Intelligent fracture diagnosis system based on image recognition

Abstract

The invention relates to the technical field of image processing, in particular to an intelligent fracture diagnosis system based on image recognition, which comprises an image analysis module, a mode recognition module, a form analysis module, a risk assessment module and an auxiliary decision module. According to the method, skeleton gray level distribution and boundary consistency are analyzed through continuous frames of X-ray images, fracture feature extraction precision and time sequence coherence are improved, key point space distribution, symmetry standards and form proportions are fused, fracture area structured quantitative evaluation is achieved, the form change trend and abnormal offset point screening are combined, and the accuracy of fracture feature extraction is improved. The method enhances abnormal trajectory recognition precision, associates bone mineral density and form offset, calibrates high-risk time periods, improves risk assessment perspectiveness and individual adaptability, dynamically corrects output content according to an inter-frame suggestion change trend and extended feedback, enhances timeliness of diagnosis suggestions and closed-loop feedback quality, and improves risk assessment accuracy. Image evolution, structural geometry and physiological data are integrally fused, and multi-dimensional intelligent judgment of fracture recognition and evaluation is achieved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an intelligent fracture diagnosis system based on image recognition. Background Art

[0002] The field of image processing technology includes related technologies for acquiring, analyzing, processing and reconstructing image or video information. Its core content includes image acquisition, image processing, feature extraction, image segmentation, image recognition and analysis. This technology field has wide applications in many industries such as medicine, industrial inspection, and security monitoring. In the medical field, image processing is widely used in auxiliary diagnosis, disease monitoring, and surgical navigation. Especially in the automatic recognition and analysis of medical images, it is gradually developing towards intelligence and precision, which has promoted the formation and progress of computer-aided diagnosis systems.

[0003] Among them, the intelligent fracture diagnosis system refers to a diagnostic assistance system that identifies and analyzes fracture sites based on medical imaging data. The system mainly focuses on the task of fracture identification using X-ray images. Specifically, it constructs an image classification model based on convolutional neural networks to realize automatic classification and positioning of fracture images, extracts anatomical feature points of bone structure through key point detection algorithms, and realizes quantitative analysis of fracture areas. At the same time, it uses fracture edge detection technology to identify fracture lines in images, thereby supporting the judgment and classification of fracture morphology. It combines image enhancement algorithms to optimize image quality to improve the accuracy of subsequent identification and analysis. By establishing training sets and label sets to supervise the model learning, the correspondence between images and fracture features is modeled.

[0004] Traditional techniques rely on fracture structural information extracted from single-frame images, ignoring the changing trends of fracture features in time-series images. This makes it difficult to reflect the dynamic evolution of fracture development, resulting in biased diagnostic results. Key point localization fails to incorporate spatial correspondence for symmetry assessment, leading to errors in the analysis of anatomical consistency in complex fracture morphologies and prone to misidentification. Existing methods often rely solely on edge detection to assess fracture edge changes. They lack mechanisms for comparing consecutive frame morphologies and identifying abnormal trajectories, making it difficult to accurately identify irregular fracture edge expansion trends and relying on subjective judgments of fracture severity. Existing systems lack a deep integration of the relationship between bone density and fracture evolution, making it difficult to dynamically identify differences in structural risk and individual bone quality, resulting in inaccurate and ineffective assessment results. Diagnostic recommendations are based on fixed model rules and fail to consider feedback on current frame feature fluctuations and regional expansion trends. This results in delayed and static recommendations and makes it difficult to dynamically adapt to the real-time evolution of the fracture region. This directly impacts the response speed, recognition accuracy, and risk prediction capabilities of fracture diagnosis systems, thereby diminishing their clinical diagnostic value. Summary of the Invention

[0005] In order to solve the technical problems existing in the prior art, the embodiment of the present invention provides an intelligent fracture diagnosis system based on image recognition. The technical solution is as follows:

[0006] In one aspect, an intelligent fracture diagnosis system based on image recognition is provided, the system comprising:

[0007] The image analysis module analyzes the grayscale distribution of the bone area in consecutive frames based on X-ray images collected by medical imaging equipment, compares the grayscale gradient changes between adjacent frames, and determines the boundary consistency of the fracture area to obtain the initial characteristic distribution of the fracture;

[0008] The pattern recognition module screens the bone key point data at the corresponding moment based on the initial fracture feature distribution, compares the corresponding position relationship of the key points in space, determines whether it meets the anatomical symmetry standard, and obtains the fracture feature quantification value based on the ratio of the length and width of the fracture area and the number and distribution density of the key points;

[0009] The morphological analysis module optimizes and selects the target area based on the fracture feature quantization value, analyzes the bone morphological changes and fracture edge fluctuations within five consecutive frames, compares the change amplitude of the fracture edge in each frame, and selects abnormal deviation points in the morphological fitting to obtain the fracture morphological deviation characteristics;

[0010] Based on the fracture morphological deviation characteristics, the risk assessment module determines the overlap with the current interval of the bone density monitoring unit, screens the time period in the predicted trajectory that exceeds the current bone density interval, compares the difference between the bone density change and the expansion trend of the fracture area, and obtains the fracture risk assessment parameters.

[0011] As a further solution of the present invention, the initial characteristic distribution of the fracture includes the regional grayscale mean, gradient change amplitude, and boundary continuity index; the fracture characteristic quantization value includes the symmetry error value, fracture morphology ratio, and key point density; the fracture morphology offset characteristics include the edge displacement amplitude, morphology fluctuation frequency, and fitting abnormal point distribution; the fracture risk assessment parameters include bone density overlap coefficient, risk period range, and expansion trend deviation value.

[0012] As a further solution of the present invention, the image analysis module includes:

[0013] The grayscale analysis submodule analyzes the grayscale distribution characteristics of the bone area in three consecutive frames of X-ray images collected by medical imaging equipment, compares the time distribution characteristics of the grayscale curve in each frame, determines whether the sampling segment is continuous and representative, and generates a periodic sampling grayscale sequence;

[0014] The gradient analysis submodule calculates the rate of change between the end and the start grayscale in the grayscale sequence between adjacent frames based on the periodic sampling grayscale sequence, compares the amplitude difference of the rate change, identifies the data combination with obvious break characteristics, and obtains the grayscale gradient interval;

[0015] The boundary consistency determination submodule calls the grayscale gradient interval to determine whether the continuity of the fracture region boundary between adjacent frames remains consistent, and screens the segments with boundary consistency and fracture amplitude to obtain the initial fracture feature distribution.

