An adaptive motion trajectory intelligent image acquisition method

Abstract

The present invention discloses an adaptive motion trajectory intelligent image acquisition method, which uses an integrated Doppler radar sensor and an inertial measurement unit to monitor the patient's tiny physiological movements during imaging and generate physiological movement data; extracts data features through a convolutional neural network for real-time tracking; builds a personalized physiological movement model based on the extracted features to predict the patient's short-term movement trajectory; dynamically adjusts image acquisition based on the model to reduce blur caused by patient movement; obtains an intermediate image by performing feature fusion on the original image, and optimizes and generates a terminal image by using adaptive adjustment correction; finally, fuses the terminal images obtained at multiple time points, applies image enhancement technology, and outputs a medical image. The present invention accurately monitors and predicts the patient's tiny physiological movements during medical imaging, and dynamically adjusts image acquisition and compensation accordingly, and finally generates high-definition and high-accuracy medical images, which significantly improves the quality and reliability of medical images.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to an adaptive motion trajectory intelligent image acquisition method. Background Art

[0002] With the development of medical technology, medical imaging technology plays an increasingly important role in the diagnosis and treatment of diseases. However, in actual operation, especially when performing high-resolution medical imaging, the patient's tiny physiological movements, such as breathing and heartbeat, will seriously affect the image quality, causing image blur, which in turn affects the doctor's accurate judgment of the disease. Therefore, how to effectively compensate or eliminate the impact of these tiny movements has become an urgent problem to be solved.

[0003] Traditional medical imaging methods usually rely on patients to maintain a fixed posture or extend the scanning time to reduce the impact of motion artifacts. However, this method not only increases the discomfort of patients and reduces work efficiency, but also makes it difficult to obtain high-quality images for patients who cannot remain still for a long time. In addition, traditional compensation methods often post-process the image after imaging, which will lose the image details to a certain extent and cannot fundamentally solve the problem.

[0004] In recent years, some studies have attempted to monitor the patient's physiological movements in real time through sensor technology and to compensate in real time by combining image processing technology. Although these technologies can improve image quality to a certain extent, the lack of consideration of individual differences and the failure to fully combine the advantages of advanced algorithms such as deep learning have limited compensation effects, especially when dealing with complex or nonlinear physiological movements, which still poses great challenges.

[0005] Chinese patent CN115862817A discloses a medical image processing system and method based on artificial intelligence, which includes collecting medical images through B-ultrasound, converting the format of the collected medical images; performing target detection and feature extraction on the converted medical images; calibrating the target area; identifying the marker line of the calibrated target area; determining the angle range between the ventricular septum line and the vector line, and classifying them. The medical image is collected through B-ultrasound, including: uploading the acquired medical image to a remote data processing center; for the upload method of the medical image, the image output port is used for local storage, and the image storage module is combined with the network to transmit it to the data processing center for processing. The present invention has the characteristics of accurate positioning and target calibration. However, although the technical solution disclosed in the patent realizes the accurate positioning and target calibration of medical images, it mainly relies on centralized processing after image acquisition, and cannot cope with the image blur problem caused by the patient's physiological movement. In contrast, the present application fundamentally reduces motion artifacts and significantly improves the clarity and accuracy of medical images by real-time monitoring and intelligent compensation of the patient's tiny physiological movements during imaging, combined with image fusion and enhancement technology.

[0006] Chinese patent CN113658116B discloses an artificial intelligence method for generating medical images in different body positions, including the following steps: collecting medical images of subjects in different body positions as training samples; preprocessing the medical images of the training samples; using the processed training samples to train the artificial intelligence network model; collecting images of a body position of a new subject, and using the trained model to generate other images in different body positions. After completing the scanning of a patient in one body position, the image generation method of the present invention can generate images of other body positions through the model without the aid of any scanning equipment, and the generated images are of high accuracy. The generation speed is very fast. However, the patent is relatively lacking in the functions of real-time monitoring and compensation of patient physiological movements and ensuring image clarity, especially in scenarios where continuous imaging or dynamic monitoring of patient micro-movements is required, the quality and accuracy of the image cannot be guaranteed.

