obtained using a traditional X-ray film. The Cobb angle annotation data of the patient can be determined from the EOS full-length spine film; the above data are uploaded to the server to obtain the patient's spinal test sample data.

[0020] In this embodiment, the posture data mainly includes: roll angle, pitch angle and yaw angle. In this embodiment, the roll angle, pitch angle and yaw angle of multiple collection points of the patient are collected by a three-dimensional electronic spine measuring instrument to obtain the posture data of the patient.

[0021] Step S20: performing alignment processing on the spinal column test sample data to obtain processed spinal column test sample data.

[0022] In this embodiment, since the posture data and distance data are mainly obtained by manual measurement using a three-dimensional electronic spine measuring instrument, the acquisition time, the length of the data volume and the overall fitting effect are all different; it is necessary to align the posture data and distance data in the spine test sample data.

[0023] As a further optimization of this embodiment, the spinal column test sample data is aligned to obtain the processed spinal column test sample data, including: Step a10: Count the data length of the spine test sample data. The data length of this embodiment mainly refers to the posture data and distance data in the spine test sample data.

[0024] Step a20: Determine whether the data length of the spinal test sample data is greater than the preset expected data length. If so, perform data elimination on the spinal test sample data based on the equal-interval elimination method to obtain the eliminated spinal test sample data, and use the eliminated spinal test sample data as the processed spinal test sample data.

[0025] In this embodiment, the preset expected data length is preferably 300 data points; wherein, the spinal test sample data is eliminated based on the equal-interval elimination method to obtain the eliminated spinal test sample data, including: Step a201: Calculate the length difference between the data length of the spinal test sample data and the preset expected data length; for example, the data length of the spinal test sample data is 400 data points, and at this time it is still 100 data points away from the preset expected data length, then the length difference is 100 data points.

[0026] Step a202: Construct the number and spacing of data to be eliminated by the equal-interval elimination method based on the length difference; at this time, the number of data eliminated is 100 data points, and the spacing of data eliminated is 4 data points, that is, 1 data point is eliminated every 4 data points, so that the data length of the spinal test sample data is 300 data points.

[0027] Step a203: performing data elimination on the spinal column test sample data based on the number and spacing of data elimination to obtain the eliminated spinal column test sample data.

[0028] As a further optimization of this embodiment, the method further includes: Step b10: If the data length of the spine test sample data is not greater than the preset expected data length, determine whether the data length of the spine test sample data is less than the preset expected data length; Step b20: If yes, the spinal test sample data is expanded based on the adjacent point linear interpolation method to obtain expanded spinal test sample data; the expanded spinal test sample data is used as the processed spinal test sample data.

[0029] In this embodiment, the function expression of the adjacent point linear interpolation method is: ; Where, Indicates the supplementary data inserted at the i-th point in the preset expected data, Represents the i-th data in the spine test sample data, Represents the i+1th data in the spine test sample data, Indicates the current interval index, Indicates the next interval index.

[0030] In this example, each column of the spinal test sample data consists of [m1, m2, m3, …, mn], but the data lengths of each group are inconsistent, that is, n is not equal. Therefore, it is impossible to uniformly incorporate them into model training or prediction. To ensure that the data lengths are equal, linear interpolation of adjacent points is used.

[0031] For example, if you need to insert data between (m1, m2), the number of points to be inserted is determined by the interval index: end_idx - start_idx + 1. The interval index is the integer obtained by taking the modulo of the current array and the desired array. This value is the number of points to be inserted between each pair of points. After the insertion between (m1, m2) is completed, the value of i is incremented, jumping to the point before (m2, m3), and then inserting data between all adjacent points in sequence.

[0032] In this embodiment, the calculation method of the current interval index start_idx is: start_idx = i * (target_size / / (n - 1)); Where i is an increasing value used in the loop, target_size represents the expected data length, n is the length of the spine test sample data, and (target_size / / (n - 1) is the number of points to be inserted between the two points.

[0033] In this embodiment, the calculation method of the next interval index end_idx is: start_idx + (target_size / / (n - 1)); The next interval index end_idx is mainly used to determine the range of two interpolation points, and finally find each Value (that is, the value to be inserted).

[0034] In this embodiment, when the data length of the spinal column test sample data is equal to the preset expected data length, no alignment processing is performed on the spinal column test sample data.

[0035] After completing the alignment of the column test sample data, each patient can obtain a set of four-dimensional feature vectors, that is, a 300*4 feature vector (roll angle, pitch angle, yaw angle and distance data); and each patient has a standard Cobb angle annotation data. At this time, a mapping relationship is established between each patient's posture data and distance data and the patient's Cobb angle annotation data, and a training set is established through the mapping relationship.

[0036] Step S30: Construct a training set based on the processed spinal test sample data, wherein the posture data and distance data in the processed spinal test sample data are used as the training data of the training set, and the Cobb angle annotation data in the processed spinal test sample data are used as the label data of the training set.

