Method, device, equipment and medium for spinal function assessment based on image analysis

Abstract

The embodiment of the present invention relates to the field of smart medical technology, and discloses a spinal function assessment method, apparatus, equipment and medium based on image analysis. The embodiment of the present invention obtains a spinal image sequence of a user; uses a pre-trained spinal function recognition model to perform spinal key point recognition on the spinal image sequence to obtain spinal key point information, performs feature engineering feature calculation on the spinal key point information to obtain a spinal state feature set; performs a fully connected feature recognition operation based on a preset target spinal function type on the spinal state feature set to obtain scores for each key factor corresponding to the target spinal function type, performs a weighted sum calculation on each score based on the preset evaluation weights of each key factor for the target spinal function type, and obtains the user's evaluation score for the target spinal function type. Utilizing the solution described in the present invention can improve the accuracy of spinal function assessment.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of smart medical technology, and in particular to a spinal function assessment method, device, equipment, and medium based on image analysis. Background Art

[0002] In existing medical technologies, the assessment of spinal function mainly relies on spinal scanning images to evaluate the spinal condition. However, due to the influence of scanning technology, spinal scans are performed multiple times and for a long time, which can easily cause damage to the patient's body cells during the scanning process.

[0003] Furthermore, traditional examination methods are unable to objectively and scientifically document the extent of spinal changes. These assessments are often influenced by the individual experience and skill level of the observer, leading to subjective and inconsistent results. Current spinal status assessment devices are unable to consistently capture images of the patient's spine during movement, making it impossible to objectively and accurately assess the true condition of the spine. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a spinal function assessment method, device, electronic device and storage medium based on image analysis, which can improve the accuracy of spinal function assessment.

[0005] To solve the above technical problems, one embodiment of the present invention provides a spinal function assessment method based on image analysis, the method comprising:

[0006] Obtaining a user's spine image sequence;

[0007] Using a pre-trained spinal function recognition model, the spinal key point recognition is performed on the spinal image sequence to obtain spinal key point information, and feature engineering feature calculation is performed on the spinal key point information to obtain a spinal state feature set;

[0008] A fully connected feature recognition operation based on a preset target spinal function type is performed on the spinal status feature set to obtain scores for each key factor corresponding to the target spinal function type, and a weighted sum calculation is performed on each score based on the preset evaluation weights of the target spinal function type for each key factor to obtain the user's evaluation score for the target spinal function type.

[0009] Another embodiment of the present invention further provides a spinal function assessment device based on image analysis, the device comprising:

[0010] A user spine scanning module, used to obtain a sequence of user spine images;

[0011] A spine feature extraction module is used to use a pre-trained spine function recognition model to perform spine key point recognition on the spine image sequence to obtain spine key point information, and perform feature engineering feature calculation on the spine key point information to obtain a spine state feature set;

[0012] The spinal function assessment module is used to perform a fully connected feature recognition operation on the spinal status feature set based on a preset target spinal function type, obtain the scores of each key factor corresponding to the target spinal function type, and perform a weighted sum calculation on each score based on the preset assessment weights of the target spinal function type for each key factor, to obtain the user's assessment score for the target spinal function type.

[0013] Other embodiments of the present invention also provide an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above-mentioned image analysis-based spinal function assessment method.

[0014] Other embodiments of the present invention further provide a computer-readable storage medium storing a computer program, which implements the above-mentioned spinal function assessment method based on image analysis when executed by a processor.

[0015] In an embodiment of the present invention, computer vision intelligent technology and algorithms are introduced to achieve measurement and analysis of the user's spinal movement, thereby avoiding the influence of the doctor's subjective experience on the evaluation results.

[0016] In addition, the acquiring of a user's spine image sequence includes:

[0017] Acquiring a sequence of motion images when the user performs a preset standard detection action;

[0018] The motion image sequence is subjected to spinal morphology prediction and recognition to obtain spinal prediction results corresponding to each extreme position action image in the motion image sequence, and each spinal prediction result is arranged to obtain a spinal image sequence.

[0019] In addition, the step of performing spinal morphology prediction and recognition on the motion image sequence to obtain spinal prediction results corresponding to each extreme position action image in the motion image sequence includes:

[0020] Performing a feature extraction operation based on human outline and human posture on the motion image sequence to obtain human outline features and human posture features;

[0021] A spine posture prediction operation is performed on the human body contour features and human body posture features to obtain a spine prediction result sequence corresponding to the motion image sequence, and spine prediction results corresponding to each extreme position action image in the motion image sequence are extracted from the spine prediction result sequence.

