A CT rib fracture artificial intelligence auxiliary diagnosis system based on image recognition

Abstract

This invention discloses an AI-assisted diagnostic system for CT rib fractures based on image recognition, belonging to the field of smart healthcare technology. It includes recognizing the patient's contour using thermal imaging data, constructing a respiratory model to analyze respiratory status, and automatically triggering CT image acquisition of the ribs at the optimal inhalation moment. Subsequently, artifact detection is performed; if qualified, the system is handed over to AI for intelligent fracture diagnosis, achieving a fully automated process from image acquisition to diagnosis. This invention uses non-contact thermal imaging to monitor respiration and body movement, constructing a respiratory model that integrates chest contour and airflow temperature changes, accurately identifying the optimal inhalation moment to trigger CT acquisition, reducing motion artifacts at the source and improving image quality. The system is highly robust, automatically determining the optimal acquisition time even under conditions of clothing or weak breathing, reducing reliance on patient cooperation, and is particularly suitable for rib fracture patients, improving examination success rate and efficiency.

Description

Technical Field

[0001] This invention relates to the field of smart medical technology, specifically to an AI-assisted diagnostic system for CT rib fractures based on image recognition. Background Technology

[0002] Relying on image recognition and deep learning technologies, the AI-assisted diagnostic system can quickly process chest CT images, automatically identify and locate rib fractures, mark the fracture type, number, and degree of displacement, and simultaneously screen for complications such as pneumothorax and hemothorax, generating structured auxiliary reports, effectively improving the efficiency of rib fracture diagnosis and treatment.

[0003] For example, the deep learning-based rib fracture auxiliary detection method and image recognition method provided in Chinese Patent Publication No. CN112699869A includes selecting a certain number of chest CT images as a training set and labeling the rib fracture regions and rib numbers in the images; performing data normalization processing on the images; using the processed images as input and the labeled rib fracture regions and rib numbers as output for model training. The training model includes: a rib detection model, a rib fracture segmentation model, and a rib numbering and segmentation model; inputting the processed chest CT image to be detected into the trained rib fracture detection model and outputting the detection result. The deep learning-based rib fracture auxiliary detection method provided in this invention effectively reduces false positives and false negatives in rib fracture detection, and the detection result provides location information of suspected rib fractures, which can assist doctors in diagnosis.

[0004] However, during a CT scan, patients are usually required to lie still and hold their breath. But rib pain makes it impossible for patients to lie still and hold their breath normally. The ribs are the main bones that make up the ribcage. They expand and contract rhythmically with breathing. This movement causes the actual position of the ribs to change during the short period of time when the CT equipment rotates and acquires data. When the system tries to reconstruct a static image from these data acquired at different times with slight positional deviations, data mismatch occurs. This is ultimately manifested as striped or blurry motion artifacts on the image, which reduces image clarity and affects the accuracy of AI-assisted diagnosis. Summary of the Invention