[0007] In order to solve the above problems, the present invention provides an adaptive motion trajectory intelligent image acquisition method, which can realize dynamic adjustment according to the actual conditions of different patients, thereby obtaining medical images with higher clarity and accuracy, greatly improving the reliability and efficiency of medical diagnosis. Summary of the invention

[0008] An adaptive motion trajectory intelligent image acquisition method comprises the following steps:

[0009] S1: Using the integrated Doppler radar sensor and inertial measurement unit, the patient's tiny physiological movements during the imaging process are monitored in real time in a non-contact manner, physiological movement data are generated, and the physiological movement data are filtered and processed;

[0010] S2: Extract feature data of physiological motion data through convolutional neural network model, track the feature data in real time, and accurately locate the patient's tiny movements;

[0011] S3: Establish a patient-specific physiological motion model based on the characteristic data of physiological motion data to dynamically capture the patient's physiological motion state;

[0012] S4: Based on key features and combined with the patient's physiological motion state, the time series analysis algorithm is used to predict the patient's motion trajectory in the short term in the future and generate a dynamic compensation model;

[0013] S5: The image acquisition device is used to acquire the initial image according to the patient's physiological movement state, and the image blurring caused by the patient's slight movement is eliminated through the dynamic compensation model;

[0014] S6: Using feature fusion technology on the initial image, the detailed information of the initial image is used to generate a mid-range image;

[0015] S7: Use the adaptive adjustment correction function to process the mid-end image, dynamically adjust the parameters of the image acquisition device according to the patient's movement, and generate the terminal image;

[0016] S8: After completing a single or multiple image acquisitions, the terminal images acquired at different time points are fused using image fusion technology and image enhancement algorithms to improve image quality, ensuring that the final output medical images have high clarity and accuracy.

[0017] Preferably, the S1 further comprises:

[0018] The Doppler radar sensor monitors the patient's breathing rate and heart rate changes in real time during the imaging process, and the inertial measurement unit simultaneously collects accelerometer and gyroscope data to detect changes in the patient's body position;

[0019] The filtering process of the physiological motion data adopts a bandpass filter H(s), where:

[0020]

[0021] Among them, ω0 and ω1 are the low-frequency cutoff frequency and high-frequency cutoff frequency respectively, and ω2 is the quality factor of the filter to remove irrelevant noise signals and obtain a purer physiological motion signal.

[0022] Preferably, S1 performs smoothing on the physiological motion signal before filtering, using a sliding average algorithm y(t), where:

[0023]

[0024] Here n represents the size of the sliding window, and x(t-i+1) represents the data of the i-th sampling point at time t, so as to reduce the fluctuation caused by the inherent noise of the sensor.

[0025] Preferably, S2 further comprises:

[0026] The pre-trained convolutional neural network model is used to extract features from the filtered physiological motion data. The convolutional neural network model processes the physiological motion data through convolutional layers, pooling layers, and fully connected layers to extract the significant features of the periodic patterns in the patient's physiological motion. The Kalman filter is used to track the extracted feature data in real time. in:

[0027]

[0028] Here represents the posterior estimate at time t, represents the prior estimate at time t, K t is the Kalman gain, z t is the observed value, H t It is an observation matrix that enables precise positioning of the patient’s tiny movements.

[0029] Preferably, S3 further comprises:

[0030] Based on the extracted feature data, a machine learning algorithm is used to build a personalized physiological movement model, which can reflect the patient's specific physiological movement pattern;

[0031] The personalized physiological motion model is constructed by taking into account individual differences in age, gender, and weight, and is constructed by continuously updating feature data using an online learning algorithm w(t+1), where:

[0032]

[0033] Here w(t) is the weight vector, η is the learning rate, y(t) is the target output, is the model prediction output, and x(t) is the input feature vector, which enables the model to adapt to changes in the patient's physiological state in real time.

[0034] Preferably, the S4 further comprises:

[0035] The time series analysis algorithm is used to predict the patient's movement trend in the next few seconds. The algorithm is based on historical physiological movement data and calculates X through the ARIMA model. t In the future movement trends, among them:

[0036] (1-φ1B-…-φ p Bp )(1-B) d X t =(1+θ1B+…+θ q B q )∈ t

[0037] Among them, X t Usually represents the observation value at time point t, φ i is the autoregressive coefficient, d is the difference order, θ i is the moving average coefficient, ∈ t is the error term, and the ARIMA model obtained through training is used to predict the future movement trajectory;

[0038] In the prediction process, the Kalman filtering method is used to dynamically compensate for the nonlinear characteristics of the patient's physiological movement.

