Method for judging affected side of benign paroxysmal positional vertigo of terrestrial horizontal tubestone

Abstract

The invention belongs to the field of computer-aided diagnosis, and relates to a method for judging the affected side of terrestrial horizontal tubulous benign paroxysmal positional vertigo, which combines artificial intelligence with a bending and supine test and a rolling test, and performs simple and effective affected side judgment on the horizontal tubulous benign paroxysmal positional vertigo by utilizing a nystagmus incubation period. The method comprises the steps of creating an affected side judgment model, testing and collecting nystagmus data, preprocessing the nystagmus data, conducting affected side judgment, combining a neck bending test, a sitting and lying position test and a 180-degree rolling test, collecting the nystagmus data through collecting equipment, and inputting the nystagmus data into the affected side judgment model for affected side judgment. The method disclosed by the invention is scientific and reasonable, simple and coherent in operation steps, low in operation cost, high in efficiency, small in interference and high in practicability, optimizes the determination method of the nystagmus incubation period, meets the clinical convenience requirement, and is suitable for large-scale popularization and application.

Description

Technical field:

[0001] The present invention belongs to the field of computer-aided diagnosis, and relates to a method for determining the affected side of geotropic horizontal canalite benign paroxysmal positional vertigo, which is used for making a simple and effective affected side judgment on the horizontal canalite benign paroxysmal positional vertigo. Background technology:

[0002] Benign paroxysmal positional vertigo (BPPV) is a common peripheral vestibular disease in clinical practice and the most common vertigo disease originating from the inner ear. When the head moves to a certain position, it can induce a short period of vertigo, accompanied by nystagmus and autonomic nervous system symptoms. It can be seen in all age groups, but is more common in the elderly. The incidence of posterior semicircular canal BPPV is the highest, followed by horizontal canalolith BPPV. Since nystagmus can be induced on both sides of patients with horizontal canalolith BPPV during the roll test, and the intensity of nystagmus on both sides of some patients is not significantly different, it is difficult to directly determine the affected side. About 10-19.6% of cases induce symmetrical nystagmus, with an accuracy rate of 62-75%. Failure to determine the affected side may cause the patient to transform into horizontal crestal BPPV after repositioning treatment, and crestoliths are more difficult to reposition than canaloliths, which increases the patient's pain. The key to the successful reduction of horizontal canalolithic BPPV lies in the determination of the affected side, so many auxiliary methods have been developed to determine the affected side: head-down and head-up test, sitting and supine test, supine mid-position 3 nystagmus, pseudo-spontaneous nystagmus, nystagmus disappearance plane, head shaking test, bow and supine test (BST), etc. The existing auxiliary determination methods are still controversial, or the induction rate is limited, or the method is cumbersome. When various previous methods cannot determine the affected side, it is necessary to repeat the test and combine it with vestibular function examination for comprehensive evaluation. If the patient's physical condition allows, the reduction treatment can be tried, but this operation method will bring physical and economic burden to the patient.

[0003] Tianjin Medical University and Tianjin Institute of Otolaryngology reported that the latency of nystagmus (LON) of 179 patients with horizontal canalolithic BPPV who underwent a 90° roll test (RT) to turn their heads to the affected and healthy sides was 1.922±1.501s and 1.447±0.855s, respectively; Hebei University reported that the LON of 93 patients with horizontal canalolithic BPPV who underwent a 90° RT to turn their heads to the affected and healthy sides was 2.01±0.89s and 1.85±0.76s, respectively. The data show that the LON of turning the head to the affected side in both groups is basically greater than that to the healthy side, but the difference between the affected and healthy sides of LON is not significant compared with this technology. Since the judgment of LON is affected by factors such as the speed of the patient turning over or turning his head to both sides, the patient's blinking, the sliding of the video goggles during rolling, and the error of the physician's judgment, the difference in LON between the two sides is even more unclear, so the judgment of the affected side of horizontal canalolithic BPPV based on LON has not been widely used.

[0004] In the prior art, Chinese Patent CN111191639A discloses a method, device, medium and electronic device for identifying the type of vertigo based on nystagmus. By acquiring the nystagmus video of a vertigo patient and extracting the nystagmus signal from the nystagmus video, where the nystagmus signal includes the horizontal nystagmus signal and the vertical nystagmus signal, and finally identifying the type of vertigo of the vertigo patient according to the nystagmus signals in each direction. The specific way to identify the type of vertigo is to input the nystagmus signal into a neural network model, and the neural network model automatically identifies the type of vertigo of the vertigo patient. On the premise of considering the nystagmus signal, using the neural network model to identify the type of vertigo of the vertigo patient can achieve automatic identification of the type of vertigo. However, it cannot be used to determine the affected side.

