Electronic nasopharyngolaryngoscopy management method and system based on artificial intelligence

Abstract

The embodiment of the specification discloses an artificial intelligence-based electronic nasopharyngeal laryngoscopy examination management method, which comprises the following steps of: after acquiring an image to be examined every time, executing examination operation on the image to be examined, namely judging whether the image quality of the image to be examined meets the requirement or not; if yes, whether the to-be-examined image corresponds to a certain nose and throat preset site or not is judged; if the to-be-examined image corresponds to a certain nose and throat preset site, taking the to-be-examined image as a target image corresponding to the nose and throat preset site; or judging whether the to-be-examined image corresponds to a certain nose and throat preset site or not; if the to-be-examined image corresponds to a certain nose and throat preset site, judging whether the image quality of the to-be-examined image meets the requirement or not; if the to-be-examined image meets the requirement, taking the to-be-examined image as a target image corresponding to the nose and throat preset site; and if the nose and throat preset station where the corresponding target image is not obtained exists, generating guidance information for indicating the collection of the images corresponding to the remaining stations and the image collection rule.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an artificial intelligence-based method and system for managing electronic nasopharyngoscopy examinations. Background Technology

[0002] Electronic laryngoscopy is a commonly used examination method in otolaryngology to observe the anatomical structures and lesions of the pharynx and larynx. It is a key tool for diagnosing nasopharyngeal diseases, especially screening for early malignant tumors (such as squamous cell carcinoma of the head and neck), in otolaryngology-head and neck surgery. However, in traditional electronic laryngoscopy, the examination procedure and image acquisition often rely on the doctor's personal experience and habits, lacking standardized procedures. The quality of this examination largely depends on the operator's personal experience and habits, and for a long time, there has been a lack of objective, unified quality control standards and real-time monitoring mechanisms. This non-standardized operating mode easily leads to problems such as omissions of key anatomical sites (i.e., the formation of "blind spots"), incomplete or non-standardized images, and inconsistent image acquisition quality, thus affecting the accuracy of diagnosis and causing missed diagnoses of early lesions. Existing studies have warned that the missed diagnosis rate of early hypopharyngeal cancer by routine electronic nasopharyngoscopy can be as high as 84.6%, highlighting the extreme urgency of standardizing the examination procedure and implementing full-process quality control.

[0003] However, existing technologies still have the following three core shortcomings: 1. Currently, there is no effective method to supervise and evaluate the quality of laryngoscopy, nor is there a quantitative assessment to confirm whether the physician has conducted a comprehensive examination of each anatomical site. This examination model, which relies on personal experience, is prone to missed or misdiagnosed important lesions (especially early-stage tumors), which may have serious consequences for patient prognosis and function.

[0004] 2. The lack of a real-time quality control mechanism for the examination process makes it impossible to ensure that all key anatomical sites are observed and recorded. Industry guidelines such as expert consensus remain at the macro-level of recommendations, lacking effective technical tools and specific implementation methods to translate them into standardized operating procedures that can be monitored and quantified in real-time in clinical practice.

[0005] 3. Post-inspection analysis rather than in-process quality control: Existing artificial intelligence applications mainly focus on offline analysis of acquired images or lesion characterization. Real-time quality control during the inspection process, such as automatically identifying the current inspection site, evaluating image clarity in real time, dynamically monitoring the integrity of the inspection and providing immediate feedback, is still a technological gap.

[0006] 4. Technology isolation rather than system integration: Currently, there is no intelligent inspection and quality control solution that can both perform standardized and normalized inspections and organically combine standardized and normalized inspections with real-time artificial intelligence analysis and closed-loop feedback functions.

[0007] Therefore, there is an urgent need in this field for an innovative technical solution that can overcome the above-mentioned defects and achieve intelligent and standardized quality control throughout the entire process of electronic nasopharyngoscopy. Summary of the Invention

[0008] This specification provides an AI-based electronic nasopharyngoscopy examination management method and system to address the technical problem of how to conduct full-process, intelligent, and standardized quality control of electronic nasopharyngoscopy examinations.

[0009] To address the aforementioned technical problems, the embodiments in this specification provide the following technical solutions: This specification provides an embodiment of an artificial intelligence-based electronic nasopharyngoscopy examination management method, which is applied to a computer and includes: After each image to be examined is acquired, the artificial intelligence model on the computer performs the following operations: An inspection operation is performed on the image to be inspected; wherein the inspection operation includes: determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, then using the image to be inspected as the target image corresponding to the nasopharyngeal preset site; or, the inspection operation includes: determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, then using the image to be inspected as the target image corresponding to the nasopharyngeal preset site. After each examination of the images to be examined, determine whether the corresponding target images have been acquired at each preset nasopharyngeal site; If there are nasopharyngeal preset sites for which the corresponding target image has not yet been acquired, guidance information corresponding to the remaining sites is generated. The guidance information is used to instruct the acquisition of the image corresponding to the remaining sites and to instruct the image acquisition rules corresponding to the next nasopharyngeal preset site. The remaining sites are nasopharyngeal preset sites for which the corresponding target image has not yet been acquired.

[0010] Preferably, before acquiring the image to be examined, the method further includes: Continuously acquire images from the nasopharyngeal imaging acquisition device, and determine whether the image acquisition conditions of a certain preset nasopharyngeal site are met based on the acquired images. If so, a collection prompt message is generated, which prompts the image to be collected at the target site; wherein, the target site is a preset nasopharyngeal site that meets the image collection conditions.

[0011] Preferably, for any pre-defined nasopharyngeal site, the image acquisition conditions for that pre-defined nasopharyngeal site include: The nasopharyngeal imaging acquisition device has reached the image acquisition area of ​​the preset nasopharyngeal station; And / or, The imaging range of the nasopharyngeal imaging acquisition device covers the image acquisition area of ​​the preset nasopharyngeal site.

[0012] As a preferred option, the image acquisition conditions for the preset nasopharyngeal site also include: The target image corresponding to the preset nasopharyngeal site has not yet been obtained.

[0013] Preferably, the acquisition prompt information also includes the image acquisition rules for the target site.

[0014] Preferably, the method further includes: The newly acquired images after each acquisition prompt message is generated are used as the images to be examined.

[0015] Preferably, before acquiring the image to be examined, the method further includes: Generate first guidance information corresponding to the first site. The first guidance information is used to instruct the acquisition of images corresponding to the first site and to instruct the image acquisition rules corresponding to the first site. The first site is the nasopharyngeal preset site that is ranked first.

[0016] Preferably, the method further includes: For each nasopharyngeal preset site, a reference image for that nasopharyngeal preset site is preset. The reference image is used to determine whether the image to be examined corresponds to the nasopharyngeal preset site.

[0017] Preferably, the nasopharyngeal preset station corresponds to a specific area of ​​the nasopharynx; And / or, The next nasopharyngeal preset station is the nasopharyngeal preset station that is ranked first among the remaining stations.

[0018] This specification provides an embodiment of an artificial intelligence-based electronic nasopharyngoscopy examination and management system, the system comprising: The image acquisition unit is used to acquire images to be inspected. An artificial intelligence processing unit is used to run an artificial intelligence model and perform an inspection operation on the image to be inspected; wherein the inspection operation includes: determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, then using the image to be inspected as the target image corresponding to the nasopharyngeal preset site; or, the inspection operation includes: determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, then using the image to be inspected as the target image corresponding to the nasopharyngeal preset site. After each examination of the images to be examined, determine whether the corresponding target images have been acquired at each preset nasopharyngeal site; If there are nasopharyngeal preset sites for which the corresponding target image has not yet been acquired, guidance information corresponding to the remaining sites is generated. The guidance information is used to instruct the acquisition of the image corresponding to the remaining sites and to instruct the image acquisition rules corresponding to the next nasopharyngeal preset site. The remaining sites are nasopharyngeal preset sites for which the corresponding target image has not yet been acquired.