[0005] Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides an AI-assisted diagnostic system for CT rib fractures based on image recognition. This system solves the problem that during CT scans, rib movement caused by the patient's breathing leads to inconsistencies in the collected data in time and space, ultimately resulting in motion artifacts during image reconstruction, which affect diagnostic clarity and the accuracy of AI-assisted diagnosis.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an AI-assisted diagnostic system for CT rib fractures based on image recognition, comprising the following specific modules: a patient thermal imaging acquisition module: acquiring patient thermal imaging data, including the number of pixels, pixel coordinates, and RGB color values; marking pixels according to the RGB color values ​​corresponding to human body temperature to generate a complex contour; matching the complex contour with a preset human body contour template to identify the patient contour; further matching the mouth and nose contour and chest contour; and defining the airflow temperature detection area with the center of the mouth and nose contour as the origin; a patient respiratory analysis module: setting a stationary factor by comparing the moving distance of the patient contour center point with a dynamic moving threshold; obtaining the chest change distance value by calculating the temporal difference between the chest contour feature points and the center point; and converting the average RGB color value of the airflow temperature detection area into HSL. The system calculates the temperature change value by analyzing the difference between cool and warm values ​​in the hue analysis. Based on the resting factor, chest distance, and temperature change value, a patient respiratory model is constructed. The respiratory status is determined by calculating the ratio of the patient's respiratory model at adjacent time points. The inspiratory volume stability value is obtained by integrating the patient's respiratory model during inspiration, and the time corresponding to the maximum stable inspiratory volume value is selected as the inspiration optimization time. The CT image capture module controls the CT equipment to acquire rib CT images during the inspiration optimization time. The CT image artifact analysis module extracts the rib contour from the rib CT image using an image segmentation algorithm and detects artifacts using a contour tracking algorithm. If artifacts are present, the system returns to the patient thermal imaging acquisition module; otherwise, it executes the AI ​​rib fracture diagnosis module. The AI ​​rib fracture diagnosis module uses a deep learning convolutional neural network to perform end-to-end analysis of the rib CT image, achieving automatic detection and diagnosis of rib fractures.

[0008] Furthermore, the specific method for obtaining the stillness factor is as follows: Set the stillness factor, calculate the center point coordinates of the patient contour based on the number of pixels and pixel coordinates of the patient contour, calculate the Euclidean distance between the center point coordinates of the patient contour at the next moment and the center point coordinates of the patient contour at the previous moment based on the time series, obtain the movement distance, preset a dynamic movement threshold, compare the movement distance with the dynamic movement threshold, if the movement distance is within the dynamic movement threshold, it indicates that the patient is still, and the stillness factor is assigned a value of 1; if the movement distance is outside the dynamic movement threshold, it indicates that the patient is moving, and the stillness factor is assigned a value of 0.

[0009] Furthermore, the specific method for obtaining the patient's respiratory model is as follows: Euclidean distance is calculated on both sides of the chest contour according to the time series to obtain the chest change distance value; the temperature change in the airflow temperature detection area is calculated according to the time series to obtain the temperature change value; and the patient's respiratory model is obtained by comprehensively calculating the rest factor, the chest change distance value, and the temperature change value. ;

[0010] in, This represents a patient's respiratory model. Represents the static factor. Indicates the temperature change value. This represents the distance value of the chest change.

[0011] Furthermore, the specific method for obtaining the chest change distance value is as follows: Euclidean distance is calculated based on the number of pixels and the coordinates of the pixels within the chest contour to obtain the coordinates of the center point of the chest contour. The coordinates of some edge pixels of the chest contour are set as feature point coordinates. Euclidean distance is calculated between the feature point coordinates and the center point coordinates of the chest contour to obtain the chest distance value. The difference between the chest distance value at the next moment and the chest distance value at the previous moment is calculated based on the time series to obtain the chest change distance value.

[0012] Further, the specific method for obtaining the temperature change value is as follows: The RGB color values ​​corresponding to the pixel coordinates within the airflow temperature detection area are averaged to obtain the average RGB color value. The average RGB color value is then converted to the degree of the HSL color space to obtain the conversion degree. The degree range of the HSL color space is from 0 to 360 degrees. The higher the temperature, the warmer the color corresponding to the average RGB color value, and the warmer the color, the closer the degree of the HSL color space is to 0 degrees. Conversely, the colder the temperature, the cooler the color corresponding to the average RGB color value, and the cooler the color, the closer the degree of the HSL color space is to 240 degrees. The difference between the conversion degree and 0 is calculated, and the absolute value is taken to obtain the warmer value. The difference between the conversion degree and 240 is calculated, and the absolute value is taken to obtain the cooler value. The difference between the cooler value and the warmer value is calculated to obtain the temperature change value.

