An ultrasound-guided lateral skull base targeted drug delivery system for effective treatment of tinnitus and vertigo

Abstract

The present application belongs to the technical field of medical ultrasound image guidance and tinnitus and vertigo targeted drug delivery, and discloses a side cranial base targeted drug delivery technology system for effectively treating tinnitus and vertigo under ultrasound guidance, which comprises a side cranial base ultrasound data acquisition module, a multi-source feature fusion construction module, a tinnitus and vertigo phenotype clustering module, a side cranial base target area clustering positioning module, a targeted drug delivery path optimization module and a drug delivery feedback clustering update module. The side cranial base fusion feature matrix, the tinnitus and vertigo phenotype clustering result, the side cranial base target area candidate cluster, the side cranial base risk structure cluster, the side cranial base safe path cluster, the side cranial base target drug delivery target area and the side cranial base targeted drug delivery scheme are generated by using the principles of ultrasound multi-source feature fusion, tinnitus and vertigo phenotype clustering, side cranial base spatial clustering positioning, drug delivery path safety optimization and feedback clustering update, so as to realize tinnitus and vertigo phenotype grouping, side cranial base target area positioning and drug delivery path safety screening.

Description

Technical Field

[0001] This invention belongs to the field of medical ultrasound image guidance and targeted drug delivery technology for tinnitus and vertigo, and particularly relates to a lateral skull base targeted drug delivery technology system for the effective treatment of tinnitus and vertigo under ultrasound guidance. Background Technology

[0002] Tinnitus and vertigo are often related to the auditory system, vestibular function, ear microcirculation, and the condition of adjacent tissues at the lateral skull base. When performing targeted drug delivery, it is necessary to accurately determine the tinnitus and vertigo phenotype of the drug recipient, the target area of ​​the lateral skull base, and the safety of the drug delivery route. The lateral skull base region is surrounded by bony boundaries, blood flow areas, soft tissue interfaces, and acoustic interference areas. If the drug delivery location and route are determined solely by human experience, they are easily affected by individual anatomical differences, differences in ultrasound image quality, and adjacent vascular structures. Therefore, there is a need for an ultrasound-guided lateral skull base targeted drug delivery technology system for the effective treatment of tinnitus and vertigo, which performs multi-source feature fusion, tinnitus and vertigo phenotype clustering, lateral skull base target area clustering and localization, and targeted drug delivery route optimization of lateral skull base ultrasound data.

[0003] The existing technology has at least the following problems that need to be improved: (1) Existing tinnitus and vertigo medication regimens rely heavily on symptom descriptions and human experience judgment. They lack a mechanism for joint cluster analysis of tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data, previous medication records data, and lateral skull base fusion feature matrix, resulting in insufficient adaptability between medication regimens and different tinnitus and vertigo phenotypes.

[0004] (2) Existing methods for lateral skull base drug delivery location lack spatial clustering identification of candidate drug delivery areas, vascular adjacent areas, bony obstruction areas, acoustic interference areas and safe access areas. They also lack a mechanism for optimizing drug delivery pathways based on candidate clusters of lateral skull base target areas, risk clusters of lateral skull base structures, safe access clusters of lateral skull base and probe posture data, resulting in insufficient safety in target delivery location and drug delivery pathway selection. Summary of the Invention

[0005] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo. Utilizing principles of ultrasound multi-source feature fusion, tinnitus and vertigo phenotypic clustering, lateral skull base spatial clustering localization, drug delivery route safety optimization, and feedback clustering updates, it generates a lateral skull base fusion feature matrix, tinnitus and vertigo phenotypic clustering results, lateral skull base target area candidate clusters, lateral skull base risk structure clusters, lateral skull base safe pathway clusters, lateral skull base target drug delivery area, and lateral skull base targeted drug delivery protocol. This achieves tinnitus and vertigo phenotypic grouping, lateral skull base target area localization, and safe drug delivery route screening, solving the problems of insufficient individualization in existing tinnitus and vertigo drug delivery protocols, reliance on experience for target area localization, insufficient avoidance of risk structures, and unstable route safety.

[0006] This invention provides a lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo under ultrasound guidance, comprising a lateral skull base ultrasound data acquisition module, a multi-source feature fusion construction module, a tinnitus and vertigo phenotype clustering module, a lateral skull base target area clustering and localization module, a targeted drug delivery pathway optimization module, and a drug delivery feedback clustering update module.