[0039] Preferably, the S5 further comprises:

[0040] Based on the motion compensation model, the image acquisition device can automatically adjust parameters such as exposure time and frame rate when acquiring the initial image to minimize the blur caused by motion;

[0041] In the image processing stage, multiple frames of images are aligned through image registration technology, using a rigid transformation T(x,y), where:

[0042] T(x,y)=(a 11 x+a 12 y+t x ,a 21 x+a 22 y+t y )

[0043] Here a ij is the rotation and translation parameter, t x and t y is the translation vector, which further reduces image distortion caused by slight patient movement.

[0044] Preferably, the S6 further comprises:

[0045] The feature fusion technology is used to process the initial image, and the information from different sensors is integrated through multi-scale analysis and multi-modal data fusion to improve the detail richness of the image;

[0046] The fusion process also uses noise suppression technology, using Wiener filtering in:

[0047]

[0048] Among them, G(f) is the image spectrum after denoising, H(f) is the degradation function, N(f) is the noise power spectrum, S(f) is the power spectrum of the original image, and H * (f) is the conjugate function of H(f); this ensures that the image quality is not affected by the additional noise introduced during the processing.

[0049] Preferably, the S7 further comprises:

[0050] The adaptive adjustment correction function is used to dynamically adjust the magnification and focal length of the image acquisition device according to the patient's movement to make the image have high definition; the adaptive adjustment correction function is used to automatically perform white balance and color correction, and the color constancy algorithm R is used. corr ,in:

[0051]

[0052] Among them, R corr is the corrected color value, R obs is the observed color value, is the reference color value;

[0053] Ensure consistent color representation of images acquired under different conditions.

[0054] Preferably, the S8 further comprises: after completing the single or multiple frame image acquisition, using image fusion technology to synthesize the terminal images acquired at different time points into one image;

[0055] The image fusion process utilizes an image enhancement algorithm, which uses histogram equalization to enhance image quality through the cumulative distribution function cdf.

[0056]

[0057] Among them, hist[i] is the number of pixels at gray level i, N is the total number of pixels, and the cumulative distribution function cdf is used to improve the overall quality of the image. Image stitching and inpainting technology are used to make the output medical image have high definition and accuracy.

[0058] Compared with the prior art, the technical effects of the present invention are:

[0059] The present invention integrates Doppler radar sensors and inertial measurement units to monitor the patient's tiny physiological movements in real time, and combines a convolutional neural network model to extract feature data, thereby solving the problem of image blur caused by the patient's physiological movements in traditional medical imaging, thereby significantly improving the clarity and accuracy of medical images.

[0060] The present invention uses a time series analysis algorithm to predict the patient's motion trajectory in the short term in the future and generates a dynamic compensation model, which solves the problem in the prior art that it is difficult to respond to the patient's unexpected movement in real time. By dynamically adjusting the image acquisition parameters, effective compensation for motion artifacts is achieved, ensuring the quality of each image acquired.

[0061] The present invention processes mid-range images through adaptive adjustment and correction functions, and uses image fusion technology to synthesize terminal images acquired at different time points into a high-definition image, thereby solving the problem of missing details that may exist in a single image and further improving the overall quality and diagnostic value of medical images.

[0062] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application so that it can be implemented in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following is a detailed description of the preferred embodiments of the present application in conjunction with the accompanying drawings as follows.

[0063] Based on the detailed description of the specific embodiments of the present application in combination with the accompanying drawings below, those skilled in the art will become more aware of the above and other objects, advantages and features of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings without creative work. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale.

[0065] Figure 1 It is a flow chart of an adaptive motion trajectory intelligent image acquisition method;

[0066] Figure 2 This is an image comparison chart;

[0067] Figure 3 This is a comparison chart of image data. Specific embodiments

[0068] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiment of the present application. Obviously, the described embodiment is a part of the embodiment of the present application, rather than all of the embodiments. In the following description, specific details such as specific configuration and components are provided only to help fully understand the embodiments of the present application. Therefore, it should be clear to those skilled in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, for clarity and brevity, the description of known functions and structures is omitted in the embodiment.

[0069] It should be understood that the references to "one embodiment" or "this embodiment" throughout the specification mean that the specific features, structures, or characteristics associated with the embodiment are included in at least one embodiment of the present application. Therefore, the references to "one embodiment" or "this embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0070] In addition, the present application may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplicity and clarity, and does not in itself indicate the relationship between the various embodiments and / or settings discussed.