[0013] Furthermore, the specific method for determining the respiratory state by calculating the ratio of the patient's respiratory models at adjacent time points is as follows: calculate the patient's respiratory model at the next moment and the patient's respiratory model at the previous moment according to the time series to obtain the respiratory ratio; compare the respiratory ratio with 1. If the respiratory ratio is greater than 1, it indicates exhalation; if the respiratory ratio is less than 1, it indicates inhalation.

[0014] Furthermore, the specific method for obtaining the respiratory ratio is as follows: ;in, Indicates the respiratory ratio, This indicates the patient's breathing model in the next moment. This represents the patient's breathing model from the previous moment. It represents a positive real number.

[0015] Furthermore, the specific method for obtaining the inspiratory optimization time is as follows: record the consecutive moments when the respiratory ratio is less than 1 to obtain the inspiratory time, where the inspiratory time is continuous. Then, perform integral calculation based on the inspiratory time and the patient's respiratory model to obtain the stable value of inspiratory volume. Sort the stable values ​​of inspiratory volume in ascending order using a quick sorting algorithm, select the largest stable value of inspiratory volume, and record the inspiratory time of the largest stable value of inspiratory volume as the inspiratory optimization time. Therefore, the initial moment when the respiratory ratio is less than 1 is recorded as the inspiratory start moment and assigned a value of 1. Starting from 1, the calculation is performed by accumulating 1 according to the time series until the result equals the inspiratory optimization time. Then, the last moment of the inspiratory optimization time is the inspiratory optimization time.

[0016] Furthermore, the specific method for obtaining the stable inspiratory volume value is as follows: ;in, This indicates the stable value of inspiratory volume. to Indicates the inhalation time. This represents a patient's breathing model.

[0017] Beneficial effects

[0018] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0019] 1. By using thermal imaging technology to monitor patients' respiratory rhythm and body movement in real time without contact, a patient respiratory model was constructed that integrates changes in chest contour and airflow temperature changes in the oral and nasal regions. This model can accurately identify the optimal inspiratory moment during the inspiratory phase, and trigger CT equipment to acquire images based on this. This effectively avoids rib displacement caused by respiratory movements, significantly reducing motion artifacts in CT images from the source, thus obtaining clearer and higher-quality rib CT images, laying a reliable image foundation for subsequent AI-based accurate diagnosis.

[0020] 2. By comprehensively analyzing multi-dimensional information such as the patient's contour static factors, chest shape changes, and respiratory airflow temperature changes, the respiratory model judgment has higher robustness and fault tolerance. Even when the patient's clothing is thick, resulting in insignificant changes in chest contour, or when breathing is weak, resulting in subtle changes in airflow temperature, the system can still reliably analyze through complementary parameters and automatically determine the optimal time for image acquisition. This process is fully automated, reducing reliance on the patient's active cooperation. It is particularly suitable for rib fracture patients who are unable to follow traditional breath-holding instructions due to pain, thus improving the success rate and efficiency of the examination.

[0021] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0022] Figure 1 This invention provides a flowchart of an AI-assisted diagnostic method for CT rib fractures based on image recognition.

[0023] Figure 2 This invention relates to a structural diagram of an AI-assisted diagnostic system for CT rib fractures based on image recognition. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0026] Example 1:

[0027] like Figures 1-2 As shown, this embodiment of the invention provides an AI-assisted diagnostic system for CT rib fractures based on image recognition, comprising the following specific modules:

[0028] Patient thermal imaging acquisition module: A thermal imager is set up in the CT imaging room to acquire thermal imaging data of the patient, including the number of pixels, pixel coordinates, and RGB color values. The thermal imaging data can analyze not only the patient's movement but also their breathing. After noise reduction processing, the thermal imaging data uses cool colors to represent low temperatures and warm colors to represent high temperatures, that is, adjusting the capture temperature of the thermal imager to correspond with the RGB color values. Since the body temperature of the patient lying on the CT scanner is different from the ambient temperature, the color corresponding to the patient's body temperature is also different from the color corresponding to the ambient temperature. The pixel coordinates of the RGB color values ​​corresponding to the human body temperature are marked to obtain the human body coordinates. The human body coordinates in a single frame of thermal imaging data are traversed, and adjacent human body coordinates are recorded until the adjacent pixel coordinates are not human body coordinates, at which point recording stops. In order to filter out the contours composed of pixels with RGB color values ​​corresponding to the human body temperature, a miscellaneous contour is obtained, that is, the human body contour and the contours of other objects with the same temperature as the human body are present in a single frame of thermal imaging data.