[0022] In addition, after obtaining the spine prediction results corresponding to each extreme position action image in the motion image sequence, the method includes:

[0023] Performing an image noise reduction operation on the spine prediction result using a Gaussian algorithm to obtain a noise-reduced image, and performing a contrast enhancement operation on the noise-reduced image using a histogram equalization algorithm to obtain a contrast-enhanced image;

[0024] An edge detection algorithm is used to perform a spine detection and interception operation on the contrast-enhanced image to obtain an image-enhanced spine prediction result.

[0025] In addition, before using the pre-trained spinal function recognition model, the method further includes:

[0026] Obtain a pre-built spinal function recognition model and a training sample set based on clinical statistics;

[0027] Extracting a target training sample from the training sample set in sequence, performing a feature extraction operation on the target training sample using the spinal function recognition model, and performing random forest classification judgment on the feature extraction result to obtain a predicted spinal assessment result;

[0028] Obtaining, according to a random forest algorithm, a correlation score of each feature in the feature extraction result with the predicted spinal assessment result, and obtaining key features in the feature extraction result based on an arrangement result of each of the correlation scores;

[0029] Calculating the loss between the predicted spine assessment result and the true label of the target training sample using a cross entropy loss algorithm;

[0030] Calculating the minimization of the loss value to obtain model parameters, and performing a network backpropagation operation on the spinal function recognition model according to the model parameters to obtain an updated spinal function recognition model;

[0031] Determining the convergence of the loss value;

[0032] When the loss value does not converge, returning to the above step of sequentially extracting a target training sample from the training sample set, and training the spinal function recognition model based on the spinal status assessment of the key features;

[0033] When the loss value converges, the training process is stopped to obtain a trained spinal function recognition model.

[0034] In addition, the feature engineering feature calculation is performed on the spine key point information to obtain a spine state feature set, including:

[0035] Acquiring key features in the spinal function recognition model, and performing secondary construction on each of the key features based on preset feature engineering to obtain secondary features;

[0036] According to the weight configuration rules in the preset feature engineering, the key features and the secondary features are weighted to obtain a spinal state feature set.

[0037] In addition, after obtaining the user's spine image sequence, the method further includes:

[0038] Obtain the user's default lying posture and real spine image;

[0039] Performing spinal feature recognition on the real spinal image to obtain a real spinal feature set;

[0040] extracting a target spinal column image from the spinal column image sequence in sequence, and acquiring a target motion image corresponding to the target spinal column image;

[0041] identifying motion change features between the target motion image and the default lying-flat motion;

[0042] Performing spinal morphology correction on the target spinal image according to the real spinal feature set and the motion change feature to obtain a corrected spinal image corresponding to the target motion image;

[0043] The corrected spinal images corresponding to the target spinal images in the spinal image sequence are acquired to obtain a user-personalized corrected spinal image sequence.

[0044] In the embodiment of the present invention, the spinal morphology is first predicted and identified to avoid long spinal scans for the user, thereby improving user safety and greatly improving image acquisition efficiency. Then, the pain questionnaire and spinal image sequence are spliced ​​together to identify the spinal status of the user during each standard detection action, thereby improving the efficiency of spinal status assessment and avoiding the risk of subjective recording and judgment by doctors. Therefore, the present invention provides a spinal function assessment method, device, electronic device and storage medium based on image analysis, which can improve the accuracy of spinal function assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] One or more embodiments of the present invention are exemplarily illustrated by the images in the corresponding drawings. These exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements. Unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0046] Figure 1 is a flow chart of a spinal function assessment method based on image analysis according to one embodiment of the present invention;

[0047] Figure 2 is a schematic structural diagram of a spinal function assessment device based on image analysis according to another embodiment of the present invention;

[0048] Figure 3 is a structural diagram of an electronic device provided according to another embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, each embodiment of the present invention will be described in detail below with reference to the accompanying drawings. However, it will be understood by those skilled in the art that in each embodiment of the present invention, many technical details are proposed to enable readers to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in this application can be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation of the present embodiment. The various embodiments can be combined with each other and referenced to each other under the premise that there is no contradiction.

