Neuroprotection-based acupuncture treatment effect image recognition method and system

Abstract

The embodiment of the invention provides an acupuncture treatment effect image recognition method and system based on nerve protection, and relates to the technical field of image recognition, and the method comprises the steps: receiving an image and the individual condition of a user, carrying out the evaluation of the image, obtaining a quality evaluation result of the image, and carrying out the self-adaptive physiological feature extraction of the image; obtaining physiological texture features and artifact features; dividing regions in the image, and analyzing the target region and the reference region based on the physiological texture features and the artifact features to obtain physiological feature differences between the target region and the reference region; verifying the physiological feature difference based on the physiological feature difference in combination with preset physiological prior information to obtain reliable physiological features; performing weighted integration on the physiological texture features, the artifact features and the reliable physiological features to obtain comprehensive physiological indexes; and adjusting the threshold value of the comprehensive physiological index according to the individual condition of the user to obtain a quantized acupuncture treatment effect. The reliability of the treatment scheme can be improved.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and in particular to an image recognition method and system for acupuncture treatment effects based on neuroprotection. Background Technology

[0002] In related technologies, medical institutions typically perform routine maintenance and calibration of imaging equipment used to evaluate the effects of acupuncture treatment to ensure that the image acquisition quality meets the precision standards required for clinical assessment. However, during long-term continuous use, even with routine maintenance, the tiny mechanical structures inside the optical lenses of the imaging equipment may experience extremely slight positional drift in the focusing mechanism due to minor vibrations or changes in ambient temperature. Although this drift is small enough not to trigger an alarm from the device's built-in self-test system, it is enough to have a cumulative impact on image sharpness, making the acquired image less sharp, such as slight blurring at the edges of blood vessels or nerve bundles. This slight decrease in image sharpness is often difficult for clinicians to detect during routine manual interpretation, but it constitutes a substantial obstacle for image recognition methods that aim to quantify changes in skin microcirculation or subtle adjustments in neural activity after acupuncture stimulation. The neuroprotective effect of acupuncture is often reflected in very subtle physiological responses, such as the degree of microvascular dilation in specific acupoint areas, subtle changes in blood flow velocity, or the morphology and arrangement of subcutaneous nerve fiber bundles, as well as the subtle texture and color gradient information of the surrounding tissue cells. When image sharpness is insufficient, these already very subtle texture and color gradient information become even more blurred. This blurriness can severely affect the ability to identify differences, especially when precise comparative analysis of different physiological response areas of a patient is required. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes an image recognition method and system for acupuncture treatment effects based on neuroprotection, aiming to improve the reliability of treatment plans.

[0004] In a first aspect, embodiments of this application provide an image recognition method for acupuncture treatment effects based on neuroprotection, including: Receive images and individual user information, perform quality assessment on the images, and obtain the quality assessment results of the images; Based on the quality assessment results of the image, adaptive physiological feature extraction is performed on the image to obtain physiological texture features and artifact features of different levels of detail. The target region and reference region in the image are divided, and based on the physiological texture features of different fineness and the artifact features, the context association analysis of the target region and the reference region is performed to obtain the physiological feature differences between the target region and the reference region. Based on the differences in physiological characteristics and combined with preset physiological prior information, physiological prior verification is performed on the differences in physiological characteristics to obtain reliable physiological characteristics. The physiological texture features, artifact features, and reliable physiological features are weighted and integrated to obtain a comprehensive physiological index. The quantitative threshold of the comprehensive physiological indicators is adjusted according to the individual circumstances of the user to obtain a quantitative acupuncture treatment effect.

[0005] According to some embodiments of this application, the step of adaptively extracting physiological features from the image based on the image quality assessment results to obtain physiological texture features and artifact features of different levels of detail includes: The image is divided into local regions to obtain multiple local regions; For each of the aforementioned local regions, multiple frames of image sequences are continuously acquired; Based on the multi-frame image sequence, identify the transient motion pattern of each local region; Based on the transient motion pattern and the quality assessment results of the image, adaptive physiological feature extraction is performed on the local region to obtain the transient deformation features of the local region. The transient deformation features are distinguished to obtain physiological texture features and artifact features of different levels of detail.

[0006] According to some embodiments of this application, the step of dividing the image into local regions to obtain multiple local regions includes: The physiological structure density of the image is evaluated to obtain the physiological structure density; The division size is adjusted according to the physiological structure density to obtain the target division size; The image is divided into local regions based on the target segmentation size to obtain multiple local regions.

[0007] According to some embodiments of this application, the weighted integration of the physiological texture features, the artifact features, and the reliable physiological features to obtain a comprehensive physiological index includes: Receive clinical feedback information; Based on the clinical feedback information, the weights of the physiological texture features, artifact features, and reliable physiological features related to the clinical feedback information are dynamically adjusted to obtain the target weights; The physiological texture features, artifact features, and reliable physiological features are weighted and integrated according to the target weights to obtain a comprehensive physiological index.

[0008] According to some embodiments of this application, the step of evaluating the physiological structure density of the image to obtain the physiological structure density includes: Acquire the visible light image and infrared image from the image; Local texture features are extracted from the visible light image to obtain local texture features; Temperature distribution analysis was performed on the infrared image to obtain temperature distribution characteristics; Based on the local texture features and the temperature distribution features, the physiological structures in the image are preliminarily identified, and a preliminary identification result is obtained. The preliminary recognition results are subjected to morphological filtering to obtain the processed image; The processed image is subjected to connected component analysis to obtain multiple initial connected components; Physiological feature verification is performed on each of the initial connected regions to obtain the target connected regions; The area ratio of the target connected region within the preset region is calculated to obtain the physiological structure density.