[0016] As a further solution of the present invention, the pattern recognition module includes:

[0017] The data screening submodule analyzes the bone key point data at the corresponding moment based on the initial fracture feature distribution, screens the bone node monitoring results in the same frame, determines whether the data segments are complete and continuous, and arranges them in spatial order to obtain a bone key point distribution sequence;

[0018] The position comparison submodule calls the skeleton key point distribution sequence, compares the spatial position relationship of each group of skeleton key points, analyzes the corresponding distances between the difference nodes, determines whether the distances fall within the reference range, selects the nodes that meet the requirements, and obtains the skeleton key point distribution characteristics;

[0019] The fracture area quantification submodule is based on the distribution characteristics of the bone key points, analyzes the proportional relationship between the length and width of the node fracture area, calculates the morphological characteristic value of the fracture area, determines the number of nodes corresponding to the fracture area, integrates the topological relationship of the bone network, and obtains the fracture feature quantification value.

[0020] As a further solution of the present invention, the fracture area morphological characteristic value adopts the formula:

[0021]

[0022] Among them, R represents the morphological characteristic value of the fracture area, L i represents the length of the fracture area of ​​the i-th node, Represents the arithmetic mean of the length of the fracture area, W i represents the width of the fracture area of ​​the i-th node, represents the arithmetic mean of the width of the fracture area, n represents the number of nodes in the fracture area being analyzed, and i is the node number in the summation symbol.

[0023] As a further solution of the present invention, the morphological analysis module includes:

[0024] The target screening submodule screens the target area with associated characteristics based on the fracture feature quantization value, organizes the bone morphology and fracture edge data within five consecutive frames of the target area, completes data collection in frame order, and obtains the frame target area data set;

[0025] The morphological analysis submodule analyzes the correspondence between the bone morphological changes and the fracture edge fluctuations in each frame based on the frame target area dataset, compares the synchronous changes of the fracture edge in each frame, and classifies the fluctuation states in the different time periods to obtain the morphological correspondence distribution;

[0026] The abnormal point screening submodule judges the frame data in the morphological correspondence distribution, screens the frame abnormal points with deviation in the fluctuation amplitude in the morphological fitting analysis, and identifies their position and change state in the monitoring sequence to obtain the fracture morphological deviation characteristics.

[0027] As a further solution of the present invention, the risk assessment module includes:

[0028] The interval judgment submodule judges the coverage of the offset characteristics within the current interval of the bone density monitoring unit based on the fracture morphology offset characteristics, selects overlapping bone density change segments, and integrates the corresponding time series according to the segment order to obtain the interval coverage sequence;

[0029] The trajectory screening submodule screens the time periods where the predicted trajectory continuously exceeds the current bone density interval based on the interval coverage sequence, analyzes the bone density changes and fracture area expansion trends within the segments, summarizes the change characteristics of the difference segments, and obtains the trajectory offset segments;

[0030] The risk parameter generation submodule compares the difference between the bone density change in the trajectory offset segment and the fracture area expansion trend, determines the correspondence between the bone density change and the fracture expansion trend in the key segment, calculates the correspondence index value, and integrates the segment fluctuation data to obtain the fracture risk assessment parameter.

[0031] As a further solution of the present invention, the corresponding relationship index value adopts the formula:

[0032]

[0033] T Δ Represents the corresponding relationship index value, B k represents the bone density value measured along the offset trajectory in the kth sub-segment, F k represents the extended length of the corresponding fracture region in the kth sub-segment, represents the average value of the fracture extension length of the neutron segment in the key segment θ, represents the average value of the bone density value of the sub-segment in the key segment φ, ΔP θrepresents the maximum and minimum difference between the bone density gradient in the key segment θ, and m represents the number of sub-segments in the key segment.

[0034] As a further solution of the present invention, the system further includes a decision support module:

[0035] The auxiliary decision-making module adjusts the diagnostic suggestion output based on the fracture risk assessment parameters, analyzes the change range of the target suggestion in the current frame, selects the target suggestion with the best change range, executes the output, and collects the expansion range of the fracture area within the output cycle to obtain a fracture imaging diagnosis suggestion feedback report;

[0036] The fracture imaging diagnosis suggestion feedback report includes a suggested adjustment range, a regional expansion intensity, and an output stability index.

[0037] As a further solution of the present invention, the auxiliary decision module includes:

[0038] The suggestion adjustment submodule determines the relationship between the current diagnostic suggestion and the risk characteristics based on the fracture risk assessment parameters, optimizes the suggestion adjustment configuration, adjusts the available suggestions according to the frame data, screens the outputtable target suggestions, and obtains the suggestion change range;

[0039] The target comparison submodule compares the change characteristics of the corresponding target suggestions within the frame based on the suggestion change amplitude, analyzes the response performance of each target suggestion, selects the target suggestion with the best change amplitude performance, and obtains the suggestion response difference;

[0040] The output response acquisition submodule performs optimal target suggestion output based on the suggestion response difference, collects changes in the fracture area within the output cycle, and analyzes the fracture extension interval to obtain a fracture imaging diagnosis suggestion feedback report.