[0019] The above-described at least one technical solution adopted in the embodiments of this specification can achieve the following beneficial effects: The above-mentioned technical solution uses artificial intelligence technology to standardize the examination process (including the image acquisition process during electronic nasopharyngoscopy) through pre-designed standard examination stations, and ensures that each electronic nasopharyngoscopy examination covers all necessary anatomical sites and acquires high-quality target images according to unified standards. This improves the comprehensiveness, standardization, intelligence and standardization of nasopharyngeal examination, reduces missed diagnoses and misdiagnoses, improves the effectiveness and efficiency of nasopharyngeal examination, and thus improves the effectiveness and efficiency of nasopharyngeal diagnosis. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings used in the description of the embodiments of this specification or the prior art will be briefly described below. Obviously, the drawings used in some embodiments of this application are only described below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the artificial intelligence-based electronic nasopharyngoscopy examination management method provided in the first embodiment of this specification.

[0022] Figure 2 This is a schematic diagram of a nasopharyngeal pre-set site in the first embodiment of this specification.

[0023] Figure 3 This is a schematic diagram of another nasopharyngeal pre-set site in the first embodiment of this specification.

[0024] Figure 4 This is a schematic diagram of an interactive interface in the first embodiment of this specification.

[0025] Figure 5 This is a schematic diagram of another interactive interface in the first embodiment of this specification.

[0026] Figure 6 This is a schematic diagram of the model architecture in the first embodiment of this specification.

[0027] Figure 7 This is a schematic diagram of an electronic nasopharyngoscopy examination management process in the first embodiment of this specification.

[0028] Figure 8 This is a schematic diagram illustrating the experimental results of the artificial intelligence model in the first embodiment of this specification.

[0029] Figure 9 This is a schematic diagram of the system operation process in the second embodiment of this specification.

[0030] Figure 10 This is a schematic diagram of the system architecture in the second embodiment of this specification.

[0031] Figure 11 This is a schematic diagram of another system operation process in the second embodiment of this specification. Detailed Implementation

[0032] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments involved in the specific implementation are only a part of the embodiments of this application, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments in the specific implementation without creative effort should fall within the protection scope of this application.

[0033] The first embodiment of this specification (hereinafter referred to as "Embodiment 1") provides an artificial intelligence-based electronic nasopharyngoscopy examination management method. The executing entity of Embodiment 1 includes, but is not limited to, a terminal, a server, an operating system, or an application; that is, the executing entity can be diverse and can be set, used, or changed as needed. Alternatively, a third-party application can assist the executing entity in executing Embodiment 1. For example, a server can execute the artificial intelligence-based electronic nasopharyngoscopy examination management method in Embodiment 1, and a corresponding application can be installed on a terminal (which may be held by a user). Data transmission can occur between the terminal or application and the server, thereby assisting the server in executing the artificial intelligence-based electronic nasopharyngoscopy examination management method in Embodiment 1.

[0034] In summary, Embodiment 1 is applied to a computer, meaning that the executing entity of Embodiment 1 can be a computer. Here, "computer" refers to various devices with data processing or data computing capabilities, including but not limited to computers or servers.

[0035] refer to Figure 1 The AI-based electronic nasopharyngoscopy examination management method provided in Example 1 includes: Acquire images to be examined The executing entity in Embodiment 1 can acquire the image to be examined. The image to be examined can be an image transmitted to the executing entity in Embodiment 1 by another entity, such as an image acquired by an electronic nasopharyngoscope and transmitted to the executing entity in Embodiment 1.

[0036] In Example 1, based on the anatomical structure or characteristics of the pharynx and / or common lesion sites (especially high-risk areas such as nasopharyngeal carcinoma, oropharyngeal carcinoma, hypopharyngeal carcinoma, laryngeal carcinoma, and oral cancer) and / or the feasibility of artificial intelligence image recognition, several standard examination stations (hereinafter referred to as "nasopharyngeal preset stations") are pre-designed for the images to be acquired during the nasopharyngeal examination. These nasopharyngeal preset stations represent specific areas or parts of the human body, covering the main anatomical regions of the head, neck, or pharynx; that is, each nasopharyngeal preset station corresponds to a specific area of ​​the nasopharynx and larynx. Performing nasopharyngeal examinations according to these preset stations ensures the completeness of the examination of high-risk lesion areas, achieving a thorough and complete examination.

[0037] Preferably, the pre-designed nasopharyngeal sites include, but are not limited to, multiple from the following 20 specific locations or regions: left nasal cavity, right nasal cavity, left nasopharynx, right nasopharynx, oropharynx and hypopharynx (distant view), left pharyngeal folds, right pharyngeal folds, root of tongue and vallecula, soft palate (midline position), left tonsils, right tonsils, midline hypopharynx and larynx (exposing the diaphragmatic phase), midline hypopharynx (exposing the postcricoid region), left pyriform fossa, right pyriform fossa, panoramic view of the larynx (inspiratory phase), close-up view of both vocal cords (inspiratory phase), panoramic view of the oral cavity, hard palate, and floor of the mouth. In this way, the pre-designed nasopharyngeal sites can cover all key areas from the nasal cavity, nasopharynx, oropharynx, hypopharynx, larynx, and oral cavity, ensuring a comprehensive examination. For a specific example of the distribution of pre-designed nasopharyngeal sites, please refer to [reference needed]. Figure 2 or Figure 3 .

[0038] The executing entity in Embodiment 1 can guide the acquisition of nasopharyngeal images based on the aforementioned preset nasopharyngeal sites to improve the quality and efficiency of nasopharyngeal image acquisition. Specifically, guiding the acquisition of nasopharyngeal images based on the aforementioned preset nasopharyngeal sites may include: before acquiring the image to be examined, the executing entity in Embodiment 1 can continuously acquire images acquired by a nasopharyngeal image acquisition device (e.g., an endoscope or an electronic nasopharyngoscope; hereinafter, an electronic nasopharyngoscope is used as an example), and determine whether the image acquisition conditions of a certain preset nasopharyngeal site are met based on the acquired images; if so, an acquisition prompt message is generated, which prompts the acquisition of images at the target site; wherein, the target site is a preset nasopharyngeal site that meets the image acquisition conditions.

[0039] Preferably, for any pre-defined nasopharyngeal site, the image acquisition conditions for that pre-defined nasopharyngeal site include: The nasopharyngeal imaging acquisition device has reached the image acquisition area of ​​the preset nasopharyngeal station; And / or, The imaging range of the nasopharyngeal imaging acquisition device covers the image acquisition area of ​​the preset nasopharyngeal site.

[0040] In addition, for any nasopharyngeal preset site, the image acquisition conditions for that nasopharyngeal preset site may also include: the target image corresponding to that nasopharyngeal preset site has not yet been acquired.

[0041] The following is a detailed explanation of how to guide the acquisition of nasopharyngeal images based on the aforementioned preset nasopharyngeal sites: For each pre-defined nasopharyngeal site, the execution entity in Implementation Example 1 can formulate image acquisition rules corresponding to that pre-defined nasopharyngeal site. These image acquisition rules enable efficient image acquisition for that pre-defined nasopharyngeal site. The image acquisition rules include, but are not limited to, the insertion depth of the endoscope, the endoscope's orientation relative to the target anatomical structure (e.g., pitch angle, rotation angle), focal length, distance, required field of view, and / or lighting conditions.