[0009] According to some embodiments of this application, the step of performing quality assessment on the image to obtain the image quality assessment result includes: The images are received at multiple preset time points to obtain a target time series image; Inter-frame consistency analysis is performed on the target time series image to obtain the inter-frame consistency deviation; Based on the inter-frame consistency deviation, the systematic deviation is obtained; The quality assessment results are obtained based on the systematic deviations described.

[0010] According to some embodiments of this application, the step of receiving the image according to a preset plurality of time points to obtain a target time series image includes: The images are received at multiple preset time points to obtain the original time series images; The original time-series image is subjected to illumination correction to obtain an illumination-corrected time-series image; Motion trajectory analysis is performed on the sequence of each frame in the illumination-corrected time-series image to obtain the local motion trajectory; Based on the local motion trajectory, a nonlinear or non-uniform deformation mode is obtained; Deformation compensation is performed based on the nonlinear or non-uniform deformation mode to obtain a sequence of compensated images for each frame. The sequence of each compensated image frame is superimposed to obtain the target time series image.

[0011] According to some embodiments of this application, the step of performing illumination correction on the original time-series image to obtain an illumination-corrected time-series image includes: The illumination intensity distribution of the environment in which the original time-series images are located is collected to obtain the illumination intensity distribution; Based on the light intensity distribution, identify non-uniform illumination regions and instantaneous light fluctuation regions in the original time series image; Local illumination compensation is performed on the non-uniform illumination area to obtain the compensated local illumination area; The instantaneous illumination fluctuation region is smoothed in the time domain to obtain a smoothed time domain illumination region. The compensated local illumination region and the smoothed temporal illumination region are fused to obtain an illumination-corrected time-series image.

[0012] According to some embodiments of this application, the step of performing motion trajectory analysis on the sequence of each frame in the illumination-corrected time-series image to obtain a local motion trajectory includes: Deformation features are extracted from the illumination-corrected time-series images to obtain deformation features; The deformation characteristics are subjected to physiological verification to obtain deformation characteristics that pass the verification. Based on the verified deformation features, motion trajectory analysis is performed on the sequence of each frame in the illumination-corrected time-series image to obtain the local motion trajectory.

[0013] Secondly, embodiments of this application provide an image recognition system for acupuncture treatment effects based on neuroprotection, comprising: The receiving and evaluation module is used to receive images and individual information about users, and to perform quality evaluation on the images to obtain the quality evaluation results of the images; The feature extraction module is used to perform adaptive physiological feature extraction on the image based on the quality assessment results of the image, so as to obtain physiological texture features and artifact features of different levels of detail. The analysis module is used to divide the target region and the reference region in the image, and based on the physiological texture features of different fineness and the artifact features, to perform contextual association analysis on the target region and the reference region to obtain the physiological feature differences between the target region and the reference region. The verification module is used to perform physiological prior verification on the physiological feature differences based on the physiological feature differences and in combination with preset physiological prior information to obtain reliable physiological features. The weighted integration module is used to weight and integrate the physiological texture features, the artifact features, and the reliable physiological features to obtain a comprehensive physiological index. The adjustment module is used to adjust the quantification threshold of the comprehensive physiological indicators according to the individual circumstances of the user, so as to obtain a quantified acupuncture treatment effect.

[0014] The technical solution according to the embodiments of this application has at least the following beneficial effects: The image recognition method for acupuncture treatment effects based on neuroprotection disclosed in this application, by receiving images and the individual situation of the user, and performing quality assessment on the images, can effectively identify and quantify problems such as decreased image sharpness, amplified noise, and artifacts caused by minor drift of the imaging device or overcompensation of the automatic enhancement program, thereby overcoming the substantial obstacle of image quality problems to the extraction of fine physiological features in the prior art. On this basis, the method performs adaptive physiological feature extraction based on the image quality assessment results, which can obtain physiological texture features and artifact features of different levels of fineness, avoiding the dilemma of traditional methods that are difficult to accurately peel off the target area in complex backgrounds. By performing contextual association analysis on the target area and the reference area, and combining it with physiological prior information for verification, the reliability of the extracted physiological features is ensured, effectively solving the problem of frequent misjudgment by existing classification judgment logic when facing distorted and unclear feature data. Ultimately, by weighted integration of physiological texture features, artifact features, and reliable physiological features, and by adjusting the quantification threshold according to the individual user's situation, this application can obtain a quantified acupuncture treatment effect, thereby achieving an objective and accurate assessment of the neuroprotective effect of acupuncture. This provides a reliable basis for doctors to accurately guide personalized acupuncture treatment plans based on the individual response of patients.

[0015] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0016] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0017] Figure 1 A flowchart illustrating an image recognition method for acupuncture treatment effects based on neuroprotection, provided in one embodiment of this application; Figure 2 This is a schematic diagram of an image recognition system for acupuncture treatment effects based on neuroprotection, provided in one embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical methods, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] It should be noted that the meaning of "multiple" (or "more than") in the description of the embodiments of this application refers to two or more, and "greater than," "less than," "exceeding," etc. are understood to exclude the number itself, while "above," "below," "within," etc. are understood to include the number itself. If "first," "second," etc. are used in the description, they are only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0020] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, the simultaneous existence of A and B, or the existence of B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: the existence of a alone, the existence of b alone, the existence of c alone, the simultaneous existence of a and b, the simultaneous existence of a and c, the simultaneous existence of b and c, or the simultaneous existence of a, b, and c, where a, b, and c can be single or multiple.

[0021] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0022] Based on the above, this application proposes an image recognition method and system for acupuncture treatment effects based on neuroprotection, aiming to improve the reliability of treatment plans.

[0023] The acupuncture treatment effect image recognition method based on neuroprotection provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms; the software can be an application that implements the acupuncture treatment effect image recognition method based on neuroprotection, but is not limited to the above forms.