[0041] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0042] By meticulously analyzing the grayscale distribution of bone regions based on continuous X-ray images and superimposing grayscale gradient trends and boundary consistency comparisons, the accuracy of fracture feature extraction at the image level is enhanced, overcoming the limitations of static single-frame image analysis and improving the completeness and temporal consistency of initial feature recognition. By comparing the spatial position information of skeletal key points with anatomical symmetry standards and integrating the morphological proportion characteristics of the fracture region with the number of key points, a more three-dimensional and structured quantitative assessment of the fracture region is achieved, strengthening the understanding of the spatial relationships within complex fracture regions. By dynamically comparing the morphological change trends of continuous frames with the degree of fracture edge fluctuation, combined with the screening of abnormal offset points in the fitting model, the ability to accurately identify abnormal morphological evolution trajectories is ensured, enhancing the sensitivity and accuracy of identifying abnormal structural evolution. By correlating fracture morphological offset trends with bone density data, anomalies in overlap between predicted trajectories and bone density intervals are identified, accurately calibrating potential high-risk periods, and effectively improving the prospective nature and individual adaptability of fracture risk assessment. The output content is dynamically adjusted based on the stability trend of the diagnostic suggestion's change amplitude between image frames. Simultaneously, combined with feedback on the expansion trend of the fracture area, a diagnostic suggestion correction mechanism with time-series backtracking is formed, enhancing the timeliness and pertinence of diagnostic suggestions and the quality of the feedback loop. The overall processing flow integrates the evolution of image features, structural geometry, dynamic morphological changes, and physiological parameters, enabling multi-dimensional fusion judgment capabilities in image processing, significantly improving the intelligent level of fracture area identification, quantification, and risk assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 is a schematic diagram of an intelligent fracture diagnosis system based on image recognition provided by an embodiment of the present invention;

[0045] Figure 2 Schematic diagram of the system framework of the present invention;

[0046] Figure 3 This is a flow chart of the image analysis module in the present invention;

[0047] Figure 4 This is a flow chart of the pattern recognition module in the present invention;

[0048] Figure 5 This is a flow chart of the morphological analysis module in the present invention;

[0049] Figure 6 This is a flow chart of the risk assessment module in the present invention;

[0050] Figure 7 This is a flow chart of the auxiliary decision-making module in the present invention. DETAILED DESCRIPTION

[0051] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0052] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0053] In the embodiments of the present invention, the terms "image" and "picture" may be used interchangeably. It should be noted that, when the distinction between them is not emphasized, their intended meanings are the same. The terms "of," "corresponding," and "corresponding" may be used interchangeably. It should be noted that, when the distinction between them is not emphasized, their intended meanings are the same.

[0054] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0055] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0056] The embodiment of the present invention provides an intelligent fracture diagnosis system based on image recognition, such as Figure 1-2 The diagram of the intelligent fracture diagnosis system based on image recognition is shown in FIG. The system includes:

[0057] The image analysis module analyzes the grayscale distribution of the bone area in consecutive frames based on X-ray images collected by medical imaging equipment, compares the grayscale gradient changes between adjacent frames, and determines the boundary consistency of the fracture area to obtain the initial characteristic distribution of the fracture;

[0058] The pattern recognition module screens the bone key point data at the corresponding moment based on the initial fracture feature distribution, compares the corresponding position relationship of the key points in space, and determines whether it meets the anatomical symmetry standard. Combined with the proportional relationship between the length and width of the fracture area, the fracture feature quantitative value is obtained based on the number and distribution density of key points.

[0059] The morphological analysis module optimizes and selects target areas based on the quantitative values ​​of fracture features, analyzes bone morphological changes and fracture edge fluctuations within five consecutive frames, compares the change amplitude of the fracture edge in each frame, and screens abnormal deviation points in morphological fitting to obtain the fracture morphological deviation characteristics;

[0060] The risk assessment module determines the overlap with the current interval of the bone density monitoring unit based on the fracture morphological deviation characteristics, selects the time period in the predicted trajectory that exceeds the current bone density interval, compares the difference between the bone density change and the expansion trend of the fracture area, and obtains the fracture risk assessment parameters;

[0061] The auxiliary decision-making module adjusts the diagnostic suggestion output based on the fracture risk assessment parameters, analyzes the change range of the target suggestion in the current frame, selects the target suggestion with the optimal change range, executes the output, and collects the expansion range of the fracture area within the output cycle to obtain a fracture imaging diagnostic suggestion feedback report.

[0062] The initial characteristic distribution of the fracture includes the regional grayscale mean, gradient change amplitude, and boundary continuity index. The fracture characteristic quantification values ​​include the symmetry error value, fracture morphology ratio, and key point density. The fracture morphology offset characteristics include the edge displacement amplitude, morphology fluctuation frequency, and fitting abnormal point distribution. The fracture risk assessment parameters include the bone density overlap coefficient, risk period range, and expansion trend deviation value. The fracture imaging diagnosis recommendation feedback report includes the recommended adjustment amplitude, regional expansion strength, and output stability index.

[0063] Specifically, if Figure 2 、 3 As shown, the image parsing module includes:

[0064] The grayscale analysis submodule analyzes the grayscale distribution characteristics of the bone area in three consecutive frames of X-ray images collected by medical imaging equipment, compares the time distribution characteristics of the grayscale curve in each frame, determines whether the sampling segment is continuous and representative, and generates a periodic sampling grayscale sequence;

[0065] The bone regions are extracted from three consecutive image frames. Grayscale analysis techniques are used to obtain the grayscale value distribution of the bone region within each frame. The grayscale value distribution reflects the different densities and tissue characteristics of the bone structure. The temporal sequence characteristics of the grayscale values ​​are then analyzed to determine their changing trends. Specifically, by setting a grayscale value range for each frame, such as a standard grayscale range of 0 to 255, image processing algorithms such as edge detection or region segmentation are used to extract the grayscale values ​​of the bone regions. For example, if the grayscale values ​​of the bone region in one image range from 0 to 120, and in the next frame change to between 50 and 130, the standard deviation of the grayscale distribution for each frame can be calculated to assess its volatility. Based on the grayscale characteristics, the changing trends of the grayscale curves of each frame are further compared, such as using dynamic time warping (DTW) or correlation coefficient analysis to evaluate the temporal distribution characteristics of the grayscale curves. This allows the continuity and representativeness of each sampling segment to be determined. By comparing the grayscale changes of each frame, a periodic sampling grayscale sequence is generated for subsequent analysis.