[0007] The lateral skull base ultrasound data acquisition module acquires basic two-dimensional ultrasound images, color Doppler images, energy Doppler images, spectral Doppler data, probe posture data, and acquisition parameter data of the lateral skull base region of the drug recipient, and generates raw lateral skull base ultrasound data. The multi-source feature fusion construction module receives raw data from lateral skull base ultrasound and performs noise reduction and enhancement, blood flow region separation, acoustic shadow region marking, grayscale normalization, and spatial registration on basic two-dimensional ultrasound images, color Doppler images, energy Doppler images, and spectral Doppler data. It then extracts bony boundary features, blood flow course features, soft tissue interface features, acoustic shadow interference features, and local echo texture features to generate a lateral skull base fusion feature matrix. The tinnitus and vertigo phenotype clustering module receives the side skull base fusion feature matrix, obtains the tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data and previous drug administration record data of the drug administration subjects, and performs tinnitus and vertigo phenotype grouping on the drug administration subjects based on the clustering algorithm to generate tinnitus and vertigo phenotype clustering results. The lateral skull base target area clustering and localization module receives the lateral skull base fusion feature matrix, and performs spatial clustering of candidate drug delivery areas, vascular adjacent areas, bony barrier areas, acoustic interference areas and safe access areas in the lateral skull base region based on the clustering algorithm, generating lateral skull base target area candidate clusters, lateral skull base risk structure clusters and lateral skull base safe access clusters, and generates the lateral skull base target drug delivery target area based on the lateral skull base target area candidate clusters, lateral skull base risk structure clusters and lateral skull base safe access clusters; The targeted drug delivery pathway optimization module receives the tinnitus and vertigo phenotype clustering results and the lateral skull base target drug delivery area. Based on the lateral skull base target drug delivery area, lateral skull base safety pathway cluster, lateral skull base risk structure cluster, probe posture data, and drug delivery constraint data, it calculates the targeted drug delivery pathway safety score and generates a lateral skull base targeted drug delivery plan. The drug administration feedback clustering update module receives the lateral skull base targeted drug administration protocol, acquires drug administration process monitoring data, re-examination tinnitus and vertigo assessment data, and manual review results, binds the lateral skull base targeted drug administration protocol, drug administration process monitoring data, re-examination tinnitus and vertigo assessment data, and manual review results into a lateral skull base targeted drug administration sample, and updates the parameters of the tinnitus and vertigo phenotype clustering module, the lateral skull base target area clustering localization module, and the targeted drug administration pathway optimization module based on the lateral skull base targeted drug administration sample.

[0008] Furthermore, the tinnitus and vertigo phenotype clustering module normalizes and splices together tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data, previous medication records data, and lateral skull base fusion feature matrix to generate tinnitus and vertigo phenotype feature vectors, and generates tinnitus and vertigo phenotype clustering results based on the phenotypic distance between the tinnitus and vertigo phenotype feature vectors and each phenotype cluster center.

[0009] Furthermore, the lateral skull base target area clustering and localization module extracts multiple candidate region feature points based on the lateral skull base fusion feature matrix, performs density clustering on the candidate region feature points, generates lateral skull base target area candidate clusters, lateral skull base risk structure clusters, and lateral skull base safe access clusters, and calculates the target area clustering score based on the cluster density, risk structure distance, acoustic shadow coverage ratio, and accessibility of the candidate region clusters.

[0010] Furthermore, the targeted drug delivery pathway optimization module generates multiple candidate drug delivery pathways, calculates the drug delivery pathway risk based on the positional relationship between the candidate drug delivery pathways and the risk structure clusters at the lateral skull base, acoustic interference features, and bony boundary features, and calculates the targeted drug delivery pathway safety score based on the drug delivery pathway risk, target area clustering score, drug dosage level, and probe posture data.

[0011] Furthermore, the drug administration feedback clustering update module updates the phenotypic cluster center, density cluster radius, minimum sample number threshold, target area clustering score parameters, and targeted drug administration route safety score parameters based on the lateral skull base targeted drug administration samples.

[0012] The beneficial effects of this invention are as follows: (1) Using the principle of ultrasound multi-source feature fusion, the lateral skull base ultrasound data acquisition module acquires basic two-dimensional ultrasound images, color Doppler images, energy Doppler images, spectral Doppler data, probe posture data and acquisition parameter data to generate lateral skull base ultrasound raw data; the lateral skull base ultrasound raw data is processed by the multi-source feature fusion construction module to perform noise reduction and enhancement, blood flow region separation, acoustic shadow region marking, grayscale normalization and spatial registration, and extracts bony boundary features, blood flow course features, soft tissue interface features, acoustic shadow interference features and local echo texture features to generate lateral skull base fusion feature matrix. The technical effect is to improve the image expression integrity of the lateral skull base drug delivery area and solve the problem that the existing drug delivery positioning relies only on a single image or human experience and is difficult to comprehensively judge the lateral skull base tissue structure and blood flow risk.

[0013] (2) Using the principle of tinnitus and vertigo phenotype clustering, the tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data, previous drug administration record data and lateral skull base fusion feature matrix are normalized and spliced ​​through the tinnitus and vertigo phenotype clustering module to generate tinnitus and vertigo phenotype feature vectors. The tinnitus and vertigo phenotype clustering results are generated based on the phenotype distance between the tinnitus and vertigo phenotype feature vectors and the cluster centers of each phenotype. The technical effect is to achieve group adaptation for different drug administration subjects such as tinnitus-dominant type, vertigo-dominant type, tinnitus and vertigo mixed type and high recurrence risk type, and solve the problem that the existing drug administration regimen does not fully combine symptom phenotype, hearing status and vestibular function status for individual differentiation.