[0024] This application can be applied to numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via communication networks. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices. It should be noted that in various specific embodiments of this invention, when processing is required based on data related to the characteristics of an object (e.g., user attributes or sets of attribute information), permission or consent from the corresponding object is obtained first, and the collection, use, and processing of this data comply with relevant laws and standards. Furthermore, when the embodiments of the present invention need to obtain the attribute information of an object, they will obtain the separate permission or separate consent of the corresponding object through pop-up windows or redirection to a confirmation page. After obtaining the separate permission or separate consent of the corresponding object, they will then obtain the relevant data of the object necessary for the embodiments of the present invention to operate normally.

[0025] See Figure 1 , Figure 1 This is a flowchart illustrating an image recognition method for acupuncture treatment effects based on neuroprotection, provided in one embodiment of this application. The method includes, but is not limited to, steps S110 to S160, which will be described in detail below.

[0026] Step S110: Receive the image and the user's individual information, and perform a quality assessment on the image to obtain the image quality assessment result; Step S120: Based on the image quality assessment results, perform adaptive physiological feature extraction on the image to obtain physiological texture features and artifact features of different levels of detail; Step S130: Divide the target region and reference region in the image, and perform contextual association analysis on the target region and reference region based on physiological texture features and artifact features of different levels of detail to obtain the physiological feature differences between the target region and reference region. Step S140: Based on the differences in physiological characteristics and combined with the preset physiological prior information, perform physiological prior verification on the differences in physiological characteristics to obtain reliable physiological characteristics. Step S150: Weighted integration of physiological texture features, artifact features, and reliable physiological features to obtain comprehensive physiological indicators; Step S160: Adjust the quantification threshold of the comprehensive physiological indicators according to the user's individual situation to obtain the quantified acupuncture treatment effect.

[0027] It should be noted that "images" generally refer to digital image data acquired through medical imaging equipment (such as visible light cameras, infrared thermal imagers, etc.) used to evaluate the effectiveness of acupuncture treatment. These images can reflect physiological information such as microcirculation, temperature distribution, and tissue texture on the skin surface. "Individual user information" refers to clinical data related to the patient, such as age, gender, medical history, chemotherapy regimen, and acupuncture treatment plan. This information is crucial for personalized evaluation of the effectiveness of acupuncture treatment.

[0028] "Physiological texture features" refer to the subtle texture information in an image that reflects the structure and functional state of biological tissues, such as the pores of the skin, the distribution of blood vessels, and the roughness of the tissue. These features can reflect the effects of acupuncture on local microcirculation and tissue activity.

[0029] "Artifact features" refer to non-physiological, false textures or structures introduced into an image due to the acquisition equipment or processing, such as edge halos caused by over-sharpening or noise particles. Identifying and distinguishing these artifacts is crucial for accurately extracting physiological features. "Target region" typically refers to the acupoint area stimulated by acupuncture or a specific physiological region related to neuroprotection. "Reference region" refers to an area adjacent to or symmetrical to the target region that has not been stimulated by acupuncture, used for comparative analysis to quantify the physiological changes caused by acupuncture treatment. "Physiological prior information" refers to a pre-established knowledge base about the physiological structure and function of the human body, such as blood vessel distribution patterns and tissue temperature ranges under normal physiological conditions, used to verify the extracted physiological features and improve their reliability. "Comprehensive physiological indicators" are quantitative indicators that integrate multiple physiological features (such as physiological texture features, artifact features, and reliable physiological features), comprehensively reflecting the overall effect of acupuncture treatment. "Quantification threshold" is a criterion used to convert comprehensive physiological indicators into specific treatment effect evaluation results; it can be adjusted according to the user's individual circumstances to achieve personalized evaluation.

[0030] In one embodiment, firstly, by performing image quality assessment, this application can identify and quantify problems such as blurring, noise, and artifacts in the image, providing a reliable foundation for subsequent feature extraction. This step is often overlooked but crucial in existing technologies. Secondly, this application introduces an "adaptive physiological feature extraction" mechanism based on image quality assessment results. This means that the system no longer simply applies a fixed feature extraction algorithm, but can dynamically adjust the extraction strategy according to the actual quality of the image. For example, when the image quality is poor, the system will use a more robust algorithm to suppress the influence of noise and artifacts, thereby ensuring that the extracted physiological texture features and artifact features are more reliable. This adaptability significantly improves the accuracy of assessment in complex clinical environments. Furthermore, this application greatly enhances the reliability of physiological feature differences by performing "contextual association analysis" on the target region and reference region, and combining "physiological prior information" for "physiological prior verification." Existing technologies often directly compare the features of the target region, while ignoring the comparative role of the reference region and its physiological rationality. This application, by introducing a reference region for comparison, can more accurately quantify the local physiological changes caused by acupuncture treatment. It also utilizes prior physiological knowledge to verify these changes, effectively eliminating interference from non-physiological or accidental factors, making the final reliable physiological characteristics more clinically significant. Finally, this application obtains a "comprehensive physiological index" by "weighted integration" of physiological texture features, artifact features, and reliable physiological characteristics. The "quantification threshold" is then adjusted according to the "user's individual situation," ultimately yielding a "quantified acupuncture treatment effect." This series of steps realizes the transformation from multi-dimensional features to a single quantitative index and further achieves personalized evaluation. Existing technologies often use single or limited indicators for evaluation and lack consideration for individual differences. This application, by integrating multiple features and adjusting the threshold according to the patient's individual situation, can more comprehensively and accurately reflect the actual effect of acupuncture treatment, providing clinicians with more instructive evaluation results. This allows for better guidance of personalized treatment plans, minimizing the toxicity of chemotherapy drugs to the patient's nervous system, and fully leveraging the potential of acupuncture as a neuroprotective measure.