[0071] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" in this article describes another type of association object relationship, indicating that there can be two relationships. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0072] The term "at least one" in this article is merely a description of the association relationship of associated objects, indicating that there may be three relationships. For example, at least one of A and B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0073] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusions.

[0074] Example 1

[0075] This embodiment mainly describes a template model and connection method of an adaptive motion trajectory intelligent image acquisition method;

[0076] Further, if Figure 1 As shown, an adaptive motion trajectory intelligent image acquisition method includes the following steps:

[0077] S1: Using the integrated Doppler radar sensor and inertial measurement unit, the patient's tiny physiological movements during the imaging process are monitored in real time in a non-contact manner, physiological movement data are generated, and the physiological movement data are filtered and processed;

[0078] S2: Extract feature data of physiological motion data through convolutional neural network model, track the feature data in real time, and accurately locate the patient's tiny movements;

[0079] S3: Establish a patient-specific physiological motion model based on the characteristic data of physiological motion data to dynamically capture the patient's physiological motion state;

[0080] S4: Based on key features and combined with the patient's physiological motion state, the time series analysis algorithm is used to predict the patient's motion trajectory in the short term in the future and generate a dynamic compensation model;

[0081] S5: The image acquisition device is used to acquire the initial image according to the patient's physiological movement state, and the image blurring caused by the patient's slight movement is eliminated through the dynamic compensation model;

[0082] S6: Using feature fusion technology on the initial image, the detailed information of the initial image is used to generate a mid-range image;

[0083] S7: Use the adaptive adjustment correction function to process the mid-end image, dynamically adjust the parameters of the image acquisition device according to the patient's movement, and generate the terminal image;

[0084] S8: After completing a single or multiple image acquisitions, the terminal images acquired at different time points are fused using image fusion technology and image enhancement algorithms to improve image quality, ensuring that the final output medical images have high clarity and accuracy.

[0085] By using Doppler radar sensors and inertial measurement units, the patient's physiological movements can be monitored in real time without contact, and data filtering and smoothing can be performed through bandpass filters and sliding average algorithms, effectively removing irrelevant noise and improving the purity of physiological movement data.

[0086] Furthermore, the S1 further comprises:

[0087] The Doppler radar sensor monitors the patient's breathing rate and heart rate changes in real time during the imaging process, and the inertial measurement unit simultaneously collects accelerometer and gyroscope data to detect changes in the patient's body position;

[0088] The filtering process of the physiological motion data adopts a bandpass filter H(s), where:

[0089]

[0090] Among them, ω0 and ω1 are the low-frequency cutoff frequency and high-frequency cutoff frequency respectively, and ω2 is the quality factor of the filter to remove irrelevant noise signals and obtain a purer physiological motion signal.

[0091] Furthermore, S1 performs smoothing on the physiological motion signal before filtering, using a sliding average algorithm y(t), where:

[0092]

[0093] Here n represents the size of the sliding window, and x(t-i+1) represents the data of the i-th sampling point at time t, so as to reduce the fluctuation caused by the inherent noise of the sensor.

[0094] Through the pre-trained convolutional neural network model, it is possible to extract significant features from the filtered physiological motion data, and track these features in real time through the Kalman filter, thereby achieving accurate positioning of the patient's tiny movements.

[0095] Furthermore, the S2 further comprises:

[0096] The pre-trained convolutional neural network model is used to extract features from the filtered physiological motion data. The convolutional neural network model processes the physiological motion data through convolutional layers, pooling layers, and fully connected layers to extract the significant features of the periodic patterns in the patient's physiological motion. The Kalman filter is used to track the extracted feature data in real time. in:

[0097]

[0098] Here represents the posterior estimate at time t, represents the prior estimate at time t, K t is the Kalman gain, z t is the observed value, H t It is an observation matrix that enables precise positioning of the patient’s tiny movements.

[0099] Furthermore, the S3 further comprises:

[0100] Based on the extracted feature data, a machine learning algorithm is used to build a personalized physiological movement model, which can reflect the patient's specific physiological movement pattern;

[0101] The personalized physiological motion model is constructed by taking into account individual differences in age, gender, and weight, and is constructed by continuously updating feature data using an online learning algorithm w(t+1), where:

[0102]

[0103] Here w(t) is the weight vector, η is the learning rate, y(t) is the target output, is the model prediction output, and x(t) is the input feature vector, which enables the model to adapt to changes in the patient's physiological state in real time.