[0050] One embodiment of the present invention relates to a spinal function assessment method based on image analysis. In existing medical technologies, the assessment of a user's spinal function mainly relies on the doctor's clinical experience, the user's subjective description, and traditional physical examination methods. Although these methods can assess the user's condition to a certain extent, they have many shortcomings. First, traditional physical examination methods are often affected by the doctor's personal experience and skill level, and the assessment results may be subjective and inconsistent. Secondly, the user's subjective description is often affected by subjective feelings such as pain and fatigue, and it is difficult to accurately reflect their true spinal function status. In addition, existing assessment methods often lack continuous monitoring and recording of dynamic changes in spinal function, making it difficult to objectively evaluate the treatment effect.

[0051] To solve the above problems, the present invention provides a spinal function assessment method based on image analysis, such as Figure 1 As shown, the present invention specifically includes the following steps:

[0052] S1. Obtain a user's spine image sequence.

[0053] In detail, in an embodiment of the present invention, obtaining a sequence of spinal images of a user includes:

[0054] Acquiring a sequence of motion images when the user performs a preset standard detection action;

[0055] The motion image sequence is subjected to spinal morphology prediction and recognition to obtain spinal prediction results corresponding to each extreme position action image in the motion image sequence, and each spinal prediction result is arranged to obtain a spinal image sequence.

[0056] In the embodiment of the present invention, the standard detection action refers to the action of the user bending down, stretching backward, or twisting left and right under the instruction of the doctor.

[0057] In an embodiment of the present invention, a sequence of motion images of the user can be obtained through optical shooting equipment or millimeter wave, microwave radar and other equipment. The extreme position refers to a state where the movement reaches a certain limit and cannot continue. For example, when bending over to a certain degree and cannot bend over anymore, a snapshot is taken at this time to obtain an extreme position action image.

[0058] In detail, in an embodiment of the present invention, the step of performing spinal morphology prediction and recognition on the motion image sequence to obtain spinal prediction results corresponding to each extreme position action image in the motion image sequence includes:

[0059] Performing a feature extraction operation based on human outline and human posture on the motion image sequence to obtain human outline features and human posture features;

[0060] A spine posture prediction operation is performed on the human body contour features and human body posture features to obtain a spine prediction result corresponding to the motion image sequence, and the spine prediction result corresponding to each extreme position action image in the motion image sequence is extracted.

[0061] In the embodiment of the present invention, a model pre-trained with a large number of samples is used to realize the prediction of the human figure contour and human body posture and the spine.

[0062] Furthermore, in an embodiment of the present invention, after obtaining the spine prediction results corresponding to each extreme position action image in the motion image sequence, the method includes:

[0063] Performing an image noise reduction operation on the spine prediction result using a Gaussian algorithm to obtain a noise-reduced image, and performing a contrast enhancement operation on the noise-reduced image using a histogram equalization algorithm to obtain a contrast-enhanced image;

[0064] An edge detection algorithm is used to perform a spine detection and interception operation on the contrast-enhanced image to obtain an image-enhanced spine prediction result.

[0065] The Gaussian algorithm is a smoothing technique based on a Gaussian function, used to blur images and remove detail and noise. The Gaussian function is a representation of the normal distribution, and its use in image processing helps achieve a balance between preserving image edges and reducing noise.

[0066] The edge detection algorithm can identify points where image intensity (brightness) changes significantly. These changes usually correspond to changes in the outline of an object, surface edges, or other visual attributes, thereby identifying the location of the boundaries of objects in the image.

[0067] The histogram equalization algorithm is a method designed to enhance the contrast of an image by changing the histogram distribution of the image. By redistributing the gray levels of the image, the histogram (i.e., the frequency distribution diagram of the gray levels) becomes more uniform, thereby evenly using all available gray levels throughout the image. The present invention can enhance the visual effect of the image through the histogram equalization algorithm, thereby improving the accuracy of intelligent evaluation.

[0068] The present invention optimizes the spine prediction results through the above-mentioned preprocessing operation, thereby improving the recognition accuracy of spine features in subsequent images.