[0031] It's important to note that dividing an image into local regions aims to break down a complex overall image into smaller, more easily analyzed and processed units. This division can be based on a pre-defined grid, image content (e.g., physiological structure density), or a region of interest. Local region division allows for refined analysis tailored to the characteristics of different regions, avoiding information confusion that may arise from global feature extraction. Continuously acquiring multiple image frames for each local region means acquiring multiple images consecutively for each defined local region over a period of time. This sequential acquisition method can capture dynamic information about local regions over time, such as subtle physiological fluctuations, tissue deformation, or changes in blood flow. In practical applications, identifying transient motion patterns in each local region based on the multi-frame image sequence involves analyzing pixel or feature point displacements between consecutive image frames to determine minute, rapid movements occurring within the local region. These transient motion patterns may include tissue tremors, vascular pulsations, or skin elastic deformation, which are important indicators reflecting physiological states. Identification methods can employ techniques such as optical flow, feature point tracking, or deep learning-based motion estimation. Furthermore, based on transient motion patterns and image quality assessment results, adaptive physiological feature extraction is performed on local regions to obtain transient deformation features. This means that the feature extraction process is dynamically adjusted according to the detected transient motion patterns and image quality. For example, if the image quality assessment results show that a certain local region has a lot of noise or blur, the parameters of the feature extraction algorithm can be adjusted to enhance the ability to capture weak physiological signals; if significant non-physiological motion is detected, compensation or filtering can be performed to ensure that the extracted deformation features mainly originate from physiological activities. Transient deformation features can include changes in the shape, size, and texture of local regions over a short period of time. Thus, transient deformation features are distinguished to obtain physiological texture features and artifact features of different levels of detail. This step aims to further classify the extracted transient deformation features. Physiological texture features refer to texture information with different spatial frequencies and directions related to real physiological activities, such as skin microcirculation and dynamic changes in capillary networks. Artifact features refer to deformations or textures caused by non-physiological factors (such as equipment noise, ambient light fluctuations, and slight involuntary movements of the patient) that may interfere with the judgment of physiological features. By distinguishing between them, interference can be effectively filtered out, improving the accuracy of subsequent analysis.

[0032] In one embodiment, it is assumed that the therapeutic effect of acupuncture on a specific acupoint area needs to be evaluated. First, the image containing the acupoint is divided into multiple local regions, for example, the skin area surrounding the acupoint is divided into a 3x3 grid centered on the acupoint. Then, for each grid region, a multi-frame image sequence is continuously acquired using a high-speed camera to capture the microscopic dynamic changes in that region. Next, by analyzing these image sequences, for example using an optical flow-based algorithm, transient motion patterns such as minute pulsations, tremors, or deformations of the skin within each local region are identified. Simultaneously, combined with the image quality assessment results (e.g., if a region is slightly blurred, the weight of edge information is increased during feature extraction), adaptive physiological feature extraction is performed on each local region to obtain the transient deformation features of that region, such as local brightness or texture changes caused by the dilation or contraction of skin microvessels. Finally, a machine learning classifier is used to distinguish these transient deformation features. Features related to physiological activities such as blood flow and tissue elasticity are classified as physiological texture features, while features caused by camera shake, changes in ambient light, etc., are classified as artifact features, thus obtaining physiological texture features and artifact features of different levels of detail.

[0033] It's important to clarify that assessing the physiological structure density of an image involves analyzing its content to quantify the density of physiological structures (such as blood vessels, nerves, and tissue textures). Specifically, image processing techniques, such as edge detection, texture analysis, or morphological operations, can be used to identify and quantify the distribution of physiological structures in the image, thus obtaining the physiological structure density. This aims to provide a basis for subsequent local region segmentation, ensuring that the segmented regions better reflect the actual distribution of physiological structures. Furthermore, adjusting the segmentation size based on the physiological structure density means dynamically determining the size used for local region segmentation based on the assessed physiological structure density. For example, a smaller segmentation size can be used in regions with high physiological structure density to capture more refined physiological features; while a larger segmentation size can be used in regions with low physiological structure density to improve processing efficiency and avoid over-refinement. This yields the target segmentation size. Its purpose is to make local region segmentation more adaptive and refined, avoiding information loss or redundancy caused by fixed-size segmentation. Finally, segmenting the image locally based on the target segmentation size means dividing the entire image into multiple local regions of different sizes and shapes according to the adjusted target segmentation size. The boundaries and sizes of these local regions will be optimized based on the density distribution of their internal physiological structures, thereby ensuring that each local region can effectively represent the physiological information it contains.

[0034] In one embodiment, it is assumed that an image of a specific acupoint region after a patient has received acupuncture treatment needs to be analyzed. First, the physiological structure density of the acupoint region image is assessed. For example, image processing algorithms can be used to identify the density of capillary networks, sweat gland openings, or skin texture in the image. If the capillary network in a local area (e.g., the center of the acupoint) is very dense, the physiological structure density of that area is assessed as high. Conversely, if the skin texture in a local area (e.g., the edge of the acupoint) is relatively smooth and the blood vessels are sparse, the physiological structure density of that area is assessed as low. Based on this physiological structure density assessment result, the system dynamically adjusts the segmentation size of the local regions. Specifically, for the acupoint center region with high physiological structure density, the system uses a smaller segmentation size to ensure that minute changes in vasodilation, vasoconstriction, or skin microcirculation can be captured. For the acupoint edge region with low physiological structure density, the system uses a larger segmentation size to improve processing efficiency while still effectively covering the physiological information of that area. Finally, based on these adjusted target segmentation sizes, the entire acupoint region image is segmented locally to obtain a series of local regions of varying sizes but with greater physiological relevance. These local areas were then used to continuously acquire multi-frame image sequences and perform transient motion pattern recognition and adaptive physiological feature extraction, thereby more accurately reflecting the local physiological response caused by acupuncture treatment.