[0014] (3) Utilizing the principle of spatial clustering localization and safe drug delivery route optimization of the lateral skull base, density clustering is performed on the feature points of the candidate region through the lateral skull base target area clustering localization module to generate candidate clusters of lateral skull base target areas, risk structure clusters of lateral skull base, and safe pathway clusters of lateral skull base. The target drug delivery area of ​​the lateral skull base is selected based on the target area clustering score. Multiple candidate drug delivery routes are generated through the targeted drug delivery route optimization module. The safety score of the targeted drug delivery route is calculated based on the drug delivery route risk, target area clustering score, drug dosage level, and probe posture data to generate a lateral skull base targeted drug delivery plan. The technical effect is to improve the safety of lateral skull base target drug delivery target area localization and drug delivery route selection, and to solve the problem that existing drug delivery routes are difficult to avoid areas adjacent to blood vessels, bony obstruction areas, and acoustic interference areas. Attached Figure Description

[0015] The accompanying drawings are provided to further understand the present invention and form part of the specification. They are used together with the embodiments of the present invention to explain the invention and do not constitute a limitation thereof.

[0016] Figure 1 This is a flowchart of the multi-source feature fusion and phenotypic clustering proposed in this invention; Figure 2 This is a flowchart of the clustering and localization process for the lateral skull base target area proposed in this invention; Figure 3 This is a flowchart of the targeted drug delivery pathway optimization and feedback update proposed in this invention. Detailed Implementation

[0017] Example 1, see Figures 1-3 The present invention provides an ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo, comprising a lateral skull base ultrasound data acquisition module, a multi-source feature fusion construction module, a tinnitus and vertigo phenotype clustering module, a lateral skull base target area clustering and localization module, a targeted drug delivery pathway optimization module, and a drug delivery feedback clustering update module.

[0018] The lateral skull base ultrasound data acquisition module acquires basic two-dimensional ultrasound images, color Doppler images, energy Doppler images, spectral Doppler data, probe posture data, and acquisition parameter data of the lateral skull base region of the drug recipient, generates raw lateral skull base ultrasound data, and sends the raw lateral skull base ultrasound data to the multi-source feature fusion construction module.

[0019] The multi-source feature fusion construction module receives raw ultrasound data from the lateral skull base and performs noise reduction and enhancement, blood flow region separation, acoustic shadow region marking, grayscale normalization, and spatial registration on the basic two-dimensional ultrasound image, color Doppler image, energy Doppler image, and spectral Doppler data. It extracts bony boundary features, blood flow course features, soft tissue interface features, acoustic shadow interference features, and local echo texture features to generate a lateral skull base fusion feature matrix. The lateral skull base fusion feature matrix is ​​then sent to the tinnitus and vertigo phenotype clustering module and the lateral skull base target area clustering and localization module.

[0020] The tinnitus and vertigo phenotype clustering module receives the lateral skull base fusion feature matrix, obtains tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data, and previous drug administration record data of the drug administration subjects, performs tinnitus and vertigo phenotype grouping on the drug administration subjects based on the clustering algorithm, generates tinnitus and vertigo phenotype clustering results, and sends the tinnitus and vertigo phenotype clustering results to the targeted drug administration pathway optimization module.

[0021] The lateral skull base target area clustering and localization module receives the lateral skull base fusion feature matrix and performs spatial clustering on candidate drug delivery areas, vascular adjacent areas, bony obstruction areas, acoustic interference areas, and safe access areas in the lateral skull base region based on a clustering algorithm. This generates lateral skull base target area candidate clusters, lateral skull base risk structure clusters, and lateral skull base safe access clusters. The module then generates the lateral skull base target drug delivery target area based on these clusters.

[0022] The targeted drug delivery pathway optimization module receives the tinnitus and vertigo phenotype clustering results and the lateral skull base target drug delivery area. Based on the lateral skull base target drug delivery area, lateral skull base safety pathway cluster, lateral skull base risk structure cluster, probe posture data, and drug dosage constraint data, it calculates the targeted drug delivery pathway safety score and generates a lateral skull base targeted drug delivery plan. The lateral skull base targeted drug delivery plan includes the drug delivery target area, drug delivery pathway, drug dosage level, drug delivery frequency, drug delivery cycle, and follow-up time.

[0023] The drug administration feedback clustering update module receives the lateral skull base targeted drug administration protocol, acquires drug administration process monitoring data, re-examination tinnitus and vertigo assessment data, and manual review results, binds the lateral skull base targeted drug administration protocol, drug administration process monitoring data, re-examination tinnitus and vertigo assessment data, and manual review results into a lateral skull base targeted drug administration sample, and updates the parameters of the tinnitus and vertigo phenotype clustering module, the lateral skull base target area clustering localization module, and the targeted drug administration pathway optimization module based on the lateral skull base targeted drug administration sample.