[0104] The personalized physiological movement model is constructed using machine learning algorithms, which can update the model in real time based on individual differences of patients such as age, gender and weight, so that it better reflects the patient's specific physiological movement patterns.

[0105] Furthermore, the S4 further comprises:

[0106] The time series analysis algorithm is used to predict the patient's movement trend in the next few seconds. The algorithm is based on historical physiological movement data and calculates X through the ARIMA model. t In the future movement trends, among them:

[0107] (1-φ1B-…-φ p B p )(1-B) d X t =(1+θ1B+…+θ q B q )∈ t

[0108] Among them, X t Usually represents the observation value at time point t, φ i is the autoregressive coefficient, d is the difference order, θ i is the moving average coefficient, ∈ t is the error term, and the ARIMA model obtained through training is used to predict the future movement trajectory;

[0109] In the prediction process, the Kalman filtering method is used to dynamically compensate for the nonlinear characteristics of the patient's physiological movement.

[0110] The time series analysis algorithm based on historical physiological motion data and the Kalman filtering method can predict the patient's future motion trend and generate a dynamic compensation model based on it to reduce the impact of motion artifacts.

[0111] Furthermore, the S5 further comprises:

[0112] Based on the motion compensation model, the image acquisition device can automatically adjust parameters such as exposure time and frame rate when acquiring the initial image to minimize the blur caused by motion;

[0113] In the image processing stage, multiple frames of images are aligned through image registration technology, using a rigid transformation T(x,y), where:

[0114] T(x,y)=(a 11 x+a 12 y+t x ,a 21 x+a 22 y+t y )

[0115] Here a ij is the rotation and translation parameter, t x and t y is the translation vector, which further reduces image distortion caused by slight patient movement.

[0116] Based on the dynamic compensation model, the image acquisition device can automatically adjust parameters such as exposure time and frame rate to minimize motion blur, and image registration technology further reduces image distortion caused by patient movement.

[0117] Furthermore, the S6 further comprises:

[0118] The feature fusion technology is used to process the initial image, and the information from different sensors is integrated through multi-scale analysis and multi-modal data fusion to improve the detail richness of the image;

[0119] The fusion process also uses noise suppression technology, using Wiener filtering in:

[0120]

[0121] Among them, G(f) is the image spectrum after denoising, H(f) is the degradation function, N(f) is the noise power spectrum, S(f) is the power spectrum of the original image, and H * (f) is the conjugate function of H(f); this ensures that the image quality is not affected by the additional noise introduced during the processing.

[0122] Through multi-scale analysis and multi-modal data fusion technology, combined with noise suppression technology such as Wiener filtering, the richness of image details is improved and the image quality is maintained.

[0123] Further, the S7 further comprises:

[0124] The adaptive adjustment correction function is used to dynamically adjust the magnification and focal length of the image acquisition device according to the patient's movement to make the image have high definition; the adaptive adjustment correction function is used to automatically perform white balance and color correction, and the color constancy algorithm R is used. corr ,in:

[0125]

[0126] Among them, R corr is the corrected color value, R obs is the observed color value, is the reference color value;

[0127] Ensure consistent color representation of images acquired under different conditions.

[0128] The adaptive adjustment correction function can dynamically adjust the parameters of the image acquisition device, such as magnification and focal length, according to the patient's movement, ensuring high-definition images. At the same time, automatic white balance and color correction technology ensure that the image has consistent color performance.

[0129] Furthermore, the S8 further includes: after completing the single or multiple frame image acquisition, using image fusion technology to synthesize the terminal images acquired at different time points into one image;

[0130] The image fusion process utilizes an image enhancement algorithm, which uses histogram equalization to enhance image quality through the cumulative distribution function cdf.

[0131]

[0132] Among them, hist[i] is the number of pixels at gray level i, N is the total number of pixels, and the cumulative distribution function cdf is used to improve the overall quality of the image. Image stitching and inpainting technology are used to make the output medical image have high definition and accuracy.

[0133] This embodiment integrates Doppler radar sensors and inertial measurement units to monitor the patient's physiological movements in real time, and uses convolutional neural networks and machine learning algorithms to extract and track feature data, establish a personalized physiological movement model, predict movement trajectories, dynamically adjust image acquisition parameters, and combine feature fusion and image enhancement technology to ultimately generate high-definition and high-accuracy medical images, thereby significantly improving image quality and the reliability of clinical diagnosis.