[0066] The gradient analysis submodule calculates the rate of change between the end and start grayscales in the grayscale sequence between adjacent frames based on the periodic sampling grayscale sequence, compares the amplitude difference of the rate change, identifies the data combination with obvious fracture characteristics, and obtains the grayscale gradient interval;

[0067] Further analyze the grayscale change rate between each frame image and calculate the grayscale gradient between adjacent frames. Specifically, first extract the end and starting grayscale values ​​of adjacent frame images through sampling points, and calculate the change rate by comparing the difference in grayscale values. If the difference in grayscale value changes between adjacent frames is greater than a preset threshold (for example, the grayscale change exceeds 50), it is judged that the grayscale gradient change is significant. In practical applications, if the grayscale value of image A is 90 and the grayscale value of image B is 150, the change rate is calculated as (150-90) / time difference. By comparing the difference in the rate change amplitude, data with obvious fracture features can be identified. For example, the grayscale value changes in certain fracture areas are more abrupt and the rate difference is large, so they can be marked as potential fracture feature data combinations. In this way, grayscale gradient interval data is generated.

[0068] The boundary consistency judgment submodule calls the grayscale gradient interval to determine whether the continuity of the fracture area boundary between adjacent frames remains consistent, and selects the segments with boundary consistency and fracture amplitude to obtain the initial fracture feature distribution;

[0069] The continuity of the fracture region boundaries between adjacent frames is analyzed. By comparing the boundaries of the grayscale gradient intervals of adjacent frames, an image matching algorithm is used to detect boundary similarity. If the boundary displacement between adjacent frames is small and meets the preset boundary variation range (for example, the boundary displacement is less than 5 pixels), the boundary continuity of the fracture region is considered to be consistent. By screening out regions that meet boundary consistency and have a large fracture amplitude, the initial fracture feature distribution is finally obtained. In practice, the boundary continuity judgment standard is determined by setting a threshold, such as by measuring the grayscale difference of the boundary position or the similarity of morphological features, and then screening and demarcating the fracture regions that meet the conditions to obtain reliable fracture feature data.

[0070] Specifically, if Figure 2 、 4 As shown, the pattern recognition module includes:

[0071] The data screening submodule analyzes the bone key point data at the corresponding moment based on the initial fracture feature distribution, screens the bone node monitoring results in the same frame, determines whether the data segments are complete and continuous, and arranges them in spatial order to obtain the bone key point distribution sequence;

[0072] The skeleton key point data at the corresponding moment is extracted and analyzed, and then the skeleton node monitoring results within the same frame are screened. Through image processing technology, key nodes in the skeleton area, such as joint positions or fracture points, are identified. The key point data is obtained through three-dimensional image reconstruction or coordinate extraction of two-dimensional images. For example, in one frame of image, there is a key point of a joint with coordinates (150, 250). In the next frame of image, the joint will change due to movement or displacement. By judging whether the skeleton node data in each frame is complete and continuous, this is mainly calculated by comparing the node positions of adjacent frames and evaluating the corresponding distance between nodes. If the distance between two nodes changes little, it can be determined that the node data is continuous. Otherwise, the node is considered to be incomplete or not continuous. By performing a series of judgments on all nodes, the skeleton nodes that are continuous within the same frame are screened out and finally arranged in spatial order to obtain the skeleton key point distribution sequence. Suppose the coordinates of a certain skeleton key point are (150, 250), and in the next frame image, its coordinates become (152, 253). By calculating the distance between the two points, if it is less than a set threshold (such as 5 pixels), the two points are considered to be continuous and belong to the same data segment.

[0073] The position comparison submodule calls the skeleton key point distribution sequence, compares the spatial position relationship of each group of skeleton key points, analyzes the corresponding distances between the difference nodes, determines whether the distances fall within the reference range, selects the nodes that meet the requirements, and obtains the skeleton key point distribution characteristics;

[0074] Compare the spatial position relationship of each group of skeletal key points. In this process, first extract the position data of each group of skeletal key points, and analyze the spatial relationship between the key points by calculating the corresponding distance. If there are multiple nodes in the skeleton data, it is necessary to determine whether it conforms to a certain spatial distribution law by calculating the relative position difference. For example, if the relative position difference between two groups of nodes (such as joints and fracture points) in two consecutive frames of images is large (for example, the distance between the nodes is greater than 10 pixels), it is necessary to determine whether it exceeds the set reference range. Specifically, by calculating the actual distance between the nodes, determine whether the distance falls within the set reference range (such as 3 to 8 pixels). If the node distance exceeds this range, it is considered that the node data does not meet the continuity requirements and is screened out. Through the screening process, the distribution characteristics of skeletal key points that conform to the spatial position relationship are obtained.

[0075] The fracture area quantification submodule analyzes the proportional relationship between the length and width of the node fracture area based on the distribution characteristics of the key points of the skeleton, calculates the morphological characteristic value of the fracture area, determines the number of nodes corresponding to the fracture area, integrates the topological relationship of the skeleton network, and obtains the fracture feature quantification value;

[0076] Analyze the proportional relationship between the length and width of the node fracture area. The execution of the process first extracts the coordinate information of the nodes in the fracture area, and then calculates the distance between the nodes. Specifically, by calculating the ratio of the length and width between each group of nodes, it is determined whether it meets the fracture characteristic standards. If the length of a fracture area is 200 pixels and the width is 50 pixels, then the ratio of its length to width is 4:1. By counting the number of nodes that meet the ratio, it is analyzed whether it meets the set standard threshold. If the number of nodes exceeds 30, it can be determined as a larger fracture area. By integrating the topological relationship of the bone network, the connection method of the bone nodes is further analyzed to determine the specific location and range of the fracture area. Assuming that there are 20 nodes in the fracture area and the length-to-width ratio is 3:1, the characteristic quantitative value of the fracture is constructed through the position and relationship of the nodes, and finally the quantitative data of the fracture area is output;

[0077] The morphological characteristic value of the fracture area is calculated using the formula:

[0078]

[0079] Among them, R represents the morphological characteristic value of the fracture area, L i represents the length of the fracture area of ​​the i-th node, Represents the arithmetic mean of the length of the fracture area, W i represents the width of the fracture area of ​​the i-th node, represents the arithmetic mean of the width of the fracture area, n represents the number of nodes in the fracture area being analyzed, and i represents the node number in the summation symbol;

[0080] The fracture region morphological characteristic value indicates the overall degree of deviation in the proportional relationship between the length and width of each fracture region in the bone structure. By normalizing the length and width differences of all fracture regions, calculating the weighted difference, and measuring the morphological differences, it comprehensively reflects the variation in their geometric morphology. The larger the value, the greater the imbalance in the length and width distribution of the fracture region, and the more complex or abnormal the morphological structure. Conversely, the smaller the value, the more regular the morphology of the fracture region and the more stable the proportional relationship. This characteristic value serves as an important basis for quantitative assessment of fracture structure and can help determine the complexity of the fracture region and potential structural risks.