[0037] Step S40: training a preset neural network based on the training set to obtain a spinal Cobb angle recognition model, wherein the spinal Cobb angle recognition model is used to output the patient's spinal Cobb angle when the patient's spinal measured data is used as input.

[0038] In this embodiment, the preset neural network is an artificial neural network (ANN), which includes an input layer (300*4 feature vector), an input flattened fully connected layer (1200*1 feature vector), a Relu nonlinear activation function and a pooling layer connected in sequence; the input layer is used to input a four-dimensional feature vector, the input flattened fully connected layer is used to flatten the four-dimensional feature vector and capture the feature information of the front and back nodes, the Relu nonlinear activation function is used to increase the expressive power of the model and enhance the robustness of the model, the pooling layer is used for feature extraction and compression to obtain a data feature set, and finally the spinal cobb angle recognition model completes the cobb angle prediction of the data set.

[0039] In this embodiment, the model parameters of the artificial neural network include: the number of model iterations, the model tolerance and the model learning rate; wherein, the number of model iterations is 1500 to 2500 times, the model tolerance is ±2.0°, and the model learning rate is 0.00005 to 0.002; wherein, the number of model iterations is preferably 2000 times, and the model learning rate is preferably 0.001.

[0040] After the spinal Cobb angle recognition model is trained, the spinal Cobb angle recognition model is deployed to various user terminals (smart terminals such as mobile phones and computers) to perform prediction tasks. Therefore, the method further includes: Step c10: Acquire the patient's spine measured data, wherein the spine measured data includes: measured posture and measured distance.

[0041] Step c20: inputting the measured posture and the measured distance into the spinal Cobb angle recognition model to obtain the patient's spinal Cobb angle.

[0042] The present invention has the following advantages: 1. Improve the accuracy and efficiency of scoliosis diagnosis: By fusing and mapping the output data (posture data and distance data) of the three-dimensional electronic spine measuring instrument with the Cobb angle annotation data in the EOS full-length spine film, automatic and rapid assessment of the scoliosis angle can be achieved; compared with traditional manual measurement and assessment methods, this method can significantly improve the accuracy and efficiency of diagnosis and reduce human errors.

[0043] 2. Traditional scoliosis measurement methods usually require professional doctors or technicians to operate, and the measurement process is cumbersome. By using data alignment processing of fixed-length data windows and the application of ANN neural network models, the measurement process is simplified, the operation difficulty is reduced, and non-professionals can also perform accurate measurements and evaluations.

[0044] 3. The invention combines three-dimensional electronic spine measurement technology with the ANN neural network model. It is innovative and provides new ideas and technical means for the automatic and rapid assessment of scoliosis.

[0045] 4. The identification method of the present invention is not only applicable to medical institutions such as hospitals, but can also be used in schools, communities and other places for large-scale scoliosis screening; at the same time, the method is also applicable to patients with scoliosis of different ages and degrees, and has wide practicality and applicability.

[0046] Example 2 Figure 2 This is a block diagram of a spinal cobb angle recognition system based on a neural network provided by an embodiment of the present invention. Figure 2As shown, this embodiment provides a spinal Cobb angle recognition system based on a neural network, which is used to implement the spinal Cobb angle recognition method based on a neural network in Example 1. The system includes: A data acquisition module is used to acquire the patient's spinal test sample data, wherein the spinal test sample data includes: posture data, distance data and Cobb angle annotation data, wherein the posture data and distance data are collected by a three-dimensional electronic spine measuring instrument; A data processing module is used to align the spinal test sample data to obtain processed spinal test sample data; A training set construction module is used to construct a training set based on the processed spinal test sample data, wherein the posture data and distance data in the processed spinal test sample data are used as the training data of the training set, and the Cobb angle annotation data in the processed spinal test sample data are used as the label data of the training set; The model training module is used to train a preset neural network based on a training set to obtain a spinal Cobb angle recognition model. The spinal Cobb angle recognition model is used to output the patient's spinal Cobb angle when the patient's actual spinal data is used as input.

[0047] This embodiment also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the neural network-based spinal Cobb angle recognition method of the first embodiment is implemented.

[0048] This embodiment further provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the spinal Cobb angle recognition method based on a neural network in the first embodiment is implemented.

[0049] The present invention has the following advantages: 1. Improve the accuracy and efficiency of scoliosis diagnosis: By fusing and mapping the output data (posture data and distance data) of the three-dimensional electronic spine measuring instrument with the Cobb angle annotation data in the EOS full-length spine film, automatic and rapid assessment of the scoliosis angle can be achieved; compared with traditional manual measurement and assessment methods, this method can significantly improve the accuracy and efficiency of diagnosis and reduce human errors.