[0042] Examples of image acquisition rules for some pre-set nasopharyngeal sites are as follows: • Nasal Cavity: (1) Left nasal cavity, (2) Right nasal cavity. Acquisition procedure: After the endoscope enters the nasal cavity, clear panoramic images of the inferior turbinate, middle turbinate, nasal septum, olfactory cleft and nasal floor are acquired, centered on the inferior and middle turbinates respectively.

[0043] • Nasopharynx: (3) Left nasopharynx, (4) Right nasopharynx. Acquisition guidelines: The endoscope tip should pass past the posterior nasal aperture. Instruct the patient to close their mouth and inhale through their nose, or to swallow. This will fully expose the nasopharyngeal and laryngeal structures and allow observation of the movement of the carina. The image should clearly show the pharyngeal orifice of the Eustachian tube, the torus, the pharyngeal recess, and two-thirds of the posterior wall of the nasopharynx.

[0044] • Oropharynx: (5) Oropharynx and hypopharynx (distant view), acquisition standard: behind the uvula, when the patient pronounces the sound "ee", place the endoscope in the middle of the oropharynx to obtain a wide-angle image including the lower poles of both tonsils, the root of the tongue, the entire hypopharynx and larynx. (6) Left pharyngeal fold, (7) Right pharyngeal fold, acquisition standard: with the endoscope close to the corresponding side, center on the pharyngeal fold, display the pharyngeal fold and root of the tongue on the corresponding side. (8) Root of tongue and vallecula, acquisition standard: instruct the patient to extend their tongue, place the endoscope above the root of the tongue, and fully expose the middle of the root of the tongue and the vallecula on both sides. (9) Soft palate (midline position), acquisition standard: insert the endoscope through the mouth, instruct the patient to pronounce the sound "ee", place the endoscope at the junction of the soft palate and hard palate, observe the upward movement of the soft palate, and acquire images of the palatoglossal arch and palatopharyngeal arch on both sides. (10) Left tonsil, (11) Right tonsil. Collection standard: The microscope body should be close to the corresponding lateral tonsil to clearly show the surface and crypts of the tonsil.

[0045] • Hypopharynx: (12) Hypopharynx and larynx (pronounce "ee"), Acquisition guidelines: Place the endoscope in the middle position above the epiglottis, instruct the patient to pronounce "ee", observe whether the bilateral pyriform fossae are symmetrical, and expose the pyriform fossa at the same time. (13) Midline position of the hypopharynx (exposing the posterior cricoid area), Acquisition guidelines: Use the anterior neck skin traction method or the balloon blowing method to fully open the posterior wall of the hypopharynx and the posterior cricoid area, and fully expose the mucosal area behind the cricoid cartilage. (14) Left pyriform fossa, (15) Right pyriform fossa, Acquisition guidelines: Turn the endoscope to the corresponding side of the pyriform fossa, maintain an appropriate distance when acquiring the image, and the image should include the adjacent arytenoid region and the aryepiglottic fold.

[0046] •Larynx: (16) Panoramic view of the larynx (inspiratory phase), acquisition procedure: The endoscope is placed in the middle position at the level of the epiglottis. The patient is instructed to take a deep breath and open the glottis to its maximum position to obtain a panoramic view of the larynx, including the epiglottic laryngeal surface, ventricular folds, vocal cords, and subglottic area. (17) Close-up view of both vocal cords (inspiratory phase), acquisition procedure: The endoscope tip is about 0.5-1.0 cm away from the glottis to clearly show the surface vascular texture and edges of both vocal cords.

[0047] • Oral Cavity: (18) Panoramic view of the oral cavity, (19) Hard palate, (20) Floor of the mouth. Acquisition specifications: Acquire overview images of the oral cavity, hard palate, and sublingual region in sequence.

[0048] Image acquisition rules may also include: • Endoscope insertion depth: For example, during a nasal examination, the recommended distance between the endoscope tip and the nostril opening is as follows.

[0049] • The orientation of the endoscope relative to the target anatomical structure: for example, pitch angle, rotation angle, or a specific directional description (such as "the endoscope tip is tilted upward at 30 degrees").

[0050] • Focal length and distance: For example, "1-2 cm from the target area".

[0051] • Required field of view: indicated by describing the specific anatomical landmarks that the image should include or by using image guide boxes.

[0052] • Lighting requirements: For example, for sites with high detail requirements, it is recommended to adjust the brightness of the endoscope's light source.

[0053] In addition, the executing entity in Embodiment 1 can also determine the order of each nasopharyngeal preset station, and can prioritize the acquisition of images corresponding to the nasopharyngeal preset stations that are ranked higher.

[0054] In practical applications, the executing entity of Embodiment 1 can guide the image acquisition work corresponding to each nasopharyngeal preset station through the interactive interface displayed on its screen. For example, before acquiring the image to be examined, the executing entity of Embodiment 1 can generate guidance information corresponding to each nasopharyngeal preset station. The guidance information is used to instruct on the acquisition of images of the corresponding nasopharyngeal preset stations and the image acquisition rules.

[0055] Specifically, this may include guidance information for the first pre-selected nasopharyngeal site (referred to as the first site), indicating the acquisition of the image corresponding to the first site and the image acquisition rules for the first site. Once the image corresponding to the first site is acquired, if the target image corresponding to the first site (described below) is obtained, guidance information for the second pre-selected nasopharyngeal site (referred to as the second site) is displayed, indicating the acquisition of the image corresponding to the second site and the image acquisition rules for the second site. This process continues until the image corresponding to a certain pre-selected nasopharyngeal site is acquired, and if the target image corresponding to that nasopharyngeal target site is obtained, guidance information for the next pre-selected nasopharyngeal site (referred to as the nth site, where n is greater than or equal to 2) is displayed, indicating the acquisition of the image corresponding to the nth site and the image acquisition rules for the nth site. This continues until all target images corresponding to all pre-selected nasopharyngeal sites have been acquired.

[0056] During an electronic nasopharyngoscopy, the operator (such as a doctor) can follow the information provided by the execution subject in Example 1 (such as the guidance information mentioned above) to sequentially insert the electronic nasopharyngoscopy into the corresponding anatomical position at each preset nasopharyngeal station, observe each area, and acquire images. For example: • Nasal cavity and nasopharynx: First, insert the laryngoscope through one nostril to observe the walls of the left nasal cavity and the nasal passage. After reaching the nasopharynx, acquire images of the left nasopharynx. Then, withdraw the laryngoscope and insert it through the other nostril to observe the right nasal cavity and acquire images of the right nasopharynx.

[0057] • Oropharynx: Insert the laryngoscope through the mouth to observe the entire oropharynx (including the soft palate, uvula, dorsum of the tongue, and bilateral tonsils), and acquire a panoramic image of the oropharynx and hypopharynx from a distant position. Then, move the lens closer to the left and right pharyngeal folds and tonsils, acquiring images of the left and right tonsils, as well as the left and right pharyngeal folds. Continue to the level of the tongue root, acquiring images of the tongue root and the vallecula region. Instruct the patient to pronounce the "ee" sound, observe the elevation of the soft palate, and acquire an image of the soft palate in a midline position.

[0058] • Hypopharynx and larynx: The laryngoscope tip is bypassed around the epiglottis and entered the hypopharyngeal cavity. First, at the laryngeal inlet, the patient is instructed to breathe calmly, and a panoramic (inspiratory) image of the larynx is acquired. At this time, the glottis is open, allowing a full view of both vocal cords. Then, the patient is instructed to produce a high-pitched "ee" sound to cause vocal cord adduction, and images of the hypopharynx and larynx are acquired while producing the "ee" sound, focusing on vocal cord closure. The lens is slightly adjusted backward to expose the postcricoid mucosa, and an image of the midline of the hypopharynx (postcricoid region) is acquired. Next, the lens is tilted towards the left and right pyriform fossa, respectively, and images of the left and right pyriform fossa are acquired. Finally, the lens is brought close to the glottis, and close-up (inspiratory) images of both vocal cords are acquired to clearly show vocal cord surface details.