[0081] This formula is used to calculate the fracture area morphological characteristic value R bw , in order to evaluate the proportional relationship between the length and width of the bone fracture area. The following is the method of obtaining each parameter and the specific value:

[0082] L i : The length of the i-th fracture area is measured by image processing technology (such as skeleton extraction and edge detection). According to the actual measurement data, the values ​​are: L1 = 12.5 mm, L2 = 13.0 mm, L3 = 11.8 mm, L4 = 12.2 mm, L5 = 12.7 mm;

[0083] W i : The width of the i-th fracture area is also measured by image processing technology. According to the actual measurement data, the values ​​are: W1 = 2.5 mm, W2 = 2.7 mm, W3 = 2.4 mm, W4 = 2.6 mm, W5 = 2.5 mm;

[0084] The average length of all fractured regions is calculated as follows:

[0085]

[0086] The average value of all fracture zone widths is calculated as follows:

[0087]

[0088] n: the total number of fracture areas, here 5;

[0089] Substitute the above values ​​into the formula and calculate item by item:

[0090] Item 1:

[0091]

[0092] Item 2:

[0093]

[0094] Item 3:

[0095]

[0096] Item 4:

[0097]

[0098] Item 5:

[0099]

[0100] Add all the terms together and average them:

[0101] The results show that the morphological characteristic value R of the fracture area is 3.1459, which reflects the average deviation of the proportional relationship between the length and width of the bone fracture area. This value can be used to further analyze the number of corresponding nodes in the fracture area, integrate the topological relationship of the bone network, and obtain the quantitative value of the fracture characteristics.

[0102] Specifically, if Figure 2 、 5 As shown, the morphological analysis module includes:

[0103] The target screening submodule selects the target area with associated characteristics based on the fracture feature quantization value, organizes the bone morphology and fracture edge data of the target area within five consecutive frames, completes the data collection in frame order, and obtains the frame target area data set;

[0104] According to the quantitative value, the target area related to the specific fracture characteristics is screened out. The target area is defined by the characteristics of the bone and fracture area. By judging the quantitative value of the fracture characteristics, the data area that can continuously display the same characteristics within five frames is screened out. The data area should have consistent bone morphology and fracture edges in five consecutive frames. For example, if the bone morphology of a certain target area has a certain degree of similarity in five consecutive frames (such as the morphological change is less than 5%), then the target area will be selected as one of the target areas. The bone morphology and fracture edge data of the target area in five consecutive frames are sorted, and the data are collected in frame order through time series sorting. By integrating the data, a frame target area data set is obtained. For example, assuming that in each frame, the edge and morphological changes of the fracture can be expressed by the movement of the key point coordinates, if the position of a key point does not change much in five consecutive frames, then the area can be successfully screened as the target area.

[0105] The morphological analysis submodule analyzes the correspondence between the bone morphological changes and the fracture edge fluctuations in each frame based on the frame target area dataset, compares the synchronous changes of the fracture edge in each frame, and classifies the fluctuation states in the different time periods to obtain the morphological correspondence distribution;

[0106] The analysis is based on the relationship between the changes in bone morphology and the fluctuations of the fracture edge in each frame of the data set. In the specific operation process, the bone morphology in each frame is first compared with the position change of the fracture edge, and the amplitude of the change is calculated. By comparing the displacement of the fracture edge in consecutive frames, the synchronous change is determined. If the change in the fracture edge between adjacent frames is less than the set threshold (for example, the edge position change is less than 5 pixels), it is considered a synchronous change. Otherwise, it is considered to have abnormal fluctuations. The changes are classified. For example, the period of small change can be classified as "normal fluctuation" and the period of large change can be classified as "abnormal fluctuation". The classified fluctuation states are further counted to obtain the morphological correspondence distribution. For example, if the fracture edge changes from (100, 150) pixels to (102, 152) pixels in a frame of image, the fluctuation amplitude of the frame is considered small and belongs to normal fluctuation. Otherwise, it is abnormal fluctuation.

[0107] The outlier screening submodule judges the frame data in the morphological correspondence distribution, screens outliers in the frames with deviations in the fluctuation amplitude in the morphological fitting analysis, and identifies their positions and change states in the monitoring sequence to obtain the fracture morphological deviation characteristics;

[0108] It is further determined whether the fluctuation amplitude in each frame of data deviates from the normal fluctuation range of the morphological fitting analysis. Specifically, in the morphological fitting analysis of each frame, the change curve of the fracture edge and the bone morphology is calculated and compared with the set benchmark fluctuation range. If the fluctuation amplitude of a certain frame exceeds this range (for example, the change of the fracture edge is greater than 10 pixels), the frame will be marked as an abnormal point, the specific position and change state of the abnormal point in the monitoring sequence will be identified, and the corresponding bone morphological offset characteristics will be analyzed. For example, if the displacement of the fracture edge is greater than 10 pixels in a certain frame, and the fluctuation is smaller in several consecutive frames, the abnormal point of the frame will be marked, and its position and change state will be recorded. By screening the abnormal points, the fracture morphological offset characteristics can be obtained.