[0005] Chinese Patent CN115956931A discloses a BPPV diagnosis and treatment system and its application method, including a posture sensing module, an electromyogram acquisition module and / or an eye movement acquisition module. By changes in head position and eye movement, it judges the movement and position of otoliths. Through the BPPV virtual simulation module, it imports the model of the semicircular canal, adds otoliths at the specified position as needed, synchronously displays the movement of otoliths, and realizes interaction through language and text, and can be applied in various scenarios such as home and diagnosis and treatment places. The application method of the BPPV diagnosis and treatment system of the invention can guide the patient to make a first specified action, and judge the position and movement trajectory of the otolith through the posture sensing module, the electromyogram acquisition module and the eye movement acquisition module, so as to judge whether BPPV is diagnosed and prompt the subsequent operation steps; or guide the patient to make a second specified action, which can reset the otoliths in different positions of the three semicircular canals and will not cause otolith semicircular canal conversion and otolith re-entry. However, its manufacturing cost is high, the use steps are cumbersome, it cannot simply and quickly determine the affected side, and it is not suitable for clinical medicine.

[0006] The above prior art has the problem that it cannot quickly and accurately determine the affected side of horizontal canal benign paroxysmal positional vertigo. After the inventor's retrieval and analysis, there is no method in the prior art that combines artificial intelligence with the flexion-supine test and the roll test and uses the nystagmus latency to determine the affected side of horizontal canal benign paroxysmal positional vertigo. Therefore, inventing a method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo can improve the deficiencies of the prior art, simply, quickly and accurately judge the affected side of horizontal canal benign paroxysmal positional vertigo, and meet the requirements of clinical convenience. Summary of the Invention:

[0007] The object of the present invention is to overcome the shortcomings of the prior art. Based on the improvement of the prior art, a method is designed and provided, which combines artificial intelligence with the flexion-supine test and the roll test, and uses the nystagmus latency to determine the affected side of horizontal canalolithiasis benign paroxysmal positional vertigo, especially a method for determining the affected side of geotropic horizontal canalolithiasis benign paroxysmal positional vertigo, solving the problems existing in the prior art.

[0008] To achieve the above object, the present invention relates to a method for determining the affected side of geotropic horizontal canalolithiasis benign paroxysmal positional vertigo, including the following steps:

[0009] S1: Create an affected side judgment model: Create an affected side judgment model for judging the affected side through the preprocessed nystagmus data;

[0010] S2: Experimentally collect nystagmus data: Conduct an experiment on a patient with horizontal canalolithiasis benign paroxysmal positional vertigo to trigger nystagmus, and use a collection device to record the nystagmus data generated by the patient and save it in a local storage device; the collection device is a Frenzel glasses with a camera recording function;

[0011] S3: Preprocess the nystagmus data: Read the nystagmus data of the patient saved in the local storage device, and preprocess the nystagmus data to obtain the preprocessed nystagmus data;

[0012] S4: Judge the affected side: Import the preprocessed nystagmus data into the affected side judgment model to judge the affected side.

[0013] The specific steps for creating the affected side judgment model described in S1 of the present invention specifically include:

[0014] S1-1: Construct an intensity ratio judgment module, calculate the nystagmus intensity ratio, and set the threshold trigger condition as the intensity ratio ≥ 1.5;

[0015] S1-2: Establish a BST direction comparison matrix, set the association between the body position relationship and the expected direction relationship, and set a direction conflict detection algorithm;

[0016] S1-3: Set the nystagmus latency calculation formula, and set a dynamic threshold for determining the affected side;

[0017] S1-4: Multimodal feature fusion: Temporal features, frequency domain features and spatial features are fused to construct a feature cross layer;

[0018] S1-5: Adopt a rule-neural network hybrid architecture, use the Drools rule engine to process explicit threshold judgments, design a 3-layer LSTM network to process temporal features, and construct an attention mechanism to process spatial features;

[0019] S1-6: Collect 500 clinically diagnosed samples as the training data set and perform segmented training on the overall model.