[0059] • Other parts of the oral cavity: Withdraw the laryngoscope into the oral cavity and observe and acquire images of the panoramic view of the oral cavity, the hard palate, and the floor of the mouth in sequence, ensuring that these parts are also recorded.

[0060] Of course, the order of the pre-set nasopharyngeal stations is not strictly limited to a fixed absolute order, but refers to a standardized path. Operators can flexibly adjust the order within a certain range according to the patient's specific situation, but in general, it follows the principle of from top to bottom, from outside to inside, and from nose to throat, to ensure that no pre-set station is missed.

[0061] It is worth noting that during the movement of the electronic nasopharyngoscope, the scope continuously captures images (i.e., image streams or video streams) and transmits them to the execution entity of Embodiment 1. That is, the execution entity of Embodiment 1 continuously acquires the images (i.e., image streams) captured by the electronic nasopharyngoscope and determines whether the image acquisition conditions for a preset nasopharyngeal station are met based on the currently acquired images. For example, during the movement of the electronic nasopharyngoscope, the execution entity of Embodiment 1 continuously identifies the acquired images to determine the current position of the electronic nasopharyngoscope and / or the current image acquisition range of the electronic nasopharyngoscope. When it is determined that the electronic nasopharyngoscope has reached the image acquisition area of ​​a preset nasopharyngeal station (hereinafter referred to as station A), and / or that the current image acquisition range of the electronic nasopharyngoscope has covered the image acquisition area of ​​station A, and the target image corresponding to station A has not yet been acquired, then it is determined that the image acquisition conditions for station A are met (station A can be used as the target station). The execution entity of Embodiment 1 then generates acquisition prompt information, prompting the acquisition of images at station A. Furthermore, the acquisition prompt information may also include the image acquisition rules for site A. Specifically, the execution entity in Embodiment 1 can perform frame-by-frame processing on the acquired image stream and determine whether the image acquisition conditions are met by identifying the image frames.

[0062] The aforementioned guidance information and acquisition prompt information can be inclusive. For example, when the image acquisition conditions of station A (assuming station A is the m-th (m≥1) station, i.e., the m-th station) are met, the generated acquisition prompt information can include the guidance information corresponding to the m-th station, which is used to instruct the acquisition of the image corresponding to the m-th station and to instruct the image acquisition rules corresponding to the m-th station.

[0063] After each acquisition prompt message is generated, the operator can acquire images using an appropriate method, such as through the imaging system of an electronic nasopharyngoscope (including but not limited to taking photos or videos, specifically by stepping on or using the camera button on the handle). The acquired images are transmitted to the execution entity of Embodiment 1. If the acquired images are in image format, the execution entity of Embodiment 1 can use one or more newly acquired images after each acquisition prompt message is generated as the images to be examined; if the acquired images are in video format, the execution entity of Embodiment 1 can use the newly acquired video after each acquisition prompt message is generated as the images to be examined.

[0064] Inspection Management After acquiring the image to be inspected, the artificial intelligence model deployed on the execution entity of Example 1 can perform the following operations: S101: After each acquisition of an image to be inspected, perform an inspection operation on the newly acquired image.

[0065] The inspection operation can include two methods. The first method includes: determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, then the image to be inspected is used as the target image corresponding to that nasopharyngeal preset site, and the target image can be stored according to preset rules. The second method includes: determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, then the image to be inspected is used as the target image corresponding to that nasopharyngeal preset site.

[0066] The first method of inspection will be explained in more detail below: Once the image to be inspected (e.g., an image or a video clip) is acquired, it is input into an artificial intelligence model to determine whether the image quality of the latest acquired image meets the requirements. For example, it determines whether there are issues such as sharpness, exposure, artifacts (e.g., blur, reflection, bleeding, mucus coverage), or misalignment with the target area. Furthermore, it can determine whether the image meets the quality requirements based on preset thresholds. Specific quality requirements and evaluation methods are not limited in Example 1.

[0067] If the image quality of the latest acquired image to be examined meets the requirements, then it is determined whether the latest acquired image to be examined corresponds to a certain nasopharyngeal preset site.

[0068] Preferably, for each pre-defined nasopharyngeal site, the execution entity in Embodiment 1 can pre-define a reference image for that pre-defined nasopharyngeal site. The reference image is used to determine whether the image to be examined corresponds to the corresponding pre-defined nasopharyngeal site. In this way, the newly acquired image to be examined can be matched with the reference image of each pre-defined nasopharyngeal site, and a confidence level of matching with each pre-defined nasopharyngeal site can be obtained. If it is determined that the newly acquired image to be examined successfully matches the reference image of a certain pre-defined nasopharyngeal site or that the confidence level meets the conditions, then it is determined that the newly acquired image to be examined corresponds to that pre-defined nasopharyngeal site.

[0069] For example, once the image acquisition conditions for site A are met and the image to be examined is obtained, it is determined whether the most recently acquired image corresponds to a pre-defined nasopharyngeal site. Theoretically, under proper operator conditions, the most recently acquired image should be for site A; that is, the most recently acquired image should correspond to site A. However, due to various factors (such as the patient's condition or the operator's ad-hoc judgment), the most recently acquired image may not be for site A. Therefore, after each image acquisition, it is determined whether the most recently acquired image corresponds to a pre-defined nasopharyngeal site.

[0070] If it is determined that the latest acquired image to be examined corresponds to a certain nasopharyngeal preset site (hereinafter referred to as site B, site B and site A may be the same site or different sites), then the latest acquired image to be examined is used as the target image corresponding to site B, and the target image corresponding to site B can be stored according to preset rules.

[0071] If the image quality of the newly acquired image to be examined does not meet the requirements or there is no corresponding preset nasopharyngeal site, it will not be used as the target image. Furthermore, the execution entity in Embodiment 1 can provide corresponding prompts to promptly reacquire the image.

[0072] If the image to be inspected is in video format, the artificial intelligence model can process it into frames and then perform the above-mentioned inspection operations on each of the segmented image frames.

[0073] S103: After each examination of the image to be examined, determine whether the corresponding target image has been acquired at each preset nasopharyngeal station.

[0074] S105: If there are nasopharyngeal preset sites for which no corresponding target image has been acquired, guidance information corresponding to the remaining sites is generated. This guidance information is used to instruct the acquisition of images corresponding to the remaining sites and to instruct the image acquisition rules corresponding to the next nasopharyngeal preset site. The remaining sites are nasopharyngeal preset sites for which no corresponding target image has been acquired. The next nasopharyngeal preset site can be the first-ranked nasopharyngeal preset site among the remaining sites. Therefore, the guidance information can include the aforementioned guidance information corresponding to the next nasopharyngeal preset site.

[0075] Similarly, after acquiring the image to be examined again, continue the above examination operation until all the preset nasopharyngeal stations have corresponding target images.

[0076] The second method of examination involves first performing site matching on the image to be examined, that is, determining whether the image corresponds to a certain preset nasopharyngeal site. If it corresponds to a preset nasopharyngeal site, then image quality assessment is performed. The difference between this and the first method lies in the order of site matching and image quality assessment; the specific content of site matching and image quality assessment is the same as in the first method. Both methods can be selected or set according to the actual situation. The following explanation uses only one method; the other method is similar.

[0077] Preferably, the target images are stored according to preset rules, such as associating each nasopharyngeal preset site with its corresponding target image (e.g., marking the corresponding nasopharyngeal preset site in the target image file name or metadata, or naming each target image with its corresponding nasopharyngeal preset site or associating it with a site tag in the database). This way, each target image has a clear anatomical location, facilitating subsequent retrieval, viewing, and analysis. Furthermore, the artificial intelligence model can optimize the target images before storage, storing the optimized target images.