[0029] Pre-set human body contour templates, including those for the nose and mouth, and chest, are used. The Hu moment algorithm matches the various contours against these templates. The core of the Hu moment algorithm is to summarize the contour shape into multiple key values ​​using mathematical moments. These values ​​are invariant to translation, rotation, and scale. The algorithm quantifies the shape similarity of two contours by directly calculating the difference between their corresponding moment vectors, such as the Euclidean distance. Smaller distances indicate higher matching accuracy. The patient's contour is then selected, and the Hu moment algorithm is used to match it against both the nose and mouth and chest contour templates. The patient's nose and mouth contours are then selected. A pre-set airflow temperature detection area is then used; after selecting the nose and mouth contours, the coordinates of their center point are calculated. The center point serves as the origin, dividing the area into circular regions. The temperature detected within this region represents the ambient air temperature after the patient breathes through their mouth or nose. This ambient air temperature then indicates whether the patient is inhaling or exhaling. This is because human respiration involves continuous heat exchange with the surrounding environment. During exhalation, the warm air heated by the body releases heat into the surrounding air, slightly raising its local temperature. During inhalation, the body needs to inhale cool air from the environment and heat it to body temperature. This process removes heat from the surrounding environment, slightly lowering its local temperature. In other words, human respiration actually creates a tiny, dynamically alternating cycle of warm and cold air near the nasal cavity or oral cavity. In thermal imaging, temperature changes in this area will cause changes in the corresponding RGB color values.

[0030] Patient Respiratory Analysis Module: Set the stillness factor and analyze the stillness of the patient's outline. Here, stillness refers to relative stillness, allowing for slight errors, that is, allowing slight movement of the patient. If the analysis shows that the patient is still, the stillness factor is assigned a value of 1; otherwise, it is assigned a value of 0.

[0031] The system comprehensively analyzes the changes in chest contour size and airflow temperature in the detection area based on the static factor, and performs standardization to eliminate dimensional differences and convert values ​​of different orders of magnitude into a unified numerical range to obtain a patient breathing model. The patient's breathing is then analyzed based on this model. If exhalation is detected, the analysis continues until inhalation is detected. If inhalation is detected, the continuity and volume of inhalation are comprehensively analyzed and standardized. This ensures that the patient inhales the maximum amount of air, allowing the lungs to fully expand and maximizing the air volume in the alveoli. This results in extremely high image contrast with structures such as ribs, soft tissues, and blood vessels, helping to more clearly reveal rib contours and minute fracture lines. The optimal inhalation time is then determined, and the CT image capture module is executed.

[0032] CT image capture module: During the inspiration optimization moment, the CT instrument acquires CT images of the patient's ribs. Because the CT instrument emits radiation when acquiring CT images of the ribs, we try to ensure successful acquisition in one go and use noise reduction processing to make the CT images of the ribs clear.

[0033] CT Image Artifact Analysis Module: Since patients may continue to inhale during the rib CT image acquisition time of the inspiration optimization phase, rib ghosting may occur in the rib CT image. Therefore, the rib contour is segmented using image contour segmentation and tracking algorithms, such as the Sobel algorithm. First, the rib CT image is converted to grayscale to simplify calculations. Then, it is convolved with 3×3 kernels in the horizontal and vertical directions to obtain the horizontal and vertical gradient matrices. Next, the gradient magnitude and direction are calculated; the former determines the edge, and the latter determines the edge extension direction. Finally, threshold binarization is used to mark pixels exceeding the threshold. The rib contour is obtained by using the Lucas-Kanade optical flow method, which establishes optical flow constraint equations based on the assumption of constant brightness between adjacent frames. The spatial and temporal gradients are calculated using the Sobel algorithm and frame difference. An overdetermined set of equations is constructed assuming consistent neighborhood motion, and the optical flow velocity is solved using the least squares method. Finally, tracking and position updates are performed to track the rib contour. Then, the shape of the rib contour is matched sequentially using the Hu moment algorithm. If two or more identical rib contours are matched, the process is returned to the patient thermal imaging acquisition module; otherwise, the AI ​​rib fracture diagnosis module is executed.