[0024] With the cooperation of the above modules, this system can combine ultrasound image recognition, tinnitus and vertigo phenotype clustering, lateral skull base target area clustering and localization, drug delivery route safety optimization and feedback updates to improve the individualized adaptation and localization stability of lateral skull base targeted drug delivery.

[0025] Example 2: This example is based on all the above examples. The lateral skull base ultrasound data acquisition module specifically includes: Ultrasound image acquisition; the lateral skull base ultrasound data acquisition module acquires basic two-dimensional ultrasound images, color Doppler images, energy Doppler images, and spectral Doppler data of the lateral skull base region of the drug recipient.

[0026] The acquisition status is recorded; the lateral skull base ultrasound data acquisition module synchronously records probe posture data and acquisition parameter data. The probe posture data includes probe tilt angle data, probe rotation angle data, and probe compression state data. The acquisition parameter data includes ultrasound depth, gain parameter, focal depth, and frame rate data.

[0027] Image quality screening: The lateral skull base ultrasound data acquisition module performs sharpness detection, motion artifact detection, acoustic shadowing detection, and blood flow display integrity detection on basic two-dimensional ultrasound images, color Doppler images, and energy Doppler images, and removes image frames with image quality below the quality threshold to obtain valid lateral skull base ultrasound image frames.

[0028] Blood flow display data extraction: The lateral skull base ultrasound data acquisition module extracts the lateral skull base blood flow display area from color Doppler images and power Doppler images to obtain lateral skull base blood flow display data.

[0029] Timestamp binding: The lateral skull base ultrasound data acquisition module binds the effective lateral skull base ultrasound image frames, lateral skull base blood flow display data, spectral Doppler data, probe posture data, and acquisition parameter data according to the image frame timestamps to generate the raw lateral skull base ultrasound data.

[0030] Example 3: This example is based on all the above examples. The multi-source feature fusion construction module specifically includes: The module receives raw lateral skull base ultrasound data; the multi-source feature fusion construction module receives the raw lateral skull base ultrasound data sent by the lateral skull base ultrasound data acquisition module.

[0031] Image enhancement processing: The multi-source feature fusion construction module performs speckle noise suppression, edge enhancement, local contrast enhancement and grayscale normalization on the effective lateral skull base ultrasound image frames to obtain enhanced lateral skull base ultrasound image frames.

[0032] Blood flow region separation; the multi-source feature fusion construction module spatially registers the lateral skull base blood flow display data with the enhanced lateral skull base ultrasound image frame to separate the blood flow course region and obtain blood flow course features.

[0033] Acoustic shadow region marking: The multi-source feature fusion construction module marks the acoustic shadow interference region based on the low-echo tail region, echo attenuation region and boundary occlusion region in the basic two-dimensional ultrasound image, and obtains the acoustic shadow interference features.

[0034] Local structural feature extraction; the multi-source feature fusion construction module extracts bony boundary features, soft tissue interface features and local echo texture features from the enhanced lateral skull base ultrasound image frames.

[0035] The lateral skull base fusion feature matrix is ​​generated by stitching together bony boundary features, blood flow characteristics, soft tissue interface features, acoustic interference features, and local echo texture features according to image spatial coordinates.

[0036] Example 4: This example is based on all the above examples. The tinnitus and vertigo phenotype clustering module is one of the core modules of this invention, and specifically includes: Phenotypic data acquisition: The tinnitus and vertigo phenotype clustering module acquires tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data, and previous drug administration records of the drug-treated subjects.

[0037] Phenotypic feature vector generation: The tinnitus and vertigo phenotype clustering module normalizes and splices together tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data, previous drug administration records data, and lateral skull base fusion feature matrix to generate tinnitus and vertigo phenotype feature vectors.

[0038] Cluster center initialization: The tinnitus and vertigo phenotype clustering module initializes multiple phenotype cluster centers based on tinnitus-dominant samples, vertigo-dominant samples, mixed tinnitus and vertigo samples, and high-risk recurrence samples from historical tinnitus and vertigo drug administration samples.

[0039] Phenotypic distance calculation; The tinnitus and vertigo phenotype clustering module calculates the phenotypic distance between the tinnitus and vertigo phenotype feature vector and the cluster centers of each phenotype. The calculation formula is as follows: in, It is the phenotypic feature vector of tinnitus and vertigo and the first Phenotypic distance between phenotypic cluster centers; It is the first in the phenotype feature vector of tinnitus and vertigo. One eigenvalue; It is the first The first phenotypic cluster center One central eigenvalue; It is the first The weight coefficients corresponding to each feature value; It is the number of features in the phenotype feature vector of tinnitus and vertigo.