[0078] Preferably, the execution entity in Embodiment 1 can display various information in various suitable ways, including but not limited to sound prompts, on-screen text prompts, image overlay guide lines, target area highlighting, image comparison, or progress percentage display, aiming to provide the operator with intuitive and efficient assistance. A typical interface layout can be divided into two areas: • Main display area: Occupies most of the screen, displaying high-definition images from the electronic nasopharyngoscope in real time. Auxiliary information, such as highlight boxes for target areas and focus guide lines, can be overlaid in this area.

[0079] • Quality Control Information Area: Located on the side of the screen, it provides a comprehensive overview of the quality control status.

[0080] • Site List and Status: All preset standardized autopsy sites are listed in a list format. Each site has a status indicator light, such as green for "accepted image" (meaning there is a target image), yellow for "pending inspection", and red for "acceptance failed and needs to be retaken".

[0081] • Inspection progress bar: Displays the overall inspection completion percentage graphically.

[0082] • Real-time feedback window: Displays immediate instructions in text or icon form, such as "Image is blurry, please adjust", "[Right pyriform fossa] acquisition completed", "Warning: [Left nasopharynx] has not been examined!".

[0083] In addition, for example Figure 4 Within the system, various interfaces allow for operation control, quality scoring of images to be examined (taking images as an example), maintenance of preset nasopharyngeal sites and their corresponding target image lists, and display and interaction of performance monitoring. It also allows for... Figure 5 As shown, the acquired images and the target images acquired at each preset nasopharyngeal site are displayed.

[0084] As a preferred embodiment, the executing entity can maintain an inspection status list (or "inspection map"). Whenever a target image corresponding to a new nasopharyngeal preset site is identified, the nasopharyngeal preset site is marked, and the coverage rate of the site set (i.e., the ratio of the number of sites with corresponding target images and qualified image quality to the total number of sites) is dynamically calculated based on the existing sites with corresponding target images and qualified image quality.

[0085] Preferably, after obtaining the corresponding target images at each pre-set nasopharyngeal site, the executing entity in Example 1 can automatically generate a standardized report (i.e., a quality control report). The report includes, but is not limited to, all pre-set nasopharyngeal sites covered in this inspection and the target image acquisition status corresponding to each pre-set nasopharyngeal site (e.g., whether the acquisition was successful, target image score, etc.), and includes thumbnails of each acquired target image.

[0086] Preferably, the artificial intelligence model on the execution subject in Example 1 can be used to analyze each target image and indicate whether there are any abnormal lesions. Additionally, the operator can fill in the report with visual observations and preliminary diagnostic opinions for each area. Since target images are archived for all important areas (i.e., the nasopharyngeal preset sites), the report content is more complete and objective.

[0087] The executing entity in Implementation Example 1 can provide feedback on relevant information at any time, such as checking the progress and / or results of the operation, and the determination of the target images corresponding to each preset nasopharyngeal site (these information can be collectively referred to as quality control feedback information). Implementation Example 1 is not limited to this.

[0088] The above artificial intelligence model is further explained as follows: The artificial intelligence model constructed in Example 1 employs a deep learning model, such as a convolutional neural network (CNN) architecture (preferably containing an Inception-ResNet module and / or a Squeeze-and-Excitation (SE) module) based on large-scale laryngoscope image data. Figure 6 The model (as shown) is trained to create a classification model used to classify and identify images of different anatomical sites, thereby determining whether the image to be examined corresponds to a specific nasopharyngeal pre-defined site. Training data can come from laryngoscopy images from various hospitals, each image annotated by experts with its corresponding anatomical site (such as one of the 20 pre-defined nasopharyngeal sites mentioned above). Through supervised learning, the model gradually becomes able to identify the currently displayed site from the texture, color, and structural features of the image. In addition, an image quality assessment model can be trained to detect whether the image quality of the image to be examined meets the requirements.

[0089] The following example, which combines Inception-ResNet-V2 and Squeeze-and-Excitation Network (SENet), will be used to further illustrate this point: • Model Selection: The Inception-ResNet-V2 architecture excels at extracting image features at different scales, while ResNet's residual connections facilitate training deeper networks. The SENet module, as a channel attention mechanism, allows the model to adaptively learn the importance of different feature channels, thereby focusing on the most critical image features for site identification and quality assessment, improving accuracy.

[0090] • Dataset Construction: Building the model requires a large-scale, high-quality dataset with precise expert annotations. This dataset should contain endoscopic images covering all pre-defined anatomical sites (such as the 20 sites mentioned above). Each image should be cross-annotated by at least two senior physicians, with labels including "anatomical site" and "image quality" (e.g., clear, blurry, reflective, occluded, etc.). For example, approximately 45,000 high-quality images can be selected from over 200,000 clinical images to ultimately construct a training, validation, and test set containing approximately 16,000 images.

[0091] • Model Training: Input preprocessed images (e.g., size normalized to 299x299 pixels) into the model. Use the cross-entropy loss function and perform multiple iterations of training via backpropagation and an optimizer (e.g., Adam or momentum-based SGD). After training, evaluate the model's performance on independent test sets to ensure that its accuracy, sensitivity, and specificity for each site meet clinical application requirements (e.g., overall accuracy ≥ 97%).

[0092] The classification model and the image quality assessment model can be the same model, i.e., an artificial intelligence model; or the classification model and the image quality assessment model can be different models, but are collectively referred to as artificial intelligence models. Since the artificial intelligence model is deployed on the execution subject of Implementation Example 1, the actions performed by the artificial intelligence model can be equivalent to the actions performed by the execution subject of Implementation Example 1. All the actions that need to be performed by the execution subject of Implementation Example 1 can be performed by the artificial intelligence model.

[0093] The following example further illustrates Embodiment 1: refer to Figure 7 Assuming there are 20 pre-set nasopharyngeal stations, before the electronic nasopharyngoscopy begins, the execution entity in Example 1 can display relevant information, such as the order of the pre-set nasopharyngeal stations, and can display guidance information for the first station.

[0094] The operator operates the electronic nasopharyngoscope according to the instructions. Starting from the nasal cavity, the nasopharyngoscope gradually penetrates deeper into the patient's body, continuously capturing images. The execution unit in Example 1 continuously acquires images captured by the electronic nasopharyngoscope. When the image acquisition conditions for a pre-defined nasopharyngeal site are met, an acquisition prompt is generated, indicating that image acquisition should be performed at that site. The operator acquires images (i.e., images to be examined, hereinafter assumed to be images) through the imaging system of the electronic nasopharyngoscope and transmits them to the execution unit in Example 1. The execution unit in Example 1 can display the received images to be examined (the same applies below). The artificial intelligence model on the execution unit in Example 1 examines the acquired images to be examined, determining the image quality of the latest acquired image (i.e., artificial intelligence image quality assessment) and whether it corresponds to a pre-defined nasopharyngeal site (i.e., lesion area identification).

[0095] Scenario 1 If the image quality of the latest acquired image to be examined meets the requirements and corresponds to the first site, then it is used as the target image corresponding to the first site. The execution entity of Example 1 displays guidance information, indicating that image acquisition is still required for the remaining 19 nasopharyngeal preset sites (the remaining sites can be called blind spots), as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is the second site at this time).

[0096] Scenario 2 If the image quality of the latest acquired image to be examined meets the requirements, and it corresponds to a nasopharyngeal preset site other than the first site, then it is taken as the target image corresponding to that nasopharyngeal preset site. The execution subject of Embodiment 1 displays guidance information, indicating that image acquisition needs to be performed on the remaining 19 nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is still the first site at this time).