[0109] Specifically, if Figure 2 、 6 As shown, the risk assessment module includes:

[0110] The interval judgment submodule judges the coverage of the offset characteristics within the current interval of the bone density monitoring unit based on the fracture morphology offset characteristics, selects overlapping bone density change segments, and integrates the corresponding time series according to the segment order to obtain the interval coverage sequence;

[0111] By identifying fracture edges in medical images and acquiring fracture morphological features, the image analysis module extracts fracture feature values ​​from each frame by comparing grayscale changes in consecutive frames with the morphology of the fracture edge. By calculating the timeline trend of the feature values, the fracture offset characteristics are defined. These characteristics are used to determine the interval coverage of the bone density monitoring unit. For bone density changes within an interval, the difference between the bone density values ​​of two adjacent frames is calculated. If the difference exceeds a set threshold, a significant change is considered, and the change is considered an overlapping bone density change segment. To determine the segment coverage of bone density changes, the bone density data within each segment is compared with the data of the previous and next frames to ensure data continuity and stability. Finally, the data is integrated into a coverage sequence for further trajectory screening. For example, if the bone density change range in the current interval is [0.7, 0.8], if the bone density change range in the next frame is [0.75, 0.85], the two intervals are considered overlapping. The overlapping segments are recorded in the interval coverage sequence, providing data support for subsequent analysis.

[0112] The trajectory screening submodule screens the time periods where the predicted trajectory continuously exceeds the current bone density interval based on the interval coverage sequence, analyzes the bone density changes and fracture area expansion trends within the segment, summarizes the change characteristics of the difference segment, and obtains the trajectory offset segment;

[0113] The predicted trajectory's bone density changes are evaluated to determine whether any trajectory exceeds the predetermined range within the current bone density interval. The screening process first identifies any instances of exceeding the interval by setting a time period threshold. For example, the time period threshold is set to 5 minutes. If the bone density change value of a predicted trajectory exceeds the current interval within 5 minutes, it will be marked as abnormal. Based on the trend of the bone density data, the predicted trajectory's exceeding portion (i.e., the time period continuously exceeding the current bone density interval) is calculated. The bone density change trend within this time period is further analyzed to determine whether there is a clear correlation with the expansion trend of the fracture area. For example, if the data exceeds the interval by 0.2 and this trend conforms to a certain expansion pattern, it can be inferred that the fracture area has expanded. To illustrate, suppose the bone density change exceeding the interval is [0.7, 0.9] from the 10th to the 15th minute, while the previous predicted interval is [0.75, 0.85]. This period is marked as a trajectory deviation segment, and the change characteristics within this time period are summarized. This summary provides a basis for the subsequent generation of risk parameters.

[0114] The risk parameter generation submodule compares the difference between bone density changes in the trajectory offset segment and the fracture area expansion trend, determines the correspondence between bone density changes and fracture expansion trends in the key segment, calculates the correspondence index value, and integrates the segment fluctuation data to obtain the fracture risk assessment parameter;

[0115] Compare the differences in bone density changes in different time periods to determine which segments have significant changes. Based on the preset difference threshold, set at 0.05, if the bone density change value of a segment differs from that of the other segments by more than 0.05, it is considered to be significantly different. The segments with significant differences will be marked, and the correlation between the bone density changes and the fracture expansion trend in each segment will be analyzed through induction. For example, if the bone density of a segment changes from 0.75 to 0.85, and the expansion rate of the fracture area is proportional to this change, the system will record this data and further optimize the risk assessment. The bone density fluctuation data in each segment will be integrated to form a fracture risk assessment parameter, which will be further used for fracture risk assessment, and finally the fracture risk assessment value will be obtained for medical decision support;

[0116] The corresponding relationship index value is calculated using the formula:

[0117]

[0118] T Δ Represents the corresponding relationship index value, B k represents the bone density value measured along the offset trajectory in the kth sub-segment, F k represents the extended length of the corresponding fracture region in the kth sub-segment, represents the average value of the fracture extension length of the neutron segment in the key segment θ, represents the average value of the bone density value of the sub-segment in the key segment φ, ΔP θ represents the maximum and minimum difference between the bone density gradient in the key segment θ, and m represents the number of sub-segments in the key segment;

[0119] The correspondence index value indicates the degree of synchronization or synergy between bone density changes and fracture propagation trends within the key segment. A larger value indicates that the amplitude of bone density changes within the segment is consistent with the expansion trend of the fracture area, that is, the deterioration of bone structure is more likely to trigger or aggravate fracture propagation behavior. When the value is small, it indicates that there is no obvious correlation between bone density changes and fracture propagation, and the trends of the two are not highly correlated. Therefore, when assessing fracture risk, the index value can be used to determine whether there is a high degree of coupling between mechanical degradation and tissue fragility in key areas.

[0120] In the key section, the corresponding relationship index value T between bone density change and fracture extension trend Δ Calculated by the above formula;

[0121] B k : Bone density value in the kth sub-segment, in g / cm 2 , obtained by dual-energy X-ray absorptiometry (DXA) measurement;

[0122] F k : the extended length of the fracture area in the kth sub-segment, in mm, obtained by real-time measurement of load-displacement data combined with the equivalent compliance calculation method;

[0123] The average value of the fracture extension length of all sub-segments in the key segment θ, in mm, is calculated as follows:

[0124] The average value of bone density of all sub-segments in the key segment φ, in g / cm 2 , the calculation formula is:

[0125] ΔP θ : The maximum and minimum difference between the bone density gradient in the key segment θ, in g / cm 2 , the calculation formula is: ΔP θ =max(B k )-min(B k );

[0126] m: the number of sub-segments within the key segment;

[0127] In order to unify the dimensions, the bone density value and fracture extension length need to be normalized.

[0128] The normalization method is as follows:

[0129] Calculate the bone density gradient:

[0130] ΔP θ =max(B k )-min(B k )=0.95-0.75=0.20g / cm 2 ;

[0131] Calculate the molecular part:

[0132]

[0133] Calculate each term:

[0134] Item 1:

[0135] Item 2:

[0136] Item 3:

[0137] Item 4:

[0138] Item 5:

[0139] Total: 1.704 + 1.345 + 0.850 + 1.625 + 1.249 = 6.773;

[0140] Molecular part: 6.773-0.85=5.923;

[0141] Calculate the denominator:

[0142] Final calculation T Δ :

[0143] The results show that the correspondence index between bone density changes and fracture extension trends in the key segment is 1.087, and the higher the value, the stronger the correspondence between the two.