[0020] The test for triggering nystagmus in patients described in step S2 of the present invention specifically includes the following steps:

[0021] S2-1: Perform a flexion-supine test on the patient and collect the direction of the patient's nystagmus.

[0022] S2-2: After the nystagmus completely disappears, perform a 180° roll test on the patient and collect the intensity and latency of the patient's nystagmus.

[0023] The flexion-supine test described in step S2-1 of the present invention specifically includes the following steps:

[0024] S2-1-1: Perform a neck flexion test on the patient, making the patient sit on the examination bed and quickly changing the head from the upright position to a 90° flexion position.

[0025] S2-1-2: After the nystagmus completely disappears, perform a sitting-supine test on the patient, making the patient change from the sitting position to lying supine on a 30° inclined pillow.

[0026] The 180° roll test described in step S2-2 of the present invention specifically includes the following steps:

[0027] S2-2-1: Make the patient quickly turn over or turn the head 90° to one side from the supine position at the end of the flexion-supine test, and use the flipped position as the starting side.

[0028] S2-2-2: After the nystagmus completely disappears, make the patient rotate 180° from the starting side through the supine midline position to the opposite side, and collect and record the latency, direction, intensity, and duration of the nystagmus.

[0029] S2-2-3: After the nystagmus completely disappears, make the patient rotate 180° again through the supine midline position and return to the starting side, and collect and record the latency, direction, intensity, and duration of the nystagmus.

[0030] S2-2-4: If the nystagmus is not obvious or the collection is incomplete, repeat steps S2-2-2 to S2-2-3 to confirm that the nystagmus data collection is complete and correct.

[0031] The preprocessing of the nystagmus data described in step S3 of the present invention specifically includes the following steps:

[0032] S3-1: Denoise the nystagmus signal, apply independent component analysis to separate environmental light interference and blink artifacts, and remove baseline drift through wavelet transform.

[0033] S3-2: Extract parameters from the time domain, frequency domain, and direction features of nystagmus, calculate the time difference from the body position change to the first appearance of nystagmus as the nystagmus latency, calculate the time span from the start to the complete disappearance of nystagmus as the nystagmus duration, calculate the average velocity using the sliding window method, extract the dominant frequency components through FFT, establish a three-dimensional coordinate system, and use the Hough transform to identify the dominant plane of nystagmus;

[0034] S3-3: Align the nystagmus signal with the video recording time axis, and mark the key body position time points at the moment when the body position starts to change, the moment when the body position is stable, and the start and end moments of nystagmus;

[0035] S3-4: Segment the data into the flexion-supine stage and the roll test stage according to the body position change signal;

[0036] S3-5: Verify the nystagmus data to ensure that each test stage contains valid nystagmus cycles.

[0037] The use of the affected side judgment model for affected side judgment described in step S4 of the present invention specifically includes the following steps:

[0038] S4-1: If the ratio of the nystagmus intensities on both sides collected in the roll test is greater than or equal to 1.5, then use the side with stronger nystagmus in the nystagmus data collected in the roll test as the affected side;

[0039] S4-2: If the ratio of the nystagmus intensities on both sides collected in the roll test is less than 1.5, then make a judgment based on the nystagmus data collected in the flexion-supine test. The affected side direction is the same as the nystagmus direction in the head-down position and opposite to the nystagmus direction in the supine position in the nystagmus data collected in the flexion-supine test;

[0040] S4-3: If the nystagmus directions in the head-down position and the supine position in the nystagmus data collected in the flexion-supine test are the same, or no nystagmus is elicited, and the affected side cannot be judged, then make a judgment based on the nystagmus latency collected in the roll test. When the difference in nystagmus latency is greater than or equal to 1 s, use the side with the longer nystagmus latency as the affected side.

[0041] The nystagmus data involved in the present invention includes nystagmus direction, nystagmus intensity, and nystagmus latency.

[0042] The present invention also relates to a system for determining the affected side of horizontal canal benign paroxysmal positional vertigo for implementing the above method, including:

[0043] A data preprocessing module for performing relevant preprocessing on the nystagmus data to obtain preprocessed nystagmus data;

[0044] An image processing module for constructing a time-slow phase angular velocity SPV graph according to the preprocessed nystagmus data;

[0045] A feature extraction module for feature extraction, validating and screening a time-slow phase velocity (SPV) diagram, and obtaining the nystagmus latency, intensity, direction, and duration.