[0069] Furthermore, in another embodiment of the present invention, after obtaining the user's spine image sequence, the method further includes:

[0070] Obtain the user's default lying posture and real spine image;

[0071] Performing spinal feature recognition on the real spinal image to obtain a real spinal feature set;

[0072] extracting a target spinal column image from the spinal column image sequence in sequence, and acquiring a target motion image corresponding to the target spinal column image;

[0073] identifying motion change features between the target motion image and the default lying-flat motion;

[0074] Performing spinal morphology correction on the target spinal image according to the real spinal feature set and the motion change feature to obtain a corrected spinal image corresponding to the target motion image;

[0075] The corrected spinal images corresponding to the target spinal images in the spinal image sequence are acquired to obtain a user-personalized corrected spinal image sequence.

[0076] The real spinal image may be a magnetic resonance imaging image, a CT scanning image, etc. The default lying posture is the user's lying posture at the moment the magnetic resonance imaging image is obtained.

[0077] Although spine recognition can be performed using a pre-built spine morphology prediction model (which predicts spine shape based on torso movement, body weight, and other information), individual differences exist, requiring personalized correction of the prediction results.

[0078] In addition, in another embodiment of the present invention, a pain questionnaire of the user when performing the standard detection action can also be obtained.

[0079] In another embodiment of the present invention, the user can record the process of performing the standard detection action to obtain a pain questionnaire.

[0080] The pain questionnaire includes pain type and pain level; the pain type includes tingling, dull pain and burning sensation; the pain level includes painless, mild, moderate, severe and extremely severe.

[0081] The pain questionnaire in the embodiment of the present invention can construct a key-value pair of answer number and option, where the answer number represents the standard detection action and the option represents the pain type and intensity. In the embodiment of the present invention, quantization is performed using a one-hot encoding method to obtain a pain feature set.

[0082] The features of the pain questionnaire are then spliced ​​into the corresponding positions of the spinal image sequence to improve the richness of data features, thereby improving the functional assessment accuracy of the subsequent spinal function recognition model.

[0083] S2. Using a pre-trained spinal function recognition model, the spinal key point recognition is performed on the spinal image sequence to obtain spinal key point information, and feature engineering feature calculation is performed on the spinal key point information to obtain a spinal state feature set.

[0084] In an embodiment of the present invention, the spinal function recognition model is a Transformer-based convolutional neural network, which is used to identify a targeted assessment score of the user's spinal condition based on spinal morphology.

[0085] Furthermore, in an embodiment of the present invention, before using the pre-trained spinal function recognition model, the method further includes:

[0086] Obtain a pre-built spinal function recognition model and a training sample set based on clinical statistics;

[0087] Extracting a target training sample from the training sample set in sequence, performing a feature extraction operation on the target training sample using the spinal function recognition model, and performing random forest classification judgment on the feature extraction result to obtain a predicted spinal assessment result;

[0088] Obtaining, according to a random forest algorithm, a correlation score of each feature in the feature extraction result with the predicted spinal assessment result, and obtaining key features in the feature extraction result based on an arrangement result of each of the correlation scores;

[0089] Calculating the loss between the predicted spine assessment result and the true label of the target training sample using a cross entropy loss algorithm;

[0090] Calculating the minimization of the loss value to obtain model parameters, and performing a network backpropagation operation on the spinal function recognition model according to the model parameters to obtain an updated spinal function recognition model;

[0091] Determining the convergence of the loss value;

[0092] When the loss value does not converge, returning to the above step of sequentially extracting a target training sample from the training sample set, and training the spinal function recognition model based on the spinal status assessment of the key features;

[0093] When the loss value converges, the training process is stopped to obtain a trained spinal function recognition model.

[0094] In the embodiment of the present invention, the training sample set is obtained from clinical medical statistics, and includes a user's spinal image sequence, pain questions and answers, and corresponding expert group evaluation results (true labels).

[0095] In the embodiment of the present invention, training is performed using a cross-entropy loss algorithm, and the training progress is controlled by the convergence of the loss value to obtain a trained spinal function recognition model. In addition, the present invention also uses a random forest algorithm to obtain the correlation score of each feature in the feature extraction result with the predicted spinal assessment result, thereby identifying key features in the spinal image that have a significant impact on the preset target spinal function type (specific spinal state).

[0096] The random forest is a learning method based on decision tree ensembles. It constructs multiple decision trees and aggregates their predictions (e.g., by voting or averaging) to make a final decision. It is commonly used for feature selection and importance assessment. The cross-entropy loss function is primarily used to evaluate predictions for classification problems, especially when the output is a probability distribution. It measures the difference between the actual label and the predicted label, with the goal of minimizing this difference.