[0050] 2. Traditional scoliosis measurement methods usually require professional doctors or technicians to operate, and the measurement process is cumbersome. By using data alignment processing of fixed-length data windows and the application of ANN neural network models, the measurement process is simplified, the operation difficulty is reduced, and non-professionals can also perform accurate measurements and evaluations.

[0051] 3. The invention combines three-dimensional electronic spine measurement technology with the ANN neural network model. It is innovative and provides new ideas and technical means for the automatic and rapid assessment of scoliosis.

[0052] 4. The identification method of the present invention is not only applicable to medical institutions such as hospitals, but can also be used in schools, communities and other places for large-scale scoliosis screening; at the same time, the method is also applicable to patients with scoliosis of different ages and degrees, and has wide practicality and applicability.

[0053] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or boxes.

[0055] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A spinal cobb angle recognition method based on neural network, characterized in that: The method comprises: Acquiring the patient's spinal test sample data, wherein the spinal test sample data includes: posture data, distance data, and Cobb angle annotation data, wherein the posture data and distance data are collected by a three-dimensional electronic spinal measuring instrument; Performing alignment processing on the spinal test sample data to obtain processed spinal test sample data; Constructing a training set based on the processed spinal test sample data, wherein the posture data and distance data in the processed spinal test sample data are used as the training data of the training set, and the Cobb angle annotation data in the processed spinal test sample data are used as the label data of the training set; The preset neural network is trained based on the training set to obtain a spinal Cobb angle recognition model. The spinal Cobb angle recognition model is used to output the patient's spinal Cobb angle when the patient's spinal measured data is used as input.

2. The spinal cobb angle recognition method based on neural network according to claim 1 is characterized in that: Align the spinal test sample data to obtain processed spinal test sample data, including: Statistical length of the spine test sample data; Determine whether the data length of the spinal test sample data is greater than the preset expected data length. If so, perform data elimination on the spinal test sample data based on the equal-interval elimination method to obtain the eliminated spinal test sample data, and use the eliminated spinal test sample data as the processed spinal test sample data.

3. The spinal cobb angle identification method based on neural network according to claim 2 is characterized in that: The spinal test sample data is eliminated based on the equal-interval elimination method to obtain the eliminated spinal test sample data, including: Calculate the difference between the data length of the spinal test sample data and the preset expected data length; The number and spacing of data to be eliminated by constructing the equal spacing elimination method based on the length difference; The spinal test sample data is eliminated based on the number and spacing of data elimination to obtain the eliminated spinal test sample data.

4. The spinal cobb angle recognition method based on neural network according to claim 2, characterized in that: The method further comprises: If the data length of the spine test sample data is not greater than the preset expected data length, determining whether the data length of the spine test sample data is less than the preset expected data length; If so, the spinal test sample data is expanded based on the adjacent point linear interpolation method to obtain expanded spinal test sample data; the expanded spinal test sample data is used as the processed spinal test sample data.

5. The spinal cobb angle identification method based on neural network according to claim 4 is characterized in that: The function expression of the adjacent point linear interpolation method is: ； Where, Indicates the supplementary data inserted at the i-th point in the preset expected data, Represents the i-th data in the spine test sample data, Represents the i+1th data in the spine test sample data, Indicates the current interval index, Indicates the next interval index.

6. The spinal cobb angle identification method based on neural network according to claim 1, characterized in that: The preset neural network is an artificial neural network, and the model parameters of the artificial neural network include: the number of model iterations, the model tolerance and the model learning rate; wherein, the number of model iterations is 1500 to 2500 times, the model tolerance is ±2.0°, and the model learning rate is 0.00005 to 0.

002.

7. The spinal cobb angle identification method based on neural network according to claim 1 is characterized in that: The method further comprises: Acquiring the patient's spine measured data, wherein the spine measured data includes: measured posture and measured distance; The measured posture and the measured distance are input into the spinal cobb angle recognition model to obtain the patient's spinal cobb angle.

8. A spinal cobb angle recognition system based on a neural network, used to implement the spinal cobb angle recognition method based on a neural network according to any one of claims 1 to 7, characterized in that: The system comprises: A data acquisition module is used to acquire the patient's spinal test sample data, wherein the spinal test sample data includes: posture data, distance data and Cobb angle annotation data, wherein the posture data and distance data are collected by a three-dimensional electronic spine measuring instrument; A data processing module is used to align the spinal test sample data to obtain processed spinal test sample data; A training set construction module is used to construct a training set based on the processed spinal test sample data, wherein the posture data and distance data in the processed spinal test sample data are used as the training data of the training set, and the Cobb angle annotation data in the processed spinal test sample data are used as the label data of the training set; The model training module is used to train a preset neural network based on a training set to obtain a spinal Cobb angle recognition model. The spinal Cobb angle recognition model is used to output the patient's spinal Cobb angle when the patient's actual spinal data is used as input.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the spinal cobb angle recognition method based on neural network according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the spinal cobb angle recognition method based on a neural network described in any one of claims 1 to 7 is implemented.

Images