[0144] Specifically, if Figure 2 、 7 As shown, the auxiliary decision module includes:

[0145] The suggestion adjustment submodule determines the relationship between the current diagnostic suggestion and the risk characteristics based on the fracture risk assessment parameters, optimizes the suggestion adjustment configuration, adjusts the available suggestions according to the frame data, filters the outputtable target suggestions, and obtains the suggestion change range;

[0146] Based on the fracture risk assessment parameters, the relationship between the current diagnostic recommendation and the risk profile is calculated to assess whether the recommendation aligns with the actual trend in fracture risk. This process involves a comprehensive analysis of multiple factors, including bone density, fracture extension, and changes in fracture area. Bone density data is collected for each image frame and used to generate risk assessment parameters. Based on the degree of match between the assessment parameters and the current diagnostic recommendation, the algorithm determines which recommendations are valid in the current diagnostic context. An algorithm compares and analyzes each data point. If the risk assessment result for a particular diagnostic recommendation deviates significantly from the current reality, the recommendation output is adjusted. Based on the trend of fracture area changes over the time period, irrelevant recommendation data is selectively adjusted or filtered. If the current fracture area trend indicates an increase and a higher risk, the recommendation content is optimized and adjusted to more accurately reflect the changes in the risk profile. This adjustment process is based on continuous monitoring of each image frame and data feedback, completed through data fitting and real-time updates. The magnitude of the recommendation change is determined, the target recommendation for output is marked, and the magnitude of the recommendation change is calculated.

[0147] The target comparison submodule compares the change characteristics of each target suggestion within the frame based on the suggestion change amplitude, analyzes the response performance of each target suggestion, selects the target suggestion with the best change amplitude performance, and obtains the suggestion response difference;

[0148] Based on the amplitude of change in image suggestions, the intra-frame change characteristics of different target suggestions are compared, and the response performance of each target suggestion is analyzed. The first step in this process is to collect the performance of each target suggestion in different frame data, compare the effects of different target suggestions based on the amplitude of change of the suggestions, and compare the relationship between the amplitude of change of each target suggestion and its change in indicators such as bone density and fracture area expansion, and screen out the suggestions that perform best during the change process. Assuming that within a certain time period, the amplitude of change of suggestion 1 is 0.05, and the amplitude of change of suggestion 2 is 0.02, and the corresponding bone density change trends of the two are consistent with the fracture expansion trends, suggestion 1 is preferred as the optimal target suggestion. Not only should the amplitude of the suggestions be compared, but also the response time and accuracy of each suggestion during the change process should be comprehensively analyzed. For example, if a target suggestion responds faster and has a larger response amplitude at a certain time point, then this suggestion will be considered the target suggestion with the best performance. The difference in suggestion response is obtained, which further supports subsequent diagnostic suggestions.

[0149] The output response acquisition submodule performs optimal target suggestion output based on the difference in suggestion responses, collects changes in the fracture area within the output cycle, analyzes the fracture extension interval, and obtains a fracture imaging diagnosis suggestion feedback report;

[0150] According to the difference in the response to the suggestion, the optimal target suggestion output is executed, and the changes in the fracture area within the output cycle are collected to calculate the degree of difference in each suggestion response. The suggestion with smaller difference is more in line with the actual diagnostic needs. The diagnosis result will be output according to the selected optimal target suggestion, and the changes in the fracture area will be monitored within the output cycle. In this process, the expansion of the fracture area is analyzed in real time, and the subsequent diagnostic suggestion output is adjusted based on the real-time feedback data. For example, if during the monitoring cycle of the fracture area, it is found that a certain suggestion can better predict the expansion trend of the fracture area and the changes are highly matched with the risk parameters, a fracture imaging diagnosis suggestion feedback report is obtained, and the changes in the fracture area and the expansion range are recorded in detail in the report. The feedback report not only provides real-time diagnostic basis for medical staff, but also can continuously optimize the subsequent diagnostic suggestion output process through data accumulation.

[0151] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An intelligent fracture diagnosis system based on image recognition, characterized in that: The system comprises: The image analysis module analyzes the grayscale distribution of the bone area in consecutive frames based on X-ray images collected by medical imaging equipment, compares the grayscale gradient changes between adjacent frames, and determines the boundary consistency of the fracture area to obtain the initial characteristic distribution of the fracture; The pattern recognition module screens the bone key point data at the corresponding moment based on the initial fracture feature distribution, compares the corresponding position relationship of the key points in space, determines whether it meets the anatomical symmetry standard, and obtains the fracture feature quantification value based on the ratio of the length and width of the fracture area and the number and distribution density of the key points; The morphological analysis module optimizes and selects the target area based on the fracture feature quantization value, analyzes the bone morphological changes and fracture edge fluctuations within five consecutive frames, compares the change amplitude of the fracture edge in each frame, and selects abnormal deviation points in the morphological fitting to obtain the fracture morphological deviation characteristics; Based on the fracture morphological deviation characteristics, the risk assessment module determines the overlap with the current interval of the bone density monitoring unit, screens the time period in the predicted trajectory that exceeds the current bone density interval, compares the difference between the bone density change and the expansion trend of the fracture area, and obtains the fracture risk assessment parameters.

2. The intelligent fracture diagnosis system based on image recognition according to claim 1, characterized in that: The initial characteristic distribution of the fracture includes the regional grayscale mean, gradient change amplitude, and boundary continuity index; the fracture characteristic quantification value includes the symmetry error value, fracture morphology ratio, and key point density; the fracture morphology offset characteristics include the edge displacement amplitude, morphology fluctuation frequency, and fitting abnormal point distribution; the fracture risk assessment parameters include the bone density overlap coefficient, risk period range, and expansion trend deviation value.