[0035] It should be noted that receiving clinical feedback information refers to the system acquiring actual observation and evaluation data regarding the effectiveness of acupuncture treatment from doctors, patients, or relevant medical records. This information can include the patient's subjective feelings, such as the degree of pain relief and functional improvement, as well as the doctor's objective assessment, such as changes in physical signs and improvement in lesions. It can also include other auxiliary examination results. Dynamically adjusting weights based on clinical feedback information means that the system intelligently adjusts the relative importance of physiological texture features, artifact features, and reliable physiological features in the calculation of comprehensive physiological indicators based on the received clinical feedback information. For example, if clinical feedback indicates that a certain physiological texture feature is highly correlated with the treatment effect, the weight of that feature can be increased; conversely, if a certain artifact feature has a negative impact on the evaluation results, its weight can be decreased. This adjustment is dynamic and can be personalized according to different patients, different treatment stages, or different disease types to obtain target weights that better reflect the current clinical reality. Weighted integration of physiological texture features, artifact features, and reliable physiological features based on target weights means applying the dynamically adjusted target weights to each physiological feature to calculate the final comprehensive physiological indicator. For example, by using linear weighted summation, each feature value can be multiplied by its corresponding target weight and then summed to obtain a quantitative indicator that can more accurately reflect the effect of acupuncture treatment.

[0036] In one embodiment, a patient receives acupuncture treatment to relieve chronic pain. Before treatment, the system acquires images and extracts physiological texture features, artifact features, and reliable physiological features. During treatment, the doctor and patient provide clinical feedback, such as the patient reporting a decrease in pain score from 8 to 3, or the doctor observing an increase in local skin temperature and improved microcirculation. Upon receiving this clinical feedback, the system analyzes which physiological features are most correlated with these changes. For example, if increased skin temperature and improved microcirculation are found to be highly consistent with pain relief, the system dynamically increases the weight of physiological texture features such as "skin temperature change" and "microcirculation activity," while potentially decreasing the weight of artifact features not directly related to pain relief. Therefore, in subsequent calculations of comprehensive physiological indicators, these features highly correlated with clinical efficacy will have a greater weight, allowing the final quantified acupuncture treatment effect to more accurately reflect the patient's actual improvement. This dynamic adjustment ensures the clinical applicability and personalization of the assessment results.

[0037] It should be noted that acquiring visible light and infrared images aims to obtain image information from different modalities. Visible light images provide information such as the macroscopic morphology and texture details of the skin surface, while infrared images reflect physiological activities such as temperature distribution and blood perfusion in subcutaneous tissue. Combining these two modalities allows for a more comprehensive and richer understanding of physiological structures. Furthermore, local texture features are extracted from the visible light images to capture fine texture information related to physiological structures, such as skin pores and microvascular networks. These local texture features can be implemented using various image processing algorithms, such as Gabor filters, Local Binary Patterns (LBP), or deep learning feature extraction networks. Simultaneously, temperature distribution analysis is performed on the infrared images to obtain temperature distribution features, quantifying temperature differences between different regions of the image. Physiological structures, especially blood vessels and inflamed areas, typically exhibit specific temperature distribution patterns. Temperature distribution features can be obtained through thermogram analysis and temperature gradient calculation. Therefore, based on local texture features and temperature distribution features, preliminary identification of physiological structures in the images is performed, yielding preliminary identification results. This step involves fusing features from both visible and infrared modes and using machine learning models or rule-based algorithms to perform preliminary classification or segmentation of physiological structures in the image. For example, a classifier can be trained to determine whether a pixel belongs to a physiological structure based on texture and temperature features.

[0038] In one embodiment, morphological filtering is applied to the preliminary identification results to obtain a processed image. Morphological filtering, such as opening, closing, dilation, or erosion operations, is used to remove noise, smooth structural edges, and fill small holes or breaks in the preliminary identification results, thereby optimizing the accuracy and consistency of physiological structure recognition. Subsequently, connected component analysis is performed on the processed image to obtain multiple initial connected regions. Connected component analysis aims to identify interconnected sets of pixels in the processed image as independent regions, each region potentially representing a physiological structure. Further, physiological feature verification is performed on each initial connected region to obtain target connected regions. This verification process aims to filter the initial connected regions based on pre-defined physiological prior knowledge, such as the size range, shape features, and location distribution of physiological structures, excluding regions that do not conform to physiological features, thereby ensuring that the identified regions are real physiological structures. Finally, the area ratio of the target connected region within the preset region is calculated to obtain the physiological structure density. This area ratio can intuitively reflect the density of physiological structures within a specific region, providing a quantitative basis for subsequent local region division.

[0039] It should be noted that the preset multiple time points can be understood as acquiring images continuously or intermittently during the image acquisition process, at a certain frequency or in a specific clinical observation stage. For example, several frames of images can be acquired per second, or at key moments such as before, during, and after acupuncture treatment. This results in a series of images arranged in chronological order, i.e., the target time-series images. These image sequences can reflect the dynamic changes of the observed area over a period of time. Inter-frame consistency analysis of the target time-series images refers to evaluating the stability of the image sequence by comparing the similarity or differences between adjacent frames or multiple frames. For example, techniques such as pixel-level difference calculation, feature point matching, or optical flow can be used to quantify the degree of change of image content over time. This yields the inter-frame consistency deviation, which reflects the inconsistencies in the image sequence caused by factors such as motion, changes in illumination, and noise. Furthermore, based on the inter-frame consistency deviation, systematic deviations can be obtained. Systematic deviations refer to deviations with certain regularities caused by the equipment itself, environmental factors, or non-physiological movements of the observed object during the image acquisition process. For example, camera shake, fluctuations in ambient lighting, or slight patient positional movements can all cause inter-frame consistency deviations to exhibit certain patterns. By performing statistical analysis, filtering, or pattern recognition on these deviations, systematic biases can be identified and quantified. Ultimately, a quality assessment result can be obtained based on these systematic biases. The quality assessment result is a comprehensive judgment of the overall quality of the image sequence, which can be a quantified score, a classification, or a binary judgment indicating usability. For example, when systematic bias exceeds a preset threshold, the image quality can be considered poor and unsuitable for subsequent physiological feature extraction. This quality assessment result provides a reliable input basis for subsequent physiological feature extraction.