[0097] Scenario 3 If it is determined that the image quality of the latest acquired image to be examined does not meet the requirements or does not correspond to a certain nasopharyngeal preset site, the execution subject of Implementation Example 1 will display guidance information, indicating that image acquisition is still required for the remaining 20 nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is still the first site at this time).

[0098] Similarly, as the electronic nasopharyngoscope moves forward, when the image acquisition conditions for a certain preset nasopharyngeal station are met, an acquisition prompt message is generated, prompting the operator to acquire an image for that preset nasopharyngeal station. The operator acquires images (i.e., images to be examined) through the imaging system of the electronic nasopharyngoscope and transmits them to the execution entity in Example 1. The artificial intelligence model on the execution entity in Example 1 examines the acquired images to be examined, determining the image quality of the latest acquired images and whether they correspond to a certain preset nasopharyngeal station.

[0099] If the image quality of the latest acquired image to be examined meets the requirements and corresponds to a certain nasopharyngeal preset site, then it is used as the target image corresponding to that nasopharyngeal preset site. The executing entity in Example 1 displays guidance information, indicating that image acquisition is still required for the remaining nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is currently the first nasopharyngeal preset site in the remaining list). If it cannot be used as the target image, the executing entity in Example 1 displays guidance information, indicating that image acquisition is still required for the remaining nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is currently the first nasopharyngeal preset site in the remaining list).

[0100] This process continues until all nasopharyngeal preset sites have acquired the corresponding target images (i.e., covering all nasopharyngeal preset sites, the same below).

[0101] The following example further illustrates Embodiment 1: Assume there are 20 pre-set nasopharyngeal stations. Before the electronic nasopharyngoscopy begins, the execution entity in Example 1 can display relevant information, such as the order of the pre-set nasopharyngeal stations, and can display guidance information for the first station.

[0102] The operator operates the electronic nasopharyngoscope according to the instructions. The nasopharyngoscope moves within the patient's body, continuously capturing images. The execution unit in Example 1 continuously acquires images captured by the electronic nasopharyngoscope. When it determines that the image acquisition conditions for the first station are met, it generates an acquisition prompt, indicating that image acquisition for the first station should proceed. The operator acquires images (i.e., images to be examined) through the imaging system of the electronic nasopharyngoscope and transmits them to the execution unit in Example 1. The artificial intelligence model on the execution unit in Example 1 examines the acquired images to be examined, determining the image quality of the latest acquired image and whether it corresponds to a specific preset nasopharyngeal station.

[0103] Scenario 1 If the image quality of the latest acquired image to be examined meets the requirements and corresponds to the first site, then it is used as the target image corresponding to the first site. The execution entity of Example 1 displays guidance information, indicating that image acquisition is still required for the remaining 19 nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is the second site at this time).

[0104] Scenario 2 If the image quality of the newly acquired image to be examined is determined to be unsatisfactory or does not correspond to the first site, then the newly acquired image to be examined will not be used as the target image. The executing entity in Example 1 displays guidance information, indicating that image acquisition needs to be performed on the remaining 20 nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is still the first site). The operator continues to acquire images until the target image corresponding to the first site is determined. Then, the executing entity in Example 1 displays guidance information, indicating that image acquisition needs to be performed on the remaining 19 nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is now the second site).

[0105] Similarly, as the electronic nasopharyngoscope moves forward, when the image acquisition conditions for the second station are met, an acquisition prompt message is generated, prompting the operator to acquire images for the second station. The operator acquires images (i.e., images to be examined) through the imaging system of the electronic nasopharyngoscope and transmits them to the execution entity in Example 1. The artificial intelligence model on the execution entity in Example 1 examines the acquired images to be examined, determining the image quality of the latest acquired images and whether they correspond to a certain preset nasopharyngeal station.

[0106] Scenario 1 If the image quality of the latest acquired image to be examined meets the requirements and corresponds to the second site, then it is used as the target image corresponding to the second site. The execution entity of Example 1 displays guidance information, indicating that image acquisition is still required for the remaining 18 nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is the third site at this time).

[0107] Scenario 2 If the image quality of the newly acquired image to be examined is determined to be unsatisfactory or does not correspond to the second site, then the newly acquired image to be examined will not be used as the target image. The executing entity in Example 1 displays guidance information, indicating that image acquisition needs to be performed on the remaining 19 nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is still the second site). The operator continues to acquire images until the target image corresponding to the second site is determined. Then, the executing entity in Example 1 displays guidance information, indicating that image acquisition needs to be performed on the remaining 18 nasopharyngeal preset sites, as well as the image acquisition rules corresponding to the next nasopharyngeal preset site (which is now the third site).

[0108] This process continues until all nasopharyngeal pre-defined sites have acquired their corresponding target images. It is evident that this example requires that the target images for each nasopharyngeal pre-defined site be determined sequentially according to their order.

[0109] Example 1 can achieve the following beneficial effects: In Example 1, the nasopharyngeal examination process and image acquisition process are monitored and guided in real time through preset standard examination sites and artificial intelligence models. This ensures that the electronic nasopharyngoscopy is performed according to the preset sites, guaranteeing that the anatomical location corresponding to the target image accurately corresponds to the preset site. Each preset site can accurately obtain the corresponding target image that meets the quality requirements. Each electronic nasopharyngoscopy examination covers all key anatomical locations of the nasopharynx and larynx, ensuring that the operation by different personnel at different times follows a unified standard. This avoids omissions of examination sites due to factors such as personal habits, experience, or operating methods. It achieves precise and efficient quality control (referred to as quality control) and management of the nasopharyngeal examination process and image acquisition process, making the electronic nasopharyngoscopy examination process comprehensive, complete, standardized, regulated, intelligent, and traceable. This effectively improves the quality and efficiency of the examination and overcomes the technical difficulties of existing technologies, such as the lack of objective standards in the examination process, non-standard image acquisition, inconsistent image quality, easy omission of blind spots, inability to guarantee the integrity of the examination, and lack of AI quality control capabilities, which lead to a high rate of missed diagnoses. The management of nasopharyngoscopy in Example 1 provides complete imaging evidence for subsequent diagnosis and treatment, which is conducive to improving the detection rate of early lesions, reducing missed diagnoses and misdiagnoses, and helping to improve the diagnostic level and medical quality of nasopharyngeal diseases. It has significant clinical value and prospects for promotion.

[0110] In Example 1, real-time monitoring and tracking of pre-defined nasopharyngeal sites with acquired target images, along with dynamic alerts for remaining sites, effectively avoids overlooking hidden areas such as the tonsils, hypopharyngeal pyriform fossa, and postcricoid region, thereby significantly improving the detection rate of precancerous lesions and early-stage head and neck cancer. Example 1 can promptly identify oversights or imaging problems during the examination, prompting doctors to make corrections, thus ensuring the examination is completed according to standards, significantly improving the consistency and reliability of the examination, and providing patients with more reliable medical services.

[0111] Example 1 transforms the qualitative examination process into quantitative site coverage data and image quality assessment results, establishing a clear quality evaluation basis for electronic nasopharyngoscopy. By recording the target image corresponding to each pre-set nasopharyngeal site, it effectively achieves a quantitative assessment of the examination quality, such as whether all pre-set nasopharyngeal sites have corresponding target images and whether the images to be examined meet the requirements of the pre-set nasopharyngeal sites. This provides objective indicators for quality control in hospital endoscopy centers and also provides objective, traceable evidence and reference standards for medical quality management, physician training, and performance evaluation.