[0134] Example 2

[0135] This embodiment is based on Embodiment 1, and mainly describes the implementation steps in detail:

[0136] Environment Setup

[0137] Device: Includes a Doppler radar sensor and an inertial measurement unit, integrated into a small wearable device.

[0138] Subject: An adult patient with heart disease undergoing CT scan.

[0139] Data collection and processing

[0140] Sensor data: The Doppler radar sensor detected a respiratory rate of 16 beats per minute and a heart rate of 75 beats per minute. The accelerometer and gyroscope data collected by the IMU showed that the patient had slight body shaking during the scan.

[0141] Data filtering: The physiological motion data were filtered using a bandpass filter, with the low-frequency cutoff frequency set to 0.1 Hz, the high-frequency cutoff frequency set to 1 Hz, and the quality factor set to 0.5 to effectively remove environmental noise.

[0142] Smoothing: The sliding average algorithm is used, and the window size is set to 5 sampling points to reduce data fluctuations caused by the inherent noise of the sensor.

[0143] Feature extraction and real-time tracking

[0144] Feature extraction: The convolutional neural network model successfully extracted the periodic features of the patient’s breathing and heartbeat from the processed data.

[0145] Real-time tracking: Through the Kalman filter, the system is able to track these feature data in real time and accurately locate the patient even if there is a slight movement during the scan.

[0146] Personalized physiological movement model

[0147] Model construction: Based on the patient's specific physiological parameters (age 45 years old, male, weight 80 kg), a personalized physiological movement model was constructed.

[0148] Model update: The online learning algorithm updates the model once per second to reflect the patient's real-time physiological state changes.

[0149] Motion compensation and image acquisition

[0150] Movement prediction: The ARIMA model is used to predict the patient's movement trend in the next 5 seconds. The results show that the patient may move slightly up and down due to breathing.

[0151] Compensation model: The Kalman filter method is used to dynamically compensate for nonlinear motion and generate a compensation model.

[0152] Image acquisition: The image acquisition device automatically adjusts the exposure time to 20 ms and the frame rate to 30 frames / s according to the compensation model to reduce motion blur.

[0153] Image Processing

[0154] Feature fusion: The initial image is processed through feature fusion technology, using multi-scale analysis and multi-modal data fusion to increase image details.

[0155] Noise suppression: The Wiener filter technique was used to reduce noise, and the image spectrum showed that the noise power was reduced by 20dB.

[0156] Image registration: Multiple frames of images are aligned using rigid transformation with a rotation angle less than 0.5 degrees and a translation distance less than 1 mm.

[0157] Final image generation

[0158] Adaptive adjustment: The image acquisition device automatically adjusts the magnification to 2X and the focal length to 50mm according to the patient's movement to ensure image clarity.

[0159] Image fusion: Images at different time points are synthesized through image fusion technology, and histogram equalization shows that the image contrast is improved by 30%.

[0160] This embodiment is illustrated by specific examples. The method significantly reduces the artifacts caused by motion in the image by real-time monitoring and compensating for the patient's tiny physiological movements; the feature fusion technology and noise reduction processing of multi-source information are integrated to make the final generated image have higher detail richness and clarity; the image acquisition parameters are dynamically adjusted according to the specific physiological conditions of each patient to ensure the consistency and accuracy of the image; the real-time tracking and prediction algorithm ensures that timely response can be made even in the face of rapidly changing situations, thereby improving imaging efficiency.

[0161] Example 3

[0162] This embodiment is based on Embodiments 1 and 2, and mainly describes the comparison of images generated by the prior art and the technical solution of the present application;

[0163] like Figure 2 As shown, the left picture is a picture taken by the prior art, and the right picture is a picture taken by the present technical solution. The left picture is divided into four points: A, B, C, and D, and the right picture is divided into four points: A', B', C', and D'. By comparing the left and right pictures, it can be found that the CT images under the prior art often have motion artifacts, resulting in blurred images and serious loss of detail information, because they fail to effectively compensate for the effects of micro-physiological movements such as breathing and heartbeat during the imaging process. These problems not only reduce the image quality, but also directly affect the doctor's accurate judgment of the condition, especially in diagnostic scenarios that require high-resolution images. These problems are particularly prominent.