[0040] In one embodiment, during image acquisition, a series of images are first continuously received at multiple preset time points, such as at a fixed sampling frequency or when triggered by a specific event, thereby forming an original time-series image. This original time-series image contains visual information of physiological regions captured at different time points. Illumination correction is performed on the original time-series image to eliminate or reduce the impact of changes in ambient lighting (such as brightness, color temperature, and shadows) on image quality. Illumination correction ensures that the image sequence has relatively consistent lighting conditions at different time points, thus avoiding the misidentification of lighting differences as physiological changes. In practical applications, illumination correction can employ various techniques, such as histogram equalization, gamma correction, the Retinex algorithm, or deep learning models for illumination normalization. Furthermore, motion trajectory analysis is performed on the sequence of each frame in the illumination-corrected time-series image to identify and quantify pixel displacement or deformation caused by minute user movements (such as breathing, heartbeat, muscle tremors, or slight posture adjustments). Local motion trajectory analysis can be achieved using techniques such as optical flow, feature point tracking, or block-matching-based motion estimation to accurately capture the motion patterns of various regions in the image. Based on the obtained local motion trajectories, nonlinear or non-uniform deformation patterns can be further analyzed and identified. Nonlinear deformation patterns refer to inconsistent motion in different regions of the image, such as rotation or distortion in some areas, while non-uniform deformation patterns refer to uneven spatial distribution of deformation. These patterns reflect the complexity of user physiological activities or involuntary movements. Deformation compensation is then performed based on the obtained nonlinear or non-uniform deformation patterns. The purpose of deformation compensation is to correct each frame in the image sequence back to a unified reference coordinate system through methods such as inverse transformation or interpolation, thereby eliminating or significantly reducing motion-induced deformation. The compensated sequence of each frame will be more stable, reducing motion artifacts. Finally, the compensated sequences of each frame are superimposed to obtain the target time-series image. The overlay operation can integrate the corrected image information from multiple frames, such as by averaging, median filtering, or weighted averaging, to further improve the signal-to-noise ratio of the image and generate a high-quality image sequence that is more stable in time and more consistent in space, which can then be used as input for subsequent quality assessment and physiological feature extraction.

[0041] In one embodiment, it is assumed that when evaluating the effect of acupuncture treatment on a user, it is necessary to acquire image sequences of specific acupoint areas. During the acquisition process, due to slight fluctuations in ambient lighting in the clinic, and the user's involuntary slight breathing movements or muscle tremors when remaining still for a long time, the image acquisition device first continuously receives images of the acupoint area according to a preset sampling frequency (e.g., 30 frames per second), obtaining the original time-series image. Next, to eliminate the influence of lighting fluctuations, the system performs illumination correction on the original time-series image. For example, a method based on local adaptive histogram equalization can be used to standardize the brightness distribution of each frame image, obtaining an illumination-corrected time-series image. Subsequently, to handle the user's slight movements, the system performs motion trajectory analysis on the sequence of each frame image in the illumination-corrected time-series image. For example, the Lucas-Kanade optical flow algorithm can be used to calculate the displacement vectors of key feature points in the image between consecutive frames, thereby obtaining local motion trajectories. Based on these local motion trajectories, the system can further identify nonlinear or non-uniform deformation patterns, such as the periodic expansion and contraction of skin areas caused by respiration.

[0042] The system performs deformation compensation based on identified nonlinear or non-uniform deformation patterns. For example, a deformation model based on thin-plate splines (TPS) can be used to geometrically correct each frame of the image, aligning it with a reference frame in the sequence, thus obtaining a sequence of compensated images. Finally, these compensated image sequences are superimposed, for example, by calculating the average pixel value of all compensated frames, to obtain the final target time-series image. This target time-series image will have higher stability and fewer illumination and motion artifacts, providing a reliable data foundation for subsequent quality assessment and physiological feature extraction.

[0043] It should be noted that acquiring the illumination intensity distribution of the environment in which the original time-series images were captured refers to obtaining brightness information of the image shooting environment through sensors or image analysis techniques. Ambient lighting conditions can be quantified using the image's average pixel values, brightness histograms, or dedicated lighting sensors. Furthermore, identifying non-uniform illumination regions and transient illumination fluctuation regions in the original time-series images based on the illumination intensity distribution means distinguishing between two main types of lighting problems in the image based on the acquired illumination intensity distribution data. Non-uniform illumination regions typically manifest as large local brightness differences in the image, such as shadows or local overexposure; transient illumination fluctuation regions manifest as rapid brightness changes in the image sequence over time. Identification methods can include spatial gradient analysis, local contrast calculation, and inter-frame brightness difference analysis. Performing local illumination compensation on non-uniform illumination regions to obtain compensated local illumination regions refers to adjusting the brightness of areas with inconsistent brightness in the image using locally adaptive algorithms. For example, local histogram equalization, gamma correction, or Retinex-based algorithms can be used to eliminate local shadows or highlights, making the brightness distribution of the image more uniform. The aim is to improve the spatial consistency of the image. Simultaneously, temporal illumination smoothing is performed on areas of instantaneous illumination fluctuation to obtain smoothed temporal illumination regions. This refers to processing brightness jumps caused by unstable light sources in the image sequence along the time dimension. For example, moving average filtering, Kalman filtering, or optical flow-based temporal smoothing algorithms can be used to reduce or eliminate inter-frame brightness fluctuations, ensuring the continuity and stability of the image sequence. The aim is to improve the temporal consistency of the image. Finally, the compensated local illumination regions and the smoothed temporal illumination regions are fused to obtain the illumination-corrected time-series image. This involves integrating the results of spatial local compensation and temporal smoothing. The fusion process can employ weighted averaging, multi-scale fusion, or machine learning-based methods, comprehensively utilizing the two correction results to generate the final illumination-corrected time-series image. The purpose is to comprehensively address illumination problems that may be encountered during image acquisition, providing high-quality input for subsequent physiological feature extraction.