[0040] Tinnitus and vertigo phenotype clustering results are generated; the tinnitus and vertigo phenotype clustering module divides the drug-treated subjects into corresponding phenotype cluster centers according to phenotype distance, and generates tinnitus and vertigo phenotype clustering results.

[0041] Regarding parameter adjustments: Step 1: Adjust the number of cluster centers; When new symptom combinations appear in historical tinnitus and vertigo drug administration samples, corresponding phenotype cluster centers are added; When the phenotypic distance between multiple phenotypic cluster centers is continuously lower than the merging threshold, the corresponding phenotypic cluster centers will be merged.

[0042] Step 2: Feature weight adjustment; When tinnitus frequency data and tinnitus loudness data have a significant impact on the follow-up tinnitus and vertigo assessment data, the weighting coefficients corresponding to tinnitus frequency data and tinnitus loudness data should be increased. When the frequency and duration of vertigo attacks have a significant impact on the follow-up tinnitus and vertigo assessment data, the weighting coefficients corresponding to the frequency and duration of vertigo attacks should be increased.

[0043] Step 3: Handling abnormal samples; When the phenotypic distance between the tinnitus and vertigo phenotypic feature vector and the cluster center of all phenotypic groups is greater than the abnormal distance threshold, the sample is marked as a sample to be manually reviewed. When a sample awaiting manual verification is found to belong to a known phenotype after manual verification, the anomaly distance threshold is lowered and the corresponding phenotype cluster center is updated.

[0044] Step 4: Update phenotypic cluster centers; When the tinnitus and vertigo assessment data are reviewed and the drug regimen is found to be highly suitable, the corresponding tinnitus and vertigo phenotype feature vector will be included in the current phenotype cluster center. When a review of tinnitus and vertigo assessment data indicates low suitability of the medication regimen, the update weight of the corresponding sample for the current phenotypic cluster center is reduced.

[0045] By performing the above operations, the tinnitus and vertigo phenotype clustering module can group the subjects according to their tinnitus and vertigo symptoms, hearing status, vestibular function status, and previous drug administration, providing an individualized basis for the generation of subsequent targeted drug administration regimens.

[0046] Example 5 This embodiment is based on all the above embodiments. The lateral skull base target area clustering and localization module is another core module of the present invention, specifically including: The lateral skull base fusion feature matrix is ​​received; the lateral skull base target area clustering and localization module receives the lateral skull base fusion feature matrix sent by the multi-source feature fusion construction module.

[0047] Candidate region feature point generation; The lateral skull base target area clustering and localization module extracts multiple candidate region feature points based on the bony boundary features, blood flow characteristics, soft tissue interface features, acoustic interference features, and local echo texture features in the lateral skull base fusion feature matrix.

[0048] Density clustering processing: The lateral skull base target area clustering and localization module performs density clustering processing on the candidate region feature points, aggregating spatially adjacent and similar candidate region feature points into candidate region clusters.

[0049] Lateral skull base target area candidate cluster generation: The lateral skull base target area clustering and localization module selects candidate region clusters located in the vicinity of bony boundaries, continuous soft tissue interfaces and with low blood flow risk from the candidate region clusters, and generates lateral skull base target area candidate clusters.

[0050] Lateral skull base risk structure clusters are generated; the lateral skull base target area clustering and localization module selects candidate region clusters with high hemorrhage intensity, obvious acoustic interference coverage, or obvious bony boundary obstruction from the candidate region clusters, and generates lateral skull base risk structure clusters.

[0051] Lateral skull base safety access cluster generation; the lateral skull base target area clustering and localization module filters out regional clusters that can connect the surface entry point with the lateral skull base target area candidate clusters based on the spatial interval between the lateral skull base target area candidate clusters and the lateral skull base risk structure clusters, and generates lateral skull base safety access clusters.

[0052] Target area clustering score calculation: The lateral skull base target area clustering localization module calculates the target area clustering score based on the intra-cluster density of candidate region clusters, distance to risk structures, acoustic shadowing coverage ratio, and pathway accessibility. The calculation formula is as follows: in, It is a target region clustering score; It is the intra-cluster compactness of the candidate region cluster; It is the risk structure distance between the candidate region cluster and the lateral skull base risk structure cluster; It refers to the sound and image coverage ratio; It refers to accessibility; , , and These are the weight coefficients for the corresponding items.

[0053] The target area for drug delivery to the lateral skull base is generated; the target area clustering and localization module selects the target area for drug delivery to the lateral skull base from the candidate clusters of target areas based on the target area clustering score.

[0054] Regarding parameter adjustments: Step 1: Adjust the density cluster radius; When the feature points of the candidate region are sparsely distributed and the boundary of the real target region is continuous, the density clustering radius is increased so that the feature points of adjacent candidate regions can be merged into candidate region clusters. When non-target area tissue feature points are mistakenly aggregated, the density clustering radius is reduced and the feature similarity constraint is increased.