[0034] AI rib fracture diagnosis module: Utilizing deep learning convolutional neural networks, the system performs end-to-end analysis of rib CT images. The AI ​​model first accurately locates and segments the structure of each rib using a target detection algorithm. Then, it simultaneously extracts multi-dimensional features on high-resolution images to identify subtle abnormal patterns such as bone continuity interruptions, cortical bone distortion, and fracture lines. Finally, the system performs intelligent classification and severity assessment based on the learned fracture features and then terminates the process.

[0035] Example 2 differs from Example 1 in that:

[0036] The specific method for analyzing the static state of the patient's contours is as follows:

[0037] Based on the number and coordinates of pixels in the patient's outline, the coordinates of the center point of the patient's outline are calculated. According to the time series, the Euclidean distance between the center point coordinates of the patient's outline at the next moment and the center point coordinates of the patient's outline at the previous moment is calculated to obtain the movement distance. A preset dynamic movement threshold is set, which is in the form of an interval. Within the dynamic movement threshold, the movement distance is allowed to be greater than zero. The movement distance is compared with the dynamic movement threshold. If the movement distance is within the dynamic movement threshold, it means that the patient is stationary. If the movement distance is outside the dynamic movement threshold, it means that the patient has moved.

[0038] The specific method for obtaining the patient's respiratory model is as follows:

[0039] Since the chest contour enlarges during inhalation and shrinks during exhalation, the Euclidean distance between the two sides of the chest contour is calculated based on the time series to obtain the chest change distance value. Furthermore, the temperature of the airflow temperature detection area decreases during inhalation and increases during exhalation. Therefore, the temperature change of the airflow temperature detection area is calculated based on the time series to obtain the temperature change value. The size change of the chest contour is synchronous with the temperature change of the airflow temperature detection area. Therefore, the patient's breathing model is obtained by comprehensively calculating based on the resting factor, chest change distance value, and temperature change value, and then standardizing the data.

[0040] ;

[0041] in, This represents a patient breathing model, reflecting whether the patient is inhaling or exhaling. Subsequent acquisition of rib CT images during inspiration... Let represent the stationary factor. If the stationary factor is 1, then... , Indicates the temperature change value. This represents the distance change in chest temperature. When the temperature change is greater than zero, it indicates the patient is exhaling because the temperature in the airflow temperature detection area gradually increases. Simultaneously, the distance change in chest temperature is less than zero because exhalation causes the patient's chest contour to shrink. To ensure that the temperature change is always greater than zero, To ensure that the distance change value of the chest is always greater than zero, if the temperature change value is greater than zero, then... Increasing incrementally, while simultaneously, the distance value of the chest change is less than zero, then... Decreasing, that is Incrementing, if the temperature change is less than zero, then Decreasing, while at the same time, the distance value of the chest change is greater than zero, then Increasing, that is Decreasing;

[0042] Additionally, if the patient is wearing thick clothing, making changes in the chest area less noticeable (i.e., the distance of chest change is zero), then... This means that the patient's breathing can only be detected by changes in chest contour; if the temperature change is greater than zero, then... Incrementing, if the temperature change is less than zero, then Decreasing, when the patient's airflow is weak and the temperature change in the airflow temperature detection area is not significant, i.e., the temperature change value is equal to 0, then... If the distance value of the chest change is greater than zero, then Decreasing; if the distance of change in the chest area is less than zero, then... Incremental analysis combines chest distance changes with temperature changes to improve the accuracy of the patient respiratory model. However, even when either chest distance change or temperature change is abnormal (i.e., 0), it does not affect the patient respiratory model, thus improving its stability and robustness. If the quiescence factor is 0, then... This indicates that the patient's respiratory model has no monotonicity.