[0044] It should be noted that deformation features refer to changes in pixel position, shape, or brightness in an image caused by physiological activities (such as pulse or respiration) or external disturbances (such as slight shaking). These features can be extracted using various image processing techniques, such as optical flow, block matching, feature point tracking (such as SIFT, SURF, ORB, etc.), or deep learning models for displacement estimation of feature points or regions. The aim is to quantify the various motion and deformation information present in an image sequence. Physiological verification can be understood as checking the biological plausibility of the extracted deformation features. Specifically, thresholds or models can be set based on known physiological laws (e.g., deformation amplitude, frequency range, deformation direction, etc. in a specific region). For example, drastic, random, or non-periodic deformations inconsistent with physiological activities such as heartbeat or respiration can be excluded, or deformations exceeding normal physiological ranges can be excluded. The purpose is to filter out non-physiological deformations to ensure the accuracy of subsequent motion trajectory analysis. In practical applications, motion trajectory analysis is performed on each frame of a time-series image after illumination correction, based on verified deformation features. Specifically, this involves using physiologically verified deformation information to construct the motion path of various local regions or key points in the image over time. For example, the positional changes of specific physiological structures (such as blood vessels and skin texture) in consecutive frames can be tracked to obtain their motion trajectory in two-dimensional or three-dimensional space. The aim is to accurately capture the true motion patterns reflecting physiological activities in the image. This application's solution addresses the problem of non-physiological noise in traditional motion trajectory analysis by introducing deformation feature extraction and physiological verification. First, by extracting deformation features from the illumination-corrected time-series image, various motion and deformation information present in the image can be comprehensively captured. Physiological verification of these deformation features effectively identifies and eliminates non-physiological deformations, such as those caused by external interference or image noise, ensuring that the analyzed deformation features truly reflect physiological activities. It is precisely this pre-screening by physiology that allows subsequent motion trajectory analysis to focus more on physiological movements, improving the accuracy and reliability of the motion trajectory.

[0045] In one embodiment, a series of light-corrected facial images are acquired when evaluating the efficacy of acupuncture treatment in a patient. To accurately analyze facial microcirculation or physiological muscle movement, deformation features are first extracted from these images. For example, pixel displacement vectors between adjacent frames can be calculated using the Lucas-Kanade optical flow method to obtain preliminary deformation features. Subsequently, these deformation features are physiologically verified. Specifically, a physiological frequency range can be set (e.g., pulse rate of 1-2 Hz, respiratory rate of 0.2-0.5 Hz), and severe or abnormal deformations exceeding this frequency range can be filtered out. Furthermore, based on anatomical knowledge of facial physiological structures, deformation patterns that should not appear in specific areas can be excluded. For example, if rapid, large-amplitude deformations in the eye area that are inconsistent with the blinking frequency are detected, they may be marked as non-physiological deformations. After physiological verification, only those deformation features that conform to physiological laws are retained. Finally, based on these verified deformation features, motion trajectory analysis is performed on specific regions (such as the skin around acupoints) in the image sequence to obtain the real, weak physiological local motion trajectory of the region under acupuncture stimulation, such as the pulsation trajectory of skin microvessels or the micro-tremor trajectory of muscles.

[0046] See Figure 2 , Figure 2 This is a schematic diagram of an image recognition system for acupuncture treatment effects based on neuroprotection, provided in one embodiment of this application. The image recognition system 200 for acupuncture treatment effects based on neuroprotection includes: The receiving and evaluation module 210 is used to receive images and individual information of users, and to perform quality evaluation on the images to obtain the image quality evaluation results. The feature extraction module 220 is used to perform adaptive physiological feature extraction on the image based on the image quality assessment results, and obtain physiological texture features and artifact features of different levels of detail. The analysis module 230 is used to divide the target region and reference region in the image, and perform contextual correlation analysis on the target region and reference region based on physiological texture features and artifact features of different levels of detail, so as to obtain the physiological feature differences between the target region and the reference region. The verification module 240 is used to perform physiological prior verification on physiological feature differences based on physiological feature differences and in combination with preset physiological prior information to obtain reliable physiological features. The weighted integration module 250 is used to weight and integrate physiological texture features, artifact features and reliable physiological features to obtain comprehensive physiological indicators; The adjustment module 260 is used to adjust the quantitative threshold of comprehensive physiological indicators according to the user's individual situation, so as to obtain the quantitative acupuncture treatment effect.

[0047] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0048] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically include computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0049] The foregoing has provided a detailed description of the preferred embodiments of this application. However, this application is not limited to the above-described embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined in this application.

Claims

1. A method for image recognition of acupuncture treatment effects based on neuroprotection, characterized in that, include: Receive images and individual user information, perform quality assessment on the images, and obtain the quality assessment results of the images; Based on the quality assessment results of the image, adaptive physiological feature extraction is performed on the image to obtain physiological texture features and artifact features of different levels of detail. The target region and reference region in the image are divided, and based on the physiological texture features of different fineness and the artifact features, the context association analysis of the target region and the reference region is performed to obtain the physiological feature differences between the target region and the reference region. Based on the differences in physiological characteristics and combined with preset physiological prior information, physiological prior verification is performed on the differences in physiological characteristics to obtain reliable physiological characteristics. The physiological texture features, artifact features, and reliable physiological features are weighted and integrated to obtain a comprehensive physiological index. The quantitative threshold of the comprehensive physiological indicators is adjusted according to the individual circumstances of the user to obtain a quantitative acupuncture treatment effect.