[0055] Step 2: Adjusting the minimum sample size; When candidate clusters of the lateral skull base target area are segmented into multiple small clusters, the minimum sample number threshold is reduced, and cluster merging is performed based on continuous regions of the soft tissue interface. When false clusters are formed in areas of sound and shadow interference, the minimum sample size threshold is increased, and samples are removed based on the sound and shadow coverage ratio.

[0056] Step 3: Adjusting the risk structure distance weights; When blood flow characteristics are close to candidate region clusters, increase the weight corresponding to the distance of risk structures, so that the target drug delivery area of ​​the lateral skull base is far away from the risk structure cluster of the lateral skull base; When the confidence level of identifying risky structures at the lateral skull base is low, the boundary of the risky structures at the lateral skull base is expanded, and the target area clustering score of adjacent candidate region clusters is reduced.

[0057] Step 4: Adjusting accessibility weights; When the lateral skull base safety access cluster is continuous and does not pass through acoustic shadowing interference areas, increase the weight corresponding to access accessibility; When the lateral skull base safety pathway cluster is blocked by bony boundaries or significantly covered by acoustic shadowing, the weight corresponding to pathway accessibility is reduced, and the target drug delivery area of ​​the lateral skull base is re-screened.

[0058] By performing the above operations, the lateral skull base target area clustering and localization module can use clustering algorithms to distinguish candidate target areas, risk structures, and safe pathways under ultrasound image guidance, thereby improving the localization reliability of the lateral skull base target drug delivery area.

[0059] Example 6: This example is based on all the above examples. The targeted drug delivery pathway optimization module specifically includes: The localization result is received; the targeted drug delivery route optimization module receives the tinnitus and vertigo phenotype clustering results and the target drug delivery area at the lateral skull base.

[0060] Drug delivery pathway candidate line generation: The targeted drug delivery pathway optimization module generates multiple drug delivery pathway candidate lines, starting from the entry point on the body surface and ending at the target drug delivery area at the lateral skull base.

[0061] Pathway risk calculation: The targeted drug delivery path optimization module calculates the drug delivery path risk based on the minimum distance between the candidate drug delivery path and the risk structure cluster at the lateral skull base, the degree of overlap between the candidate drug delivery path and acoustic interference features, and the overlap relationship between the candidate drug delivery path and bony boundary features.

[0062] Dosage level matching; the targeted drug delivery pathway optimization module generates dosage levels based on tinnitus and vertigo phenotype clustering results, lateral skull base target delivery area, and dosage constraint data.

[0063] Targeted drug delivery route safety score calculation: The targeted drug delivery route optimization module calculates the targeted drug delivery route safety score based on drug delivery route risk, target area clustering score, drug dosage level, and probe posture data. The calculation formula is as follows: in, It is a safety score for targeted drug delivery routes; It is a target region clustering score; It is a risk associated with the route of drug administration; It refers to the dosage level; It's a probe attitude deviation; , , and These are the weight coefficients for the corresponding items.

[0064] Lateral skull base targeted drug delivery protocol generation; The targeted drug delivery route optimization module sorts multiple drug delivery route candidate lines according to the targeted drug delivery route safety score, selects the drug delivery route candidate line with the highest targeted drug delivery route safety score and meets the safety threshold, and generates a lateral skull base targeted drug delivery protocol by combining the tinnitus and vertigo phenotype clustering results.

[0065] Example 7: This example is based on all the above examples. The drug administration feedback clustering update module specifically includes: The dosing regimen is received; the dosing feedback clustering update module receives the lateral skull base targeted dosing regimen.

[0066] Data acquisition during drug administration; the drug administration feedback clustering update module acquires real-time ultrasound image data, target area offset data, local blood flow change data, local tissue echo change data, and operational status data during the drug administration process.

[0067] Data acquisition for follow-up tinnitus and vertigo assessment; data acquisition for follow-up tinnitus frequency, tinnitus loudness, vertigo attack frequency, vertigo duration, hearing test, and vestibular function test using the drug administration feedback clustering update module.

[0068] Lateral skull base targeted drug delivery sample generation; the drug delivery feedback clustering update module binds the lateral skull base targeted drug delivery protocol, drug delivery process monitoring data, follow-up tinnitus and vertigo assessment data, and manual review results to generate lateral skull base targeted drug delivery samples.

[0069] Sample clustering update; the drug administration feedback clustering update module updates the phenotypic cluster center, density cluster radius, minimum sample number threshold, target area clustering score parameters, and targeted drug administration route safety score parameters based on the lateral skull base targeted drug administration samples.

[0070] Abnormal sample screening: The dosing feedback clustering update module filters out phenotypic clustering error samples, target area clustering error samples, path risk misjudgment samples, and dosing regimen mismatch samples from the lateral skull base targeted dosing samples.