[0043] The specific method for obtaining the chest change distance value is as follows:

[0044] The center point coordinates of the chest contour are obtained by calculating the Euclidean distance based on the number and coordinates of pixels within the chest contour. The coordinates of some edge pixels of the chest contour are set as feature point coordinates. The Euclidean distance between the feature point coordinates and the center point coordinates of the chest contour is calculated to obtain the chest distance value. The difference between the chest distance value at the next moment and the chest distance value at the previous moment is calculated based on the time series to obtain the chest change distance value. If the chest change distance value is greater than zero, it means that the chest distance value at the next moment is greater than the chest distance value at the previous moment. That is, the distance between the feature point coordinates at the next moment and the center point coordinates of the chest contour is greater than the distance between the feature point coordinates at the previous moment and the center point coordinates of the chest contour, which indicates that the chest contour has become larger, that is, the patient is inhaling. Conversely, the patient is exhaling.

[0045] The specific method for obtaining temperature change values ​​is as follows:

[0046] The average RGB color value is calculated by averaging the RGB color values ​​corresponding to the pixel coordinates within the airflow temperature detection area. This average RGB color value is then converted to HSL color space degrees to obtain the conversion degree. Since RGB color values ​​include red, green, and blue channel values, which are inconvenient to calculate, the HSL color space degree becomes a single value. The HSL color space degree ranges from 0 to 360 degrees. The higher the temperature, the warmer the color corresponding to the average RGB color value, and the warmer the color, the closer the degree in the HSL color space is to 0 degrees. Conversely, the cooler the temperature, the cooler the color corresponding to the average RGB color value, and the cooler the color, the closer the degree in the HSL color space is to 240 degrees. The difference between the conversion degree and 0 is calculated, and the absolute value is taken to obtain the warmer value. The difference between the conversion degree and 240 is calculated, and the absolute value is taken to obtain the cooler value. The difference between the cooler value and the warmer value is calculated to obtain the temperature change value.

[0047] ;

[0048] in, Indicates the temperature change value. Indicates a cool value. This indicates the warming value. When the conversion degree is close to 0, the warming value is less than the cooling value, meaning the temperature change is greater than zero. When the conversion degree is close to 240, the warming value is greater than the cooling value, meaning the temperature change is less than zero.

[0049] The specific method for analyzing a patient's respiratory status based on a patient respiratory model is as follows:

[0050] The respiratory ratio is calculated by comparing the patient's respiratory model at the next moment with that at the previous moment based on the time series.

[0051] The respiratory ratio is compared with 1. If the respiratory ratio is greater than 1, that is, the value of the patient's respiratory model is increasing, it indicates expiration. If the respiratory ratio is less than 1, that is, the value of the patient's respiratory model is decreasing, it indicates inhalation.

[0052] The specific method for obtaining the respiratory ratio is as follows:

[0053] ;

[0054] in, Indicates the respiratory ratio, This indicates the patient's breathing model in the next moment. This represents the patient's breathing model from the previous moment. Represent positive real numbers, avoid It is 0 and does not affect the ratio of the patient's respiratory model in the next moment to that in the previous moment.