[0097] In addition, during the model training process, feature engineering feature calculations need to be performed on the spinal key point information to obtain a spinal state feature set.

[0098] Specifically, during the model training process of the embodiment of the present invention, the feature engineering feature calculation is performed on the spine key point information to obtain a spine state feature set, including:

[0099] Acquiring key features in the spinal function recognition model, and performing secondary construction on each of the key features based on preset feature engineering to obtain secondary features;

[0100] According to the weight configuration rules in the preset feature engineering, the key features and the secondary features are weighted to obtain a spinal state feature set.

[0101] Specifically, when a key feature is identified, such as the vector (ab) between the midpoints of both hips (a) and the seventh cervical vertebra (b), the key feature can be reconstructed according to the preset feature engineering to obtain secondary features, such as calculating the angle between the vector (ab) and the frontal plane of the human body.

[0102] Furthermore, after model training is completed, the present invention obtains the average of the correlation scores corresponding to each key feature in the trained model, ultimately obtaining the correlation score of each feature, thereby assigning corresponding weights to obtain a set of spinal status features, for example [key features (weight 1): midpoints of both hips, seventh cervical vertebra point, distal points of both fingers, unilateral wrist point...; and secondary features (weight 0.8): spinal flexion - overall, spinal flexion - finger-to-ground distance, spinal extension, scoliosis - left / right, scoliosis - distance (L / R)...] …], wherein the spinal flexion-overall refers to [the angle between the midpoints of both hips (a) and the seventh cervical vertebra (b) is calculated using the vector (ab) between them and the frontal plane to determine the change in flexion angle during trunk flexion]; the spinal flexion-finger-ground distance refers to [the distance between the distal points of both hands and the ground is calculated and averaged]; and the spinal extension refers to [the angle between the midpoints of both hips and the seventh cervical vertebra is calculated using the vector between them and the frontal plane to determine the change in trunk movement angle during trunk extension], etc. These key features evaluated by the model are not detailed here.

[0103] S3. Perform a fully connected feature recognition operation on the spinal state feature set based on a preset target spinal function type to obtain the scores of the key factors corresponding to the target spinal function type, and perform a weighted sum calculation on each of the scores based on the preset evaluation weights of the target spinal function type for each of the key factors to obtain the evaluation score of the user in terms of the target spinal function type. In the embodiment of the present invention, the target spinal function type is a certain specific spinal state, such as ankylosing spondylosis, scoliosis, etc., and this application does not impose specific restrictions on this. The key factors may include spinal flexion, spinal extension, scoliosis, cervical scoliosis, ankle joint spacing, etc., and this application does not impose specific restrictions on this.

[0104] In the embodiment of the present invention, a spinal status assessment operation is implemented through a feature random forest full connection operation to obtain an assessment score, which can be expressed as a normalized score. The larger the score, the worse the spinal status, and vice versa.

[0105] In the embodiment of the present invention, the spinal morphology is first predicted and identified to avoid long spinal scans for the user, thereby improving user safety and greatly increasing image acquisition efficiency. Then, by splicing a pain questionnaire and a spinal image sequence, the spinal state of the user during each standard test action is identified, thereby improving assessment efficiency and avoiding the risk of subjective recording and judgment by the doctor. Therefore, the present invention provides a spinal function assessment method based on image analysis that can improve the accuracy of spinal function assessment.

[0106] The above is a basic process. In practice, different methods and techniques may be used depending on the specific situation. It is worth noting that the field of image processing and computer vision is complex and diverse, and it is necessary to select appropriate algorithms and techniques based on the specific problem and application scenario to extract key feature points of the spine.

[0107] The steps of the above method are divided only for the purpose of clear description. During implementation, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent. Adding insignificant modifications or introducing insignificant designs to the algorithm or process without changing the core design of the algorithm and process are all within the scope of protection of this patent.