[0071] Parameter updates: The drug delivery feedback clustering update module updates the parameters of the tinnitus and vertigo phenotype clustering module based on phenotypic clustering error samples, updates the parameters of the lateral skull base target area clustering localization module based on target area clustering error samples, and updates the parameters of the targeted drug delivery route optimization module based on path risk misjudgment samples.

[0072] Through the above feedback and update process, this system can continuously correct phenotypic clustering, target area clustering, and pathway safety score parameters based on the monitoring results of the drug administration process and the results of the follow-up tinnitus and vertigo assessment, thereby improving the individualized adaptation capability and safety of subsequent lateral skull base targeted drug administration regimens.

Claims

1. A lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo under ultrasound guidance, characterized in that: It includes a lateral skull base ultrasound data acquisition module, a multi-source feature fusion construction module, a tinnitus and vertigo phenotype clustering module, a lateral skull base target area clustering and localization module, a targeted drug delivery pathway optimization module, and a drug delivery feedback clustering update module; The lateral skull base ultrasound data acquisition module acquires basic two-dimensional ultrasound images, color Doppler images, energy Doppler images, spectral Doppler data, probe posture data, and acquisition parameter data of the lateral skull base region of the drug recipient, and generates raw lateral skull base ultrasound data. The multi-source feature fusion construction module receives raw data from lateral skull base ultrasound and performs noise reduction and enhancement, blood flow region separation, acoustic shadow region marking, grayscale normalization, and spatial registration on basic two-dimensional ultrasound images, color Doppler images, energy Doppler images, and spectral Doppler data. It then extracts bony boundary features, blood flow course features, soft tissue interface features, acoustic shadow interference features, and local echo texture features to generate a lateral skull base fusion feature matrix. The tinnitus and vertigo phenotype clustering module receives the side skull base fusion feature matrix, obtains the tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data and previous drug administration record data of the drug administration subjects, and performs tinnitus and vertigo phenotype grouping on the drug administration subjects based on the clustering algorithm to generate tinnitus and vertigo phenotype clustering results. The lateral skull base target area clustering and localization module receives the lateral skull base fusion feature matrix, and performs spatial clustering of candidate drug delivery areas, vascular adjacent areas, bony barrier areas, acoustic interference areas and safe access areas in the lateral skull base region based on the clustering algorithm, generating lateral skull base target area candidate clusters, lateral skull base risk structure clusters and lateral skull base safe access clusters, and generates the lateral skull base target drug delivery target area based on the lateral skull base target area candidate clusters, lateral skull base risk structure clusters and lateral skull base safe access clusters; The targeted drug delivery pathway optimization module receives the tinnitus and vertigo phenotype clustering results and the lateral skull base target drug delivery area. Based on the lateral skull base target drug delivery area, lateral skull base safety pathway cluster, lateral skull base risk structure cluster, probe posture data, and drug delivery constraint data, it calculates the targeted drug delivery pathway safety score and generates a lateral skull base targeted drug delivery plan. The drug administration feedback clustering update module receives the lateral skull base targeted drug administration protocol, acquires drug administration process monitoring data, re-examination tinnitus and vertigo assessment data, and manual review results, binds the lateral skull base targeted drug administration protocol, drug administration process monitoring data, re-examination tinnitus and vertigo assessment data, and manual review results into a lateral skull base targeted drug administration sample, and updates the parameters of the tinnitus and vertigo phenotype clustering module, the lateral skull base target area clustering localization module, and the targeted drug administration pathway optimization module based on the lateral skull base targeted drug administration sample.

2. The ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo according to claim 1, characterized in that: The lateral skull base ultrasound data acquisition module performs sharpness detection, motion artifact detection, acoustic shadowing detection, and blood flow display integrity detection on the basic two-dimensional ultrasound images, color Doppler images, and energy Doppler images. Image frames with image quality below the quality threshold are removed to obtain valid lateral skull base ultrasound image frames. The lateral skull base blood flow display area is extracted from the color Doppler images and energy Doppler images to obtain lateral skull base blood flow display data. The valid lateral skull base ultrasound image frames, lateral skull base blood flow display data, spectral Doppler data, probe posture data, and acquisition parameter data are bound according to the image frame timestamps to generate the raw lateral skull base ultrasound data.

3. The ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo according to claim 2, characterized in that: The multi-source feature fusion construction module performs speckle noise suppression, edge enhancement, local contrast enhancement, and grayscale normalization on the effective lateral skull base ultrasound image frames to obtain enhanced lateral skull base ultrasound image frames; spatial registration is performed between the lateral skull base blood flow display data and the enhanced lateral skull base ultrasound image frames to separate the blood flow course region and obtain blood flow course features; Based on the hypoechoic tail region, echo attenuation region, and boundary occlusion region in the basic two-dimensional ultrasound image, the acoustic shadowing interference region is marked, and acoustic shadowing interference features are obtained. Bone boundary features, soft tissue interface features, and local echo texture features are extracted from the enhanced lateral skull base ultrasound image frame. The bone boundary features, blood flow course features, soft tissue interface features, acoustic shadowing interference features, and local echo texture features are stitched together according to the image space coordinates to generate the lateral skull base fusion feature matrix.