[0055] The specific method for obtaining the inhalation optimization timing is as follows:

[0056] Because normal inhalation is continuous, and when a patient has a rib fracture, the expansion of the chest cavity can cause the broken rib ends to rub against each other or move, thereby stimulating the abundant pain nerves distributed on the periosteum. This can potentially trigger severe pain, forcing the patient to interrupt inhalation or causing uncontrolled body movements during inhalation. Consequently, rib CT images cannot be acquired at these times. Therefore, we record consecutive moments when the respiratory ratio is less than 1 to obtain the inhalation time. Since the inhalation time is continuous, we then integrate the inhalation time with the patient's respiratory model to obtain a stable inhalation volume value, reflecting the inhalation volume during a stable inhalation within the inhalation time. Using a quicksort algorithm, we sort the stable inhalation volume values ​​in ascending order and select the largest stable inhalation volume value, i.e., the patient's inhalation volume is maximized within this inhalation time. This inhalation time is recorded as the optimal inhalation time. Therefore, the initial moment when the respiratory ratio is less than 1 is recorded as the inhalation start moment and assigned a value of 1. Starting from 1, we accumulate and increment by 1 according to the time series until the result equals the optimal inhalation time. The last moment of the optimal inhalation time is the optimal inhalation moment.

[0057] The specific method for obtaining the stable inspiratory volume value is as follows:

[0058] ;

[0059] in, This indicates the stable value of inspiratory volume. to Indicates the inhalation time. This represents a patient's breathing model.

[0060] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An AI-assisted diagnostic system for CT rib fractures based on image recognition, characterized in that: Includes the following specific modules: Patient thermal imaging acquisition module: Acquires patient thermal imaging data, including the number of pixels, pixel coordinates and RGB color values. Marks pixels according to the RGB color values ​​corresponding to human body temperature to generate a rough contour. Matches the rough contour with a preset human body contour template to identify the patient contour. Further matches the mouth and nose contour and chest contour. Delineates the airflow temperature detection area with the center of the mouth and nose contour as the origin. The patient breathing analysis module sets a stationary factor by comparing the distance the patient's contour center point moves with the dynamic movement threshold. It obtains the chest change distance value by calculating the temporal difference between the distances between the chest contour feature points and the center point. It converts the average RGB color value of the airflow temperature detection area to HSL hue and calculates the difference between the cool and warm values ​​to obtain the temperature change value. Based on the stationary factor, the chest change distance value, and the temperature change value, it constructs a patient breathing model. It judges the breathing state by calculating the ratio of the patient breathing models at adjacent time points. It obtains the stable inspiratory volume value by integrating the patient breathing model during the inspiratory process and selects the time corresponding to the maximum stable inspiratory volume value as the inspiratory optimization time. CT image capture module: Controls the CT equipment to acquire rib CT images during the optimal inspiration time; CT image artifact analysis module: Extracts rib contours from rib CT images using image segmentation algorithms, and detects artifacts using contour tracking algorithms. If artifacts are present, the system returns to the patient thermal imaging acquisition module; otherwise, it executes the AI ​​rib fracture diagnosis module. AI rib fracture diagnosis module: Utilizes deep learning convolutional neural networks to perform end-to-end analysis of rib CT images, enabling automatic detection and diagnosis of rib fractures.

2. The CT rib fracture artificial intelligence-assisted diagnosis system based on image recognition according to claim 1, characterized in that: The specific method for obtaining the static factor is as follows: A stationary factor is set. Based on the number of pixels and their coordinates in the patient's outline, the center point coordinates of the patient's outline are calculated. The Euclidean distance between the center point coordinates of the patient's outline at the next moment and the center point coordinates of the patient's outline at the previous moment is calculated according to the time series to obtain the movement distance. A dynamic movement threshold is preset. The movement distance is compared with the dynamic movement threshold. If the movement distance is within the dynamic movement threshold, it means that the patient is stationary and the stationary factor is assigned a value of 1. If the movement distance is outside the dynamic movement threshold, it means that the patient is moving and the stationary factor is assigned a value of 0.