[0108] Another embodiment of the present invention relates to a spinal function assessment device based on image analysis. Figure 2 As shown, the spinal function assessment device based on image analysis includes: a user spine scanning module 201 for acquiring a sequence of spinal images of a user;

[0109] The spine feature extraction module 202 is used to use a pre-trained spine function recognition model to perform spine key point recognition on the spine image sequence to obtain spine key point information, and perform feature engineering feature calculation on the spine key point information to obtain a spine state feature set;

[0110] The spinal function evaluation module 203 is used to perform a fully connected feature recognition operation on the spinal status feature set based on a preset target spinal function type, obtain the scores of each key factor corresponding to the target spinal function type, and perform a weighted sum calculation on each score based on the preset evaluation weights of the target spinal function type for each key factor, to obtain the user's evaluation score for the target spinal function type.

[0111] It is not difficult to find that this embodiment is a device embodiment corresponding to the above-mentioned method embodiment, and this embodiment can be implemented in conjunction with the above-mentioned method embodiment. The relevant technical details mentioned in the above-mentioned method embodiment are still valid in this embodiment. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above-mentioned method embodiment.

[0112] It's worth noting that all modules involved in this embodiment are logical modules. In actual applications, a logical unit can be a physical unit, a portion of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovations of this embodiment, this embodiment does not include units that are not closely related to solving the technical problems raised by this embodiment. However, this does not mean that other units do not exist in this embodiment.

[0113] Another embodiment of the present invention relates to an electronic device, such as Figure 3 As shown, it includes at least one processor 301; and a memory 302 that is communicatively connected to the at least one processor; wherein the memory 302 stores instructions that can be executed by the at least one processor 301, and the instructions are executed by the at least one processor 301 to enable the at least one processor 301 to perform the spinal function assessment method based on image analysis as described above.

[0114] The memory 302 and processor 301 are connected using a bus. The bus can include any number of interconnected buses and bridges, connecting various circuits of one or more processors 301 and memory 302. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits. These are all well known in the art and are therefore not described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor 301 is transmitted over a wireless medium via an antenna. Furthermore, the antenna receives data and transmits it to the processor 301.

[0115] The processor 301 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. The memory 302 can be used to store data used by the processor 301 when performing operations.

[0116] Another embodiment of the present invention relates to a computer-readable storage medium storing a computer program, which implements the above method embodiment when executed by a processor.

[0117] That is, those skilled in the art will understand that all or part of the steps in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a program, which is stored in a storage medium and includes a number of instructions for causing a device (which may be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0118] Those skilled in the art will appreciate that the above-mentioned embodiments are specific examples for implementing the present embodiment, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present embodiment.

Claims

1. A spinal function assessment method based on image analysis, characterized in that: The method comprises: Obtaining a spinal image sequence corresponding to when the user performs a preset standard detection action; Obtain the user's default lying posture and real spine image; Performing spinal feature recognition on the real spinal image to obtain a real spinal feature set; extracting a target spinal column image from the spinal column image sequence in sequence, and acquiring a target motion image corresponding to the target spinal column image; identifying motion change features between the target motion image and the default lying-flat motion; Performing spinal morphology correction on the target spinal image according to the real spinal feature set and the motion change feature to obtain a corrected spinal image corresponding to the target motion image; Acquire a corrected spinal image corresponding to each target spinal image in the spinal image sequence to obtain a user-personalized corrected spinal image sequence; Obtaining a pain questionnaire of the user performing the standard detection action, wherein the pain questionnaire constructs a key-value pair between the standard detection action and the pain level; quantifying the key-value pairs to obtain a pain feature set, and splicing the features of the pain questionnaire to corresponding positions of the spinal image sequence; Using a pre-trained spinal function recognition model, the spinal key point recognition is performed on the spinal image sequence to obtain spinal key point information, and feature engineering feature calculation is performed on the spinal key point information to obtain a spinal state feature set; A fully connected feature recognition operation based on a preset target spinal function type is performed on the spinal status feature set to obtain scores for each key factor corresponding to the target spinal function type, and a weighted sum calculation is performed on each score based on the preset evaluation weights of the target spinal function type for each key factor to obtain the user's evaluation score for the target spinal function type.

2. The spinal function assessment method based on image analysis according to claim 1, characterized in that: The step of obtaining a sequence of images of the user's spine comprises: Acquiring a sequence of motion images when the user performs a preset standard detection action; The motion image sequence is subjected to spinal morphology prediction and recognition to obtain spinal prediction results corresponding to each extreme position action image in the motion image sequence, and each spinal prediction result is arranged to obtain a spinal image sequence.