[0046] An intelligent data analysis platform equipped with a affected side judgment model to determine the affected side of geotropic horizontal canal benign paroxysmal positional vertigo.

[0047] The present invention also relates to a computer device, including a processor and a memory. The processor is used to execute a computer program stored in the memory to implement the method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo.

[0048] The present invention also relates to a computer-readable storage medium. The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo is implemented.

[0049] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention uses the prone-supine test as an auxiliary determination method, with an elicitation rate of 38% - 96% and a correct rate of 61 - 90%. Compared with other methods, both the elicitation rate and the correct rate are relatively high. The supine position of the sitting-supine position test and the first movement of the roll test are repeated, reducing the pain caused by the patient performing the test and improving the determination efficiency. Moreover, this test is simple to operate, only requiring observing the direction of nystagmus without observing indicators such as intensity. (2) The 180° roll test extends the running track of the otolith, making it easier to overcome the viscosity of the tube wall, generating a greater difference in nystagmus strength, thereby increasing the positive rate of the test and improving the examination efficiency. (3) The difference in nystagmus latency between the two sides of the 180° roll test is significant, which is convenient for collection and easy to judge the affected side with high accuracy. (4) The affected side judgment model created by the present invention can quickly and accurately judge the collected nystagmus data, improving the efficiency and accuracy. The method involved in the present invention is scientific and reasonable, the operation steps are simple and coherent, the operation cost is low, the efficiency is high, it optimizes the determination method of nystagmus latency, is less affected by interference, has strong practicability, meets the clinical convenience requirements, and is suitable for large-scale promotion and use. Description of the Drawings:

[0050] Figure 1 It is a flowchart showing the method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo according to the present invention;

[0051] Figure 2 It is a curve graph showing the accuracy of judging the affected side by BST, RT, and LON according to Embodiment 2 of the present invention.

[0052] Figure 3 It is a schematic diagram of the horizontal semicircular canal in the supine position according to Embodiment 2 of the present invention.

[0053] Description of the Drawings: LON represents the experimental data group for judging the affected side using nystagmus latency data, RT represents the experimental data group for judging the affected side using roll test data, and BST represents the experimental data group for judging the affected side using flexion-supine test data; the left horizontal semicircular canal A, the head position in the middle B, the right horizontal semicircular canal C, the short arm a, the forearm b, the posterior arm c, the ampulla d, the single foot e, and the utricle f. Detailed Implementation Manner:

[0054] The present invention will be further described below in conjunction with the drawings and embodiments.

[0055] Embodiment 1:

[0056] This embodiment relates to a method for determining the affected side of geotropic horizontal canalolith benign paroxysmal positional vertigo, including the following steps:

[0057] S1: Create an affected side judgment model for judging the affected side through preprocessed nystagmus data;

[0058] S2: Conduct a test on a patient with horizontal canalolith benign paroxysmal positional vertigo to trigger nystagmus, and use a collection device to record the nystagmus data generated by the patient and save it in a local storage device; the collection device is a Frenzel glasses with a camera recording function;

[0059] S3: Read the nystagmus data of the patient saved in the local storage device, and preprocess the nystagmus data to obtain the preprocessed nystagmus data;

[0060] S4: Import the preprocessed nystagmus data into the affected side judgment model for affected side judgment.

[0061] The specific steps for creating the affected side judgment model described in S1 of this embodiment specifically include:

[0062] S1-1: Construct an intensity ratio judgment module, calculate the nystagmus intensity ratio, and set the threshold trigger condition as intensity ratio ≥ 1.5;

[0063] S1-2: Establish a BST direction comparison matrix, set the association between the body position relationship and the expected direction relationship, and set a direction conflict detection algorithm;

[0064] S1-3: Set the nystagmus latency calculation formula and set a dynamic threshold for judging the affected side;

[0065] S1-4: Fuse multi-modal features of time domain features, frequency domain features, and spatial features to construct a feature cross layer;

[0066] S1-5: Adopt a rule-neural network hybrid architecture, use the Drools rule engine to handle explicit threshold judgments, design a 3-layer LSTM network to process temporal features, and construct an attention mechanism to process spatial features;

[0067] S1-6: Collect 500 clinically diagnosed samples as the training dataset and perform segmented training on the overall model.