4. The ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo according to claim 3, characterized in that: The tinnitus and vertigo phenotype clustering module normalizes and concatenates tinnitus frequency data, tinnitus loudness data, vertigo attack frequency data, vertigo duration data, hearing test data, vestibular function test data, previous drug administration records, and lateral skull base fusion feature matrix to generate a tinnitus and vertigo phenotype feature vector. Multiple phenotype clustering centers are initialized based on tinnitus-dominant, vertigo-dominant, mixed tinnitus and vertigo, and high-risk recurrence samples from historical tinnitus and vertigo drug administration samples. The phenotypic distance between the tinnitus and vertigo phenotype feature vector and each phenotype clustering center is calculated, and the drug-administered subjects are assigned to the corresponding phenotype clustering centers based on the phenotypic distance, generating tinnitus and vertigo phenotype clustering results.

5. The lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo under ultrasound guidance according to claim 4, characterized in that: The lateral skull base target area clustering and localization module extracts multiple candidate region feature points based on bony boundary features, blood flow characteristics, soft tissue interface features, acoustic interference features, and local echo texture features in the lateral skull base fusion feature matrix; density clustering is performed on the candidate region feature points to aggregate spatially adjacent and feature-similar candidate region feature points into candidate region clusters.

6. The ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo according to claim 5, characterized in that: The lateral skull base target area clustering and localization module selects candidate region clusters from the candidate region clusters that are located in the region adjacent to the bony boundary, the continuous region of the soft tissue interface and have low blood flow risk, and generates lateral skull base target area candidate clusters; it also selects candidate region clusters with high blood flow intensity, obvious acoustic shadowing interference coverage or obvious bony boundary obstruction from the candidate region clusters, and generates lateral skull base risk structure clusters. Based on the spatial interval between the candidate clusters of the lateral skull base target area and the clusters of risk structures of the lateral skull base, regional clusters that can connect the surface entry points with the candidate clusters of the lateral skull base target area are selected to generate lateral skull base safe access clusters.

7. The ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo according to claim 6, characterized in that: The lateral skull base target area clustering and localization module calculates the target area clustering score based on the cluster density, risk structure distance, acoustic shadow coverage ratio, and pathway accessibility of the candidate region clusters, and selects the lateral skull base target drug delivery target area from the candidate clusters of the lateral skull base target area based on the target area clustering score.

8. The ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo according to claim 7, characterized in that: The targeted drug delivery pathway optimization module generates multiple candidate drug delivery pathways starting from the entry point on the body surface and ending at the target drug delivery area at the lateral skull base. It calculates the drug delivery pathway risk based on the minimum distance between the candidate drug delivery pathway and the risk structure cluster at the lateral skull base, the degree of overlap between the candidate drug delivery pathway and acoustic interference features, and the overlap relationship between the candidate drug delivery pathway and bony boundary features. It generates a drug delivery level based on the tinnitus and vertigo phenotype clustering results, the target drug delivery area at the lateral skull base, and the drug delivery constraint data. Finally, it calculates a targeted drug delivery pathway safety score based on the drug delivery pathway risk, target area clustering score, drug delivery level, and probe posture data.

9. The ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo according to claim 8, characterized in that: The targeted drug delivery pathway optimization module sorts multiple candidate drug delivery pathways based on the targeted drug delivery pathway safety score, selects the candidate drug delivery pathway with the highest targeted drug delivery pathway safety score that meets the safety threshold, and generates a lateral skull base targeted drug delivery plan by combining the tinnitus and vertigo phenotype clustering results. The lateral skull base targeted drug delivery plan includes the drug delivery target area, drug delivery pathway, drug delivery dose level, drug delivery frequency, drug delivery cycle, and follow-up time.

10. The ultrasound-guided lateral skull base targeted drug delivery system for the effective treatment of tinnitus and vertigo according to claim 9, characterized in that: The drug administration feedback clustering update module binds the lateral skull base targeted drug administration protocol, drug administration process monitoring data, re-examination tinnitus and vertigo assessment data, and manual review results to generate lateral skull base targeted drug administration samples. Based on these samples, it updates the phenotypic cluster center, density cluster radius, minimum sample number threshold, target area clustering score parameters, and targeted drug administration route safety score parameters. From the lateral skull base targeted drug administration samples, it filters out phenotypic clustering error samples, target area clustering error samples, route risk misjudgment samples, and drug administration protocol mismatch samples. Based on the phenotypic clustering error samples, it updates the parameters of the tinnitus and vertigo phenotypic clustering module; based on the target area clustering error samples, it updates the parameters of the lateral skull base target area clustering localization module; and based on the route risk misjudgment samples, it updates the parameters of the targeted drug administration route optimization module.

Images