[0164] This technical solution establishes a personalized physiological movement model by monitoring the patient's tiny physiological movements in real time and extracting feature data in combination with a convolutional neural network model. It can predict the patient's future movement trends and generate a dynamic compensation model. By dynamically adjusting the image acquisition parameters and combining feature fusion and image enhancement technology, the final generated CT image has significantly higher clarity and detail richness. From the right figure, it can be found that the four points A', B', C', and D' have obvious shadows and clearer boundaries. Specifically, the technical solution of this application ensures the high accuracy and reliability of the image and greatly improves the effect of medical diagnosis.

[0165] By comparing the CT images taken by the prior art with the CT under the scheme of the present application, this embodiment can show that the technical scheme of the present application significantly improves the clarity and accuracy of medical images by real-time monitoring and compensating for the patient's tiny physiological movements during the imaging process, combining feature extraction, dynamic compensation model and image fusion enhancement technology, solves the image blur problem caused by motion artifacts in traditional technologies, and greatly improves the reliability and efficiency of medical diagnosis.

[0166] Example 4

[0167] This embodiment is based on Embodiment 3, and mainly compares the peak signal-to-noise ratio PSNR and the structural similarity index SSIM of the prior art and the present application;

[0168] like Figure 3 As shown, for the quality and clarity of the image, the peak signal-to-noise ratio PSNR and the structural similarity index SSIM are used as evaluation indicators to compare the image quality processed by the prior art and the technical solution of the present application. The left figure is a comparison diagram of the structural similarity index SSIM, and the right figure is a comparison diagram of the peak signal-to-noise ratio PSNR; for the image area photographed by the prior art, the technical solution of the present application shoots the image area again, and obtains a significant image processing effect. The PSNR value of the prior art solution is about 25dB, while the PSNR value of the technical solution of the present application reaches 35dB, which is increased by about 40%; the SSIM value of the prior art solution is 0.7, while the SSIM value of the technical solution of the present application is 0.9, which is increased by about 28.57%, which shows that the technical solution of the present application can more effectively improve the image quality in photo acquisition and remove the blur and unclear phenomenon of the picture.

[0169] Through the comparison of the peak signal-to-noise ratio PSNR and the structural similarity index SSIM, this embodiment can know that the technical solution of the present application solves the problem of image blurring caused by the physiological movement of the patient in traditional medical imaging, improves the clarity and accuracy of the image, and ensures the high reliability and diagnostic value of medical images.

[0170] The above are only preferred embodiments of the present invention, which do not limit the scope of protection of the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any changes, modifications, replacements, integrations and parameter changes to these embodiments within the spirit and principles of the present invention through conventional substitutions or without departing from the principles and spirit of the present invention fall within the scope of protection of the present invention.

Claims

1. An adaptive motion trajectory intelligent image acquisition method, comprising the following steps: S1: Using the integrated Doppler radar sensor and inertial measurement unit, the patient's tiny physiological movements during the imaging process are monitored in real time in a non-contact manner, physiological movement data are generated, and the physiological movement data are filtered and processed; S2: Extract feature data of physiological motion data through convolutional neural network model, track the feature data in real time, and accurately locate the patient's tiny movements; S3: Establish a patient-specific physiological motion model based on the characteristic data of physiological motion data to dynamically capture the patient's physiological motion state; S4: Based on key features and combined with the patient's physiological motion state, the time series analysis algorithm is used to predict the patient's motion trajectory in the short term in the future and generate a dynamic compensation model; S5: The image acquisition device is used to acquire the initial image according to the patient's physiological movement state, and the image blurring caused by the patient's slight movement is eliminated through the dynamic compensation model; S6: Using feature fusion technology on the initial image, the detailed information of the initial image is used to generate a mid-range image; S7: Use the adaptive adjustment correction function to process the mid-end image, dynamically adjust the parameters of the image acquisition device according to the patient's movement, and generate the terminal image; S8: After completing a single or multiple image acquisitions, the terminal images acquired at different time points are fused using image fusion technology and image enhancement algorithms to improve image quality, ensuring that the final output medical images have high clarity and accuracy.

2. The adaptive motion trajectory intelligent image acquisition method according to claim 1, characterized in that: The S1 further comprises: The Doppler radar sensor monitors the patient's breathing rate and heart rate changes in real time during the imaging process, and the inertial measurement unit simultaneously collects accelerometer and gyroscope data to detect changes in the patient's body position; The filtering process of the physiological motion data adopts a bandpass filter ,in: in, and are the low frequency cutoff frequency and the high frequency cutoff frequency respectively, It is the quality factor of the filter to remove irrelevant noise signals and obtain a purer physiological motion signal.