3. The spinal function assessment method based on image analysis according to claim 2, characterized in that: The step of performing spinal morphology prediction and recognition on the motion image sequence to obtain spinal prediction results corresponding to each extreme position action image in the motion image sequence includes: Performing a feature extraction operation based on human outline and human posture on the motion image sequence to obtain human outline features and human posture features; A spine posture prediction operation is performed on the human body contour features and human body posture features to obtain a spine prediction result sequence corresponding to the motion image sequence, and spine prediction results corresponding to each extreme position action image in the motion image sequence are extracted from the spine prediction result sequence.

4. The spinal function assessment method based on image analysis according to claim 2, characterized in that: After obtaining the spine prediction results corresponding to the extreme position action images in the motion image sequence, the method includes: Performing an image noise reduction operation on the spine prediction result using a Gaussian algorithm to obtain a noise-reduced image, and performing a contrast enhancement operation on the noise-reduced image using a histogram equalization algorithm to obtain a contrast-enhanced image; An edge detection algorithm is used to perform a spine detection and interception operation on the contrast-enhanced image to obtain an image-enhanced spine prediction result.

5. The spinal function assessment method based on image analysis according to claim 1, characterized in that: Before using the pre-trained spinal function recognition model, the method further includes: Obtain a pre-built spinal function recognition model and a training sample set based on clinical statistics; Extracting a target training sample from the training sample set in sequence, performing a feature extraction operation on the target training sample using the spinal function recognition model, and performing random forest classification judgment on the feature extraction result to obtain a predicted spinal assessment result; Obtaining, according to a random forest algorithm, a correlation score of each feature in the feature extraction result with the predicted spinal assessment result, and obtaining key features in the feature extraction result based on an arrangement result of each of the correlation scores; Calculating the loss between the predicted spine assessment result and the true label of the target training sample using a cross entropy loss algorithm; Calculating the minimization of the loss value to obtain model parameters, and performing a network backpropagation operation on the spinal function recognition model according to the model parameters to obtain an updated spinal function recognition model; Determining the convergence of the loss value; When the loss value does not converge, returning to the above step of sequentially extracting a target training sample from the training sample set, and training the spinal function recognition model based on the spinal status assessment of the key features; When the loss value converges, the training process is stopped to obtain a trained spinal function recognition model.

6. The spinal function assessment method based on image analysis according to claim 1, characterized in that: The feature engineering feature calculation is performed on the spine key point information to obtain a spine state feature set, including: Acquiring key features in the spinal function recognition model, and performing secondary construction on each of the key features based on preset feature engineering to obtain secondary features; According to the weight configuration rules in the preset feature engineering, the key features and the secondary features are weighted to obtain a spinal state feature set.

7. A spinal function assessment device based on image analysis, characterized in that: The device comprises: A user spine scanning module is used to obtain a spine image sequence corresponding to when the user performs a preset standard detection action, obtain the user's default lying action and real spine image, perform spine feature recognition on the real spine image to obtain a real spine feature set, extract a target spine image from the spine image sequence in turn, and obtain a target motion image corresponding to the target spine image; identify the motion change features of the target motion image and the default lying action; perform spine morphology correction on the target spine image according to the real spine feature set and the motion change features to obtain a corrected spine image corresponding to the target motion image; obtain the corrected spine image corresponding to each target spine image in the spine image sequence to obtain a user-personalized corrected spine image sequence; obtain a pain questionnaire when the user performs the standard detection action, wherein a key-value pair between the standard detection action and the pain level is constructed in the pain questionnaire; quantify the key-value pair to obtain a pain feature set, and splice the features of the pain questionnaire to the corresponding positions of the spine image sequence; A spine feature extraction module is used to use a pre-trained spine function recognition model to perform spine key point recognition on the spine image sequence to obtain spine key point information, and perform feature engineering feature calculation on the spine key point information to obtain a spine state feature set; The spinal function assessment module is used to perform a fully connected feature recognition operation on the spinal status feature set based on a preset target spinal function type, obtain the scores of each key factor corresponding to the target spinal function type, and perform a weighted sum calculation on each score based on the preset assessment weights of the target spinal function type for each key factor, to obtain the user's assessment score for the target spinal function type.

8. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the spinal function assessment method based on image analysis as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the spinal function assessment method based on image analysis according to any one of claims 1 to 6 is implemented.

Images