3. The CT rib fracture artificial intelligence-assisted diagnosis system based on image recognition according to claim 1, characterized in that: The specific method for obtaining the patient's respiratory model is as follows: The Euclidean distance between the two sides of the chest contour is calculated based on the time series to obtain the chest change distance value. The temperature change in the airflow temperature detection area is calculated based on the time series to obtain the temperature change value. The patient's breathing model is obtained by comprehensively calculating the rest factor, chest change distance value, and temperature change value. ； in, This represents a patient's respiratory model. Represents the static factor. Indicates the temperature change value. This represents the distance value of the chest change.

4. The CT rib fracture artificial intelligence-assisted diagnosis system based on image recognition according to claim 3, characterized in that: The specific method for obtaining the chest change distance value is as follows: The center point coordinates of the chest contour are obtained by calculating the Euclidean distance based on the number and coordinates of the pixels within the chest contour. The coordinates of some edge pixels of the chest contour are set as feature point coordinates. The Euclidean distance between the feature point coordinates and the center point coordinates of the chest contour is calculated to obtain the chest distance value. The chest distance value is obtained by calculating the difference between the chest distance value at the next moment and the chest distance value at the previous moment based on the time series.

5. The CT rib fracture artificial intelligence-assisted diagnosis system based on image recognition according to claim 3, characterized in that: The specific method for obtaining the temperature change value is as follows: The average RGB color values ​​corresponding to the pixel coordinates within the airflow temperature detection area are calculated to obtain the average RGB color value. This average RGB color value is then converted to HSL color space degrees to obtain the conversion degree. The HSL color space degrees range from 0 to 360 degrees. The higher the temperature, the warmer the color corresponding to the average RGB color value, with the warmer the color tending to be near 0 degrees in the HSL color space. Conversely, the cooler the temperature, the cooler the color corresponding to the average RGB color value, with the cooler the color tending to be near 240 degrees in the HSL color space. The difference between the conversion degree and 0 is calculated, and the absolute value is taken to obtain the warmer value. The difference between the conversion degree and 240 is calculated, and the absolute value is taken to obtain the cooler value. The difference between the cooler value and the warmer value is calculated to obtain the temperature change value.

6. The CT rib fracture artificial intelligence-assisted diagnosis system based on image recognition according to claim 1, characterized in that: The specific method for determining respiratory status by calculating the ratio of patient respiratory models at adjacent time points is as follows: The respiratory ratio is calculated by comparing the patient's respiratory model at the next moment with that at the previous moment based on the time series. The respiratory ratio is then compared with 1. If the respiratory ratio is greater than 1, it indicates exhalation; if the respiratory ratio is less than 1, it indicates inhalation.

7. The CT rib fracture artificial intelligence-assisted diagnosis system based on image recognition according to claim 6, characterized in that: The specific method for obtaining the respiratory ratio is as follows: ； in, Indicates the respiratory ratio, This indicates the patient's breathing model in the next moment. This represents the patient's breathing model from the previous moment. It represents a positive real number.

8. The CT rib fracture artificial intelligence-assisted diagnosis system based on image recognition according to claim 1, characterized in that: The specific method for obtaining the inhalation optimization timing is as follows: Record each consecutive moment when the respiratory ratio is less than 1 to obtain the inspiratory time. The inspiratory time is continuous. Then, perform integral calculation based on the inspiratory time and the patient's respiratory model to obtain the stable inspiratory volume value. Sort the stable inspiratory volume values ​​in ascending order using a quicksort algorithm, and select the largest stable inspiratory volume value. Record the inspiratory time of the largest stable inspiratory volume value as the inspiratory optimization time. Therefore, the initial moment when the respiratory ratio is less than 1 is recorded as the inspiratory start moment and assigned a value of 1. Starting from 1, accumulate and increment by 1 according to the time series until the result equals the inspiratory optimization time. The last moment of the inspiratory optimization time is the inspiratory optimization moment.

9. The CT rib fracture artificial intelligence-assisted diagnosis system based on image recognition according to claim 8, characterized in that: The specific method for obtaining the stable inspiratory volume value is as follows: ； in, This indicates the stable value of inspiratory volume. to Indicates the inhalation time. This represents a patient's breathing model.

Images