3. The adaptive motion trajectory intelligent image acquisition method according to claim 2, characterized in that: The S1 performs smoothing on the physiological motion signal before filtering, using a sliding average algorithm ,in: Here represents the size of the sliding window, Indicates at time Previous The data of each sampling point can be used to reduce the fluctuation caused by the inherent noise of the sensor.

4. The adaptive motion trajectory intelligent image acquisition method according to claim 1, characterized in that: The S2 further comprises: The pre-trained convolutional neural network model is used to extract features from the filtered physiological motion data. The convolutional neural network model processes the physiological motion data through convolutional layers, pooling layers, and fully connected layers to extract the significant features of the periodic patterns in the patient's physiological motion. The Kalman filter is used to track the extracted feature data in real time. ,in: Here Indicates at time The posterior estimate of Indicates at time A priori estimate of is the Kalman gain, is the observed value, It is an observation matrix that enables precise positioning of the patient’s tiny movements.

5. The adaptive motion trajectory intelligent image acquisition method according to claim 1 or 4, characterized in that: The S3 further comprises: Based on the extracted feature data, a machine learning algorithm is used to build a personalized physiological movement model, which can reflect the patient's specific physiological movement pattern; The personalized physiological motion model is constructed by taking into account individual differences in age, gender, and weight, and by continuously updating feature data using an online learning algorithm. Build where: Here is the weight vector, is the learning rate, is the target output, is the model prediction output, is the input feature vector, which enables the model to adapt to changes in the patient's physiological state in real time.

6. The adaptive motion trajectory intelligent image acquisition method according to claim 1, characterized in that: The S4 further comprises: The time series analysis algorithm is used to predict the patient's movement trend in the next few seconds. The algorithm is based on historical physiological movement data and calculated by the ARIMA model. In the future movement trends, among them: in, Usually means at a point in time The observed value of is the autoregressive coefficient, d is the difference order, is the moving average coefficient, is the error term, and the ARIMA model obtained through training is used to predict the future movement trajectory; In the prediction process, the Kalman filtering method is used to dynamically compensate for the nonlinear characteristics of the patient's physiological movement.

7. The adaptive motion trajectory intelligent image acquisition method according to claim 1, characterized in that: The S5 further comprises: Based on the motion compensation model, the image acquisition device can automatically adjust the exposure time and frame rate parameters when acquiring the initial image to minimize the blur caused by motion; In the image processing stage, multiple frames of images are aligned through image registration technology, and rigid transformation is used. ,in: Here are the rotation and translation parameters, and is the translation vector, which further reduces image distortion caused by slight patient movement.

8. The adaptive motion trajectory intelligent image acquisition method according to claim 1, characterized in that: The S6 further comprises: The feature fusion technology is used to process the initial image, and the information from different sensors is integrated through multi-scale analysis and multi-modal data fusion to improve the detail richness of the image; The fusion process also uses noise suppression technology, using Wiener filtering ,in: in, is the spectrum of the image after denoising, is the degradation function, is the noise power spectrum, is the power spectrum of the original image, yes This ensures that the image quality is not affected by the additional noise introduced during the processing.

9. The adaptive motion trajectory intelligent image acquisition method according to claim 1, characterized in that: The S7 further comprises: Adaptive adjustment and correction function is used to dynamically adjust the magnification and focal length of the image acquisition device according to the patient's movement, so that the image has high definition; white balance and color correction are automatically performed through adaptive adjustment and correction function, and color constancy algorithm is used ,in: in, is the corrected color value, is the observed color value, is the reference color value; Ensure consistent color representation of images acquired under different conditions.

10. The adaptive motion trajectory intelligent image acquisition method according to claim 1, characterized in that: The S8 further includes: after completing the single or multiple frame image acquisition, using image fusion technology to synthesize the terminal images acquired at different time points into one image; The image fusion process uses an image enhancement algorithm, which uses histogram equalization to generate a cumulative distribution function. To enhance image quality, the cumulative distribution function for: in, Is grayscale The number of pixels, is the total number of pixels, through the cumulative distribution function Improve the overall quality of the image and use image stitching and repair technology to make the output medical images have high clarity and accuracy.

Images