[0112] In Example 1, several standard inspection stations were pre-defined, enabling the computer to understand and track the inspection process. Combined with image recognition technology from an artificial intelligence model, the system automatically identifies which areas were inspected, determines whether the image to be inspected can be used as the target image, and provides real-time feedback to the operator. This elevates electronic nasopharyngoscopy to a new stage of intelligent quality control, further ensuring the quality of the examination.

[0113] Furthermore, experiments have shown that the artificial intelligence model trained in Example 1 can accurately and quickly classify and identify anatomical sites in the images to be examined and make quality judgments. It demonstrates excellent capabilities in supervising examination quality and can accurately and quickly identify whether an image to be examined is suitable as a target image. (Reference) Figure 8 Taking the image to be examined as an example, the artificial intelligence model used in Example 1 achieved an accuracy rate of over 97% in recognizing each preset nasopharyngeal site (i.e., anatomical site), fully demonstrating the application effect of Example 1.

[0114] The artificial intelligence model provided in Example 1 can perform examination operations and provide feedback at millisecond speeds, providing real-time assistance to examining physicians. It plays a "smart navigation" role, especially for less experienced primary care physicians, which can lower the technical threshold for operators, help shorten the learning curve, and improve their operational level and examination efficiency.

[0115] In Example 1, since each pre-set nasopharyngeal station corresponds to a standardized target image, doctors can refer to these complete and standardized target images during diagnosis, which helps improve diagnostic accuracy and prevents early lesions from being missed due to oversights during electronic nasopharyngoscopy. Especially for diseases such as head and neck tumors, standardized image acquisition helps in the early detection of small lesions, thereby improving patient prognosis.

[0116] In Example 1, each target image and all related data can be stored according to preset rules. Each target image has a unified site marker, facilitating academic exchanges and remote consultations between different hospitals and doctors. It also provides a reliable basis for subsequent case analysis, teaching and training, evidence collection in medical disputes, and scientific research. Consulting doctors can quickly locate the anatomical location or preset nasopharyngeal site corresponding to the target image based on the site marker, improving communication efficiency and accuracy.

[0117] Example 1 is adaptable to different artificial intelligence model implementations and various electronic nasopharyngoscope devices, demonstrating good applicability and stability. Furthermore, the standardized, high-quality target images and recorded quality control data acquired serve as ideal data sources for training and optimizing artificial intelligence diagnostic models, further promoting the application of artificial intelligence in the medical field.

[0118] The second embodiment of this specification (hereinafter referred to as "Embodiment Two") provides an artificial intelligence-based electronic nasopharyngoscopy examination and management system corresponding to the method described in Embodiment One, including: The image acquisition unit is used to acquire images to be examined. The image acquisition unit can be the electronic nasopharyngoscope itself, containing an image sensor (such as CMOS / CCD) and a light source. It is responsible for capturing real-time high-definition video streams of the nasopharynx and larynx, and transmitting them to a computer's artificial intelligence processing unit via wired (such as HDMI / USB) or wireless (such as Wi-Fi) methods.

[0119] The computer includes an artificial intelligence processing unit for running an artificial intelligence model and performing an inspection operation on the image to be inspected; wherein the inspection operation includes: determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site (hereinafter referred to as a site); if the image to be inspected corresponds to a certain site, then the image to be inspected is used as the target image corresponding to that site. After each inspection of the image to be inspected, determine whether the corresponding target image has been acquired at each site; If there are stations that have not yet acquired the corresponding target image, guidance information corresponding to the remaining stations is generated. The guidance information is used to instruct the acquisition of the image corresponding to the remaining stations and to instruct the image acquisition rules corresponding to the next station. The remaining stations are those that have not yet acquired the corresponding target image.

[0120] The artificial intelligence processing unit, acting as the "brain" of the system, may include a high-performance processor (such as a GPU or a dedicated AI chip). The artificial intelligence processing unit may check the status list as described in Embodiment 1.

[0121] Preferably, the computer is also used to continuously acquire images acquired by the nasopharyngeal imaging acquisition device before acquiring the image to be examined, and to determine whether the image acquisition conditions of a certain site are met based on the acquired images. If so, a data acquisition prompt message is generated, which prompts the user to acquire images of the target site; wherein, the target site is a site that meets the image acquisition conditions.

[0122] Preferably, for any given site, the image acquisition conditions for that site include: The nasopharyngeal imaging equipment has reached the image acquisition area of ​​the site; And / or, The imaging range of the nasopharyngeal imaging equipment covers the image acquisition area of ​​the site.

[0123] As a preferred option, the image acquisition conditions for this site also include: The target image corresponding to this site has not yet been obtained.

[0124] Preferably, the acquisition prompt information also includes the image acquisition rules for the target site.

[0125] Preferably, the computer is also used to use the newly acquired images after each acquisition prompt information is generated as images to be inspected.

[0126] Preferably, the computer is further configured to generate first guidance information corresponding to the first station before acquiring the image to be inspected. The first guidance information is used to instruct the acquisition of the image corresponding to the first station and to instruct the image acquisition rules corresponding to the first station; wherein, the first station is the station ranked first.

[0127] Preferably, the computer is also used to preset a reference image for each site, the reference image being used to determine whether the image to be inspected corresponds to the site.

[0128] Preferably, the site corresponds to a specific area of ​​the nasopharynx; And / or, The next station is the station that is ranked first among the remaining stations.

[0129] Preferably, the artificial intelligence processing unit is also used to update the inspection coverage status of the site set in real time, that is, to update which sites have corresponding target images and which sites do not yet have corresponding target images.

[0130] Preferably, the computer further includes a feedback generation unit connected to the artificial intelligence processing unit. The feedback generation unit is used to generate quality control feedback information about the integrity of the inspection based on the inspection coverage status and / or the image quality of the image to be inspected, or to generate specific quality control feedback instructions according to preset rules.

[0131] Preferably, the computer further includes a user interaction unit for displaying various types of information, such as guidance information, instruction information, and quality control feedback information for each site. The guidance information instructs the acquisition of images corresponding to the specified site and specifies the image acquisition rules. The user interaction unit can also display real-time endoscopic images, examination instructions from the guidance information storage unit, and real-time quality control feedback information from the feedback generation unit to the operator.

[0132] The user interaction unit, serving as the system's "dashboard," may include at least one display, speaker, or indicator light, and is used to provide the quality control feedback information through methods such as sound prompts, on-screen text prompts, image overlay auxiliary lines, target area highlighting, image comparison, or inspection progress percentage display.

[0133] Preferably, the artificial intelligence processing unit employs a deep learning model as its core, which includes a convolutional neural network (CNN) architecture. Further, the CNN architecture includes an Inception-ResNet module and / or a Squeeze-and-Excitation (SE) module.

[0134] Preferably, the computer further includes a guidance information storage unit for storing each station and the guidance information corresponding to each station.

[0135] Preferably, the computer further includes a data storage unit for storing various types of data or information involved in Embodiment 1 or Embodiment 2, such as inspection operation results, correspondence results between the image to be inspected and the site, image quality judgment results, determination of target images for each site, quality control data, original images or keyframes, and reports.

[0136] Preferably, the system is integrated into the handle, controller, or endoscope unit of the electronic nasopharyngoscope device.

[0137] Here is an example: • If the image to be inspected precisely corresponds to a certain site and is of acceptable quality, the integrity assessment module updates the inspection progress, and the user interaction unit displays "Site [XX] has been completed, please inspect the next site" or the progress bar is updated.

[0138] • If the AI ​​determines that the current image to be examined does not match any of the sites (e.g., it deviates from the site that should be examined), the feedback generation unit generates feedback information such as "Please adjust the endoscope position to align with the target site [XX]", and the user interaction unit provides prompts through on-screen text, voice, or image overlay with auxiliary lines.