2. The method according to claim 1, characterized in that, Based on the quality assessment results of the image, adaptive physiological feature extraction is performed on the image to obtain physiological texture features and artifact features of different levels of detail, including: The image is divided into local regions to obtain multiple local regions; For each of the aforementioned local regions, multiple frames of image sequences are continuously acquired; Based on the multi-frame image sequence, identify the transient motion pattern of each local region; Based on the transient motion pattern and the quality assessment results of the image, adaptive physiological feature extraction is performed on the local region to obtain the transient deformation features of the local region. The transient deformation features are distinguished to obtain physiological texture features and artifact features of different levels of detail.

3. The method according to claim 2, characterized in that, The image is divided into local regions to obtain multiple local regions, including: The physiological structure density of the image is evaluated to obtain the physiological structure density; The division size is adjusted according to the physiological structure density to obtain the target division size; The image is divided into local regions based on the target segmentation size to obtain multiple local regions.

4. The method according to claim 1, characterized in that, The weighted integration of the physiological texture features, the artifact features, and the reliable physiological features yields a comprehensive physiological index, including: Receive clinical feedback information; Based on the clinical feedback information, the weights of the physiological texture features, artifact features, and reliable physiological features related to the clinical feedback information are dynamically adjusted to obtain the target weights; The physiological texture features, artifact features, and reliable physiological features are weighted and integrated according to the target weights to obtain a comprehensive physiological index.

5. The method according to claim 3, characterized in that, The step of evaluating the physiological structure density of the image to obtain the physiological structure density includes: Acquire the visible light image and infrared image from the image; Local texture features are extracted from the visible light image to obtain local texture features; Temperature distribution analysis was performed on the infrared image to obtain temperature distribution characteristics; Based on the local texture features and the temperature distribution features, the physiological structures in the image are preliminarily identified, and a preliminary identification result is obtained. The preliminary recognition results are subjected to morphological filtering to obtain the processed image; The processed image is subjected to connected component analysis to obtain multiple initial connected components; Physiological feature verification is performed on each of the initial connected regions to obtain the target connected regions; The area ratio of the target connected region within the preset region is calculated to obtain the physiological structure density.

6. The method according to claim 1, characterized in that, The process of performing a quality assessment on the image to obtain the image quality assessment result includes: The images are received at multiple preset time points to obtain a target time series image; Inter-frame consistency analysis is performed on the target time series image to obtain the inter-frame consistency deviation; Based on the inter-frame consistency deviation, the systematic deviation is obtained; The quality assessment results are obtained based on the systematic deviations described.

7. The method according to claim 6, characterized in that, The step of receiving the images according to multiple preset time points to obtain a target time series image includes: The images are received at multiple preset time points to obtain the original time series images; The original time-series image is subjected to illumination correction to obtain an illumination-corrected time-series image; Motion trajectory analysis is performed on the sequence of each frame in the illumination-corrected time-series image to obtain the local motion trajectory; Based on the local motion trajectory, a nonlinear or non-uniform deformation mode is obtained; Deformation compensation is performed based on the nonlinear or non-uniform deformation mode to obtain a sequence of compensated images for each frame. The sequence of each compensated image frame is superimposed to obtain the target time series image.

8. The method according to claim 7, characterized in that, The step of performing illumination correction on the original time-series image to obtain an illumination-corrected time-series image includes: The illumination intensity distribution of the environment in which the original time-series images are located is collected to obtain the illumination intensity distribution; Based on the light intensity distribution, identify non-uniform illumination regions and instantaneous light fluctuation regions in the original time series image; Local illumination compensation is performed on the non-uniform illumination area to obtain the compensated local illumination area; The instantaneous illumination fluctuation region is smoothed in the time domain to obtain a smoothed time domain illumination region. The compensated local illumination region and the smoothed temporal illumination region are fused to obtain an illumination-corrected time-series image.

9. The method according to claim 7, characterized in that, The step of performing motion trajectory analysis on each frame of the illumination-corrected time-series image to obtain local motion trajectories includes: Deformation features are extracted from the illumination-corrected time-series images to obtain deformation features; The deformation characteristics are subjected to physiological verification to obtain deformation characteristics that pass the verification. Based on the verified deformation features, motion trajectory analysis is performed on the sequence of each frame in the illumination-corrected time-series image to obtain the local motion trajectory.

10. An image recognition system for acupuncture treatment effects based on neuroprotection, characterized in that, include: The receiving and evaluation module is used to receive images and individual information about users, and to perform quality evaluation on the images to obtain the quality evaluation results of the images; The feature extraction module is used to perform adaptive physiological feature extraction on the image based on the quality assessment results of the image, so as to obtain physiological texture features and artifact features of different levels of detail. The analysis module is used to divide the target region and the reference region in the image, and based on the physiological texture features of different fineness and the artifact features, to perform contextual association analysis on the target region and the reference region to obtain the physiological feature differences between the target region and the reference region. The verification module is used to perform physiological prior verification on the physiological feature differences based on the physiological feature differences and in combination with preset physiological prior information to obtain reliable physiological features. The weighted integration module is used to weight and integrate the physiological texture features, the artifact features, and the reliable physiological features to obtain a comprehensive physiological index. The adjustment module is used to adjust the quantification threshold of the comprehensive physiological indicators according to the individual circumstances of the user, so as to obtain a quantified acupuncture treatment effect.

Images