[0068] The test for triggering nystagmus in patients described in step S2 of this embodiment specifically includes the following steps:

[0069] S2-1: Conduct a bow and supine test (BST) on the patient and collect the direction of the patient's nystagmus;

[0070] S2-2: After the nystagmus completely disappears, conduct a 180° roll test (RT) on the patient and collect the intensity and latency of the patient's nystagmus.

[0071] The bow and supine test described in step S2-1 of this embodiment specifically includes the following steps:

[0072] S2-1-1: Conduct a head-bending test (HBT) on the patient, making the patient sit on the examination bed and quickly change the head from the upright position to a 90° flexed position;

[0073] S2-1-2: After the nystagmus completely disappears, conduct a seated supinepositioning test (SSPT) on the patient, changing the patient from a sitting position to lying supine on a 30° inclined pillow.

[0074] The 180° roll test described in step S2-2 of this embodiment specifically includes the following steps:

[0075] S2-2-1: Make the patient quickly turn over or turn the head 90° to one side from the supine position at the end of the bow and supine test, and use the flipped position as the starting side;

[0076] S2-2-2: After the nystagmus completely disappears, make the patient rotate 180° from the starting side through the supine midline position to the opposite side, and collect and record the latency, direction, intensity, and duration of the nystagmus;

[0077] S2-2-3: After the nystagmus completely disappears, make the patient rotate 180° again through the supine midline position and return to the starting side, and collect and record the latency, direction, intensity, and duration of the nystagmus;

[0078] S2-2-4: If the nystagmus is not obvious or the acquisition is incomplete, repeat steps S2-2-2 to S2-2-3 to confirm that the nystagmus data is collected perfectly and correctly.

[0079] The preprocessing of the nystagmus data described in step S3 of this embodiment specifically includes the following steps:

[0080] S3-1: Denoise the nystagmus signal, apply independent component analysis to separate environmental light interference and blink artifacts, and remove baseline drift through wavelet transform;

[0081] S3-2: Extract parameters from the time domain, frequency domain, and direction features of the nystagmus, calculate the time difference from the body position change to the first appearance of nystagmus as the nystagmus latency, and calculate the time span from the start to the complete disappearance of the nystagmus as the duration of the nystagmus. Use the sliding window method to calculate the average speed, extract the dominant frequency component through FFT, establish a three-dimensional coordinate system, and use the Hough transform to identify the dominant plane of the nystagmus;

[0082] S3-3: Align the nystagmus signal with the video recording time axis, and mark the key body position time points of the start time of the body position change, the stable time of the body position, and the start and end times of the nystagmus;

[0083] S3-4: Segment the data into the flexion-supine stage and the roll test stage according to the body position change signal;

[0084] S3-5: Verify the nystagmus data to ensure that each test stage contains valid nystagmus cycles.

[0085] The use of the affected side judgment model to judge the affected side described in step S4 of this embodiment specifically includes the following steps:

[0086] S4-1: If the ratio of the nystagmus intensities on both sides collected in the roll test is greater than or equal to 1.5, then use the side with stronger nystagmus in the nystagmus data collected in the roll test as the affected side;

[0087] S4-2: If the ratio of the nystagmus intensities on both sides collected in the roll test is less than 1.5, then judge according to the nystagmus data collected in the flexion-supine test. The direction of the affected side is the same as the direction of the nystagmus in the head-down position in the nystagmus data collected in the flexion-supine test and opposite to the direction of the nystagmus in the supine position;

[0088] S4-3: If the direction of the nystagmus in the head-down position and the direction of the nystagmus in the supine position in the nystagmus data collected in the flexion-supine test are the same, or no nystagmus is elicited, and the affected side cannot be judged, then judge according to the nystagmus latency collected in the roll test. When the difference in nystagmus latency is greater than or equal to 1 s, use the side with the longer nystagmus latency as the affected side.

[0089] The nystagmus data involved in this embodiment includes nystagmus direction, nystagmus intensity, and nystagmus latency.

[0090] This embodiment also relates to a system for determining the affected side of horizontal canal benign paroxysmal positional vertigo for implementing the above method, including:

[0091] A data preprocessing module for performing relevant preprocessing on the nystagmus data to obtain preprocessed nystagmus data;

[0092] An image processing module for constructing a time-slow phase angular velocity (SPV) graph based on the preprocessed nystagmus data;

[0093] A feature extraction module for feature extraction, effectively screening the time-slow phase angular velocity (SPV) graph, and obtaining the nystagmus latency, intensity, direction, and duration