3. The intelligent fracture diagnosis system based on image recognition according to claim 1, characterized in that: The image analysis module includes: The grayscale analysis submodule analyzes the grayscale distribution characteristics of the bone area in three consecutive frames of X-ray images collected by medical imaging equipment, compares the time distribution characteristics of the grayscale curve in each frame, determines whether the sampling segment is continuous and representative, and generates a periodic sampling grayscale sequence; The gradient analysis submodule calculates the rate of change between the end and the start grayscale in the grayscale sequence between adjacent frames based on the periodic sampling grayscale sequence, compares the amplitude difference of the rate change, identifies the data combination with obvious break characteristics, and obtains the grayscale gradient interval; The boundary consistency determination submodule calls the grayscale gradient interval to determine whether the continuity of the fracture region boundary between adjacent frames remains consistent, and screens the segments with boundary consistency and fracture amplitude to obtain the initial fracture feature distribution.

4. The intelligent fracture diagnosis system based on image recognition according to claim 3, characterized in that: The pattern recognition module includes: The data screening submodule analyzes the bone key point data at the corresponding moment based on the initial fracture feature distribution, screens the bone node monitoring results in the same frame, determines whether the data segments are complete and continuous, and arranges them in spatial order to obtain a bone key point distribution sequence; The position comparison submodule calls the skeleton key point distribution sequence, compares the spatial position relationship of each group of skeleton key points, analyzes the corresponding distances between the difference nodes, determines whether the distances fall within the reference range, selects the nodes that meet the requirements, and obtains the skeleton key point distribution characteristics; The fracture area quantification submodule is based on the distribution characteristics of the bone key points, analyzes the proportional relationship between the length and width of the node fracture area, calculates the morphological characteristic value of the fracture area, determines the number of nodes corresponding to the fracture area, integrates the topological relationship of the bone network, and obtains the fracture feature quantification value.

5. The intelligent fracture diagnosis system based on image recognition according to claim 4 is characterized in that: The morphological characteristic value of the fracture area is calculated using the formula: Among them, R represents the morphological characteristic value of the fracture area, L i represents the length of the fracture area of ​​the i-th node, L represents the arithmetic mean of the fracture area length, W i represents the width of the fracture region of the i-th node, W represents the arithmetic mean of the fracture region width, n represents the number of nodes in the fracture region being analyzed, and i represents the node number in the summation symbol.

6. The intelligent fracture diagnosis system based on image recognition according to claim 4, characterized in that: The morphological analysis module includes: The target screening submodule screens the target area with associated characteristics based on the fracture feature quantization value, organizes the bone morphology and fracture edge data within five consecutive frames of the target area, completes data collection in frame order, and obtains the frame target area data set; The morphological analysis submodule analyzes the correspondence between the bone morphological changes and the fracture edge fluctuations in each frame based on the frame target area dataset, compares the synchronous changes of the fracture edge in each frame, and classifies the fluctuation states in the different time periods to obtain the morphological correspondence distribution; The abnormal point screening submodule judges the frame data in the morphological correspondence distribution, screens the frame abnormal points with deviation in the fluctuation amplitude in the morphological fitting analysis, and identifies their position and change state in the monitoring sequence to obtain the fracture morphological deviation characteristics.

7. The intelligent fracture diagnosis system based on image recognition according to claim 6, characterized in that: The risk assessment module includes: The interval judgment submodule judges the coverage of the offset characteristics within the current interval of the bone density monitoring unit based on the fracture morphology offset characteristics, selects overlapping bone density change segments, and integrates the corresponding time series according to the segment order to obtain the interval coverage sequence; The trajectory screening submodule screens the time periods where the predicted trajectory continuously exceeds the current bone density interval based on the interval coverage sequence, analyzes the bone density changes and fracture area expansion trends within the segments, summarizes the change characteristics of the difference segments, and obtains the trajectory offset segments; The risk parameter generation submodule compares the difference between the bone density change in the trajectory offset segment and the fracture area expansion trend, determines the correspondence between the bone density change and the fracture expansion trend in the key segment, calculates the correspondence index value, and integrates the segment fluctuation data to obtain the fracture risk assessment parameter.

8. The intelligent fracture diagnosis system based on image recognition according to claim 7, characterized in that: The corresponding relationship index value adopts the formula: T Δ Represents the corresponding relationship index value, B k represents the bone density value measured along the offset trajectory in the kth sub-segment, F k represents the extended length of the corresponding fracture region in the kth sub-segment, represents the average value of the fracture extension length of the neutron segment in the key segment θ, represents the average value of the bone density value of the sub-segment in the key segment φ, ΔP θ represents the maximum and minimum difference between the bone density gradient in the key segment θ, and m represents the number of sub-segments in the key segment.

9. The intelligent fracture diagnosis system based on image recognition according to claim 1, characterized in that: The system also includes auxiliary decision modules: The auxiliary decision-making module adjusts the diagnostic suggestion output based on the fracture risk assessment parameters, analyzes the change range of the target suggestion in the current frame, selects the target suggestion with the best change range, executes the output, and collects the expansion range of the fracture area within the output cycle to obtain a fracture imaging diagnosis suggestion feedback report; The fracture imaging diagnosis suggestion feedback report includes a suggested adjustment range, a regional expansion intensity, and an output stability index.

10. The intelligent fracture diagnosis system based on image recognition according to claim 9, characterized in that: The auxiliary decision module includes: The suggestion adjustment submodule determines the relationship between the current diagnostic suggestion and the risk characteristics based on the fracture risk assessment parameters, optimizes the suggestion adjustment configuration, adjusts the available suggestions according to the frame data, screens the outputtable target suggestions, and obtains the suggestion change range; The target comparison submodule compares the change characteristics of the corresponding target suggestions within the frame based on the suggestion change amplitude, analyzes the response performance of each target suggestion, selects the target suggestion with the best change amplitude performance, and obtains the suggestion response difference; The output response acquisition submodule performs optimal target suggestion output based on the suggestion response difference, collects changes in the fracture area within the output cycle, and analyzes the fracture extension interval to obtain a fracture imaging diagnosis suggestion feedback report.

Images