[0139] • If the AI ​​determines that the quality of the image to be inspected is unacceptable (e.g., blurry), the feedback generation unit generates a feedback message such as "Image is blurry, please adjust focus / distance / angle", and the user interaction unit prompts the operator accordingly.

[0140] • If the integrity assessment reveals any missed sites, the system will remind the operator to take additional photos. Based on real-time feedback from the user interaction unit, the operator should adjust the operation of the electronic nasopharyngoscope until all sites are effectively covered (i.e., each has a corresponding target image) and the quality is satisfactory. The system will then indicate that the inspection is complete.

[0141] An example of the working process of the system is as follows: Figure 9 As shown, the specific system architecture can also be as follows: Figure 10 As shown.

[0142] The system can be configured in the following ways: • Fully integrated: The artificial intelligence processing unit (such as an embedded artificial intelligence chip), data storage unit, and user interaction unit (such as a micro display) are directly integrated into the handle or controller of the electronic nasopharyngoscope, forming a highly portable all-in-one intelligent device.

[0143] • Partially integrated: Core modules such as the artificial intelligence processing unit are integrated into the endoscope host (video processor) and connected to the probe via the endoscope cable.

[0144] Standalone configuration: The system acts as a standalone device (such as an AI workstation or tablet computer) and connects to a standard electronic nasopharyngoscope via wired or wireless means, providing it with intelligent management and control functions.

[0145] Through hardware-level innovation, the system solidifies standardized quality control processes into equipment functions, providing a ready-to-use solution for clinical applications, thus forming a solid technological barrier.

[0146] The following example further illustrates the working process of the system described in Embodiment 2: refer to Figure 11 This example includes the following process: 1. Start inspection (201): The operator starts the system, and the system loads the preset set of standardized anatomical sites.

[0147] 2. Acquiring real-time images (202): The operator manipulates the electronic nasopharyngoscope to perform the examination, and the image acquisition unit captures the video stream in real time to obtain the image to be examined.

[0148] 3. Real-time AI analysis (203): The AI ​​processing unit performs inspection operations, including analyzing each frame of image and performing site identification and image quality assessment in parallel.

[0149] 4. Is the site a standard site (204): The artificial intelligence model determines whether the current image clearly corresponds to one of the preset sites. If yes, proceed to step (205); if no, the system can prompt "No standard site was identified, please adjust the position" through the user interaction unit and return to step (202).

[0150] 5. Is the image quality acceptable? (205): The model evaluates the image quality of the identified sites (e.g., sharpness, no glare). If acceptable, proceed to step (206); if unacceptable, the system prompts "Image quality is poor (e.g., blurry), please adjust focus / angle", and the operator returns to step (202) after adjustment.

[0151] 6. Update inspection coverage status (206): For sites that have been identified and are of acceptable quality, the system marks them as "covered" and updates the inspection progress.

[0152] 7. Are all sites covered? (207): The system determines whether the preset site set has been completely covered. If yes, proceed to step (209); if no, proceed to step (208).

[0153] 8. Generate and provide feedback (208): The system highlights the uncovered sites (blind spots) on the user interface or prompts "Please continue to check the next site: [XX]", and returns to step (202) to continue checking.

[0154] 9. Inspection Complete (209): The system prompts that the inspection is complete and can generate and store an inspection report containing quality control information.

[0155] 204 and 205 can be interchanged, meaning you can first determine the image quality and then determine if it corresponds to a specific site.

[0156] The contents not described in detail in Embodiment 1 and Embodiment 2 can be referred to each other. Embodiment 2 can achieve the same beneficial effects as Embodiment 1. The above embodiments can be used in combination.

[0157] The above description is merely an embodiment of this specification and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A management method for electronic nasopharyngoscopy examination based on artificial intelligence, characterized in that, The method is applied to a computer, and the method includes: After each image to be examined is acquired, the artificial intelligence model on the computer performs the following operations: An inspection operation is performed on the image to be inspected; wherein the inspection operation includes: determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, then using the image to be inspected as the target image corresponding to the nasopharyngeal preset site; or, the inspection operation includes: determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, then using the image to be inspected as the target image corresponding to the nasopharyngeal preset site. After each examination of the images to be examined, determine whether the corresponding target images have been acquired at each preset nasopharyngeal site; If there are nasopharyngeal preset sites for which the corresponding target image has not yet been acquired, guidance information corresponding to the remaining sites is generated. The guidance information is used to instruct the acquisition of the image corresponding to the remaining sites and to instruct the image acquisition rules corresponding to the next nasopharyngeal preset site. The remaining sites are nasopharyngeal preset sites for which the corresponding target image has not yet been acquired.

2. The method as described in claim 1, characterized in that, Before acquiring the image to be examined, the method further includes: Continuously acquire images from the nasopharyngeal imaging acquisition device, and determine whether the image acquisition conditions of a certain preset nasopharyngeal site are met based on the acquired images. If so, a collection prompt message is generated, which prompts the image to be collected at the target site; wherein, the target site is a preset nasopharyngeal site that meets the image collection conditions.

3. The method as described in claim 2, characterized in that, For any pre-defined nasopharyngeal site, the image acquisition conditions for that site include: The nasopharyngeal imaging acquisition device has reached the image acquisition area of ​​the preset nasopharyngeal station; And / or, The imaging range of the nasopharyngeal imaging acquisition device covers the image acquisition area of ​​the preset nasopharyngeal site.

4. The method as described in claim 3, characterized in that, The image acquisition conditions for this preset nasopharyngeal site also include: The target image corresponding to the preset nasopharyngeal site has not yet been obtained.

5. The method according to any one of claims 2 to 4, characterized in that, The acquisition prompt information also includes the image acquisition rules for the target site.

6. The method according to any one of claims 2 to 4, characterized in that, The method further includes: The newly acquired images after each acquisition prompt message is generated are used as the images to be examined.

7. The method as described in claim 1, characterized in that, Before acquiring the image to be examined, the method further includes: Generate first guidance information corresponding to the first site. The first guidance information is used to instruct the acquisition of images corresponding to the first site and to instruct the image acquisition rules corresponding to the first site. The first site is the nasopharyngeal preset site that is ranked first.

8. The method as described in claim 1, characterized in that, The method further includes: For each nasopharyngeal preset site, a reference image for that nasopharyngeal preset site is preset. The reference image is used to determine whether the image to be examined corresponds to the nasopharyngeal preset site.

9. The method as described in claim 1, characterized in that, The preset nasopharyngeal stations correspond to specific areas of the nasopharynx. And / or, The next nasopharyngeal preset station is the nasopharyngeal preset station that is ranked first among the remaining stations.

10. An artificial intelligence-based electronic nasopharyngoscopy examination and management system, characterized in that, The system includes: The image acquisition unit is used to acquire images to be inspected. An artificial intelligence processing unit is used to run an artificial intelligence model and perform an inspection operation on the image to be inspected; wherein the inspection operation includes: determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, then using the image to be inspected as the target image corresponding to the nasopharyngeal preset site; or, the inspection operation includes: determining whether the image to be inspected corresponds to a certain nasopharyngeal preset site; if the image to be inspected corresponds to a certain nasopharyngeal preset site, determining whether the image quality of the image to be inspected meets the requirements; if it meets the requirements, then using the image to be inspected as the target image corresponding to the nasopharyngeal preset site. After each examination of the images to be examined, determine whether the corresponding target images have been acquired at each preset nasopharyngeal site; If there are nasopharyngeal preset sites for which the corresponding target image has not yet been acquired, guidance information corresponding to the remaining sites is generated. The guidance information is used to instruct the acquisition of the image corresponding to the remaining sites and to instruct the image acquisition rules corresponding to the next nasopharyngeal preset site. The remaining sites are nasopharyngeal preset sites for which the corresponding target image has not yet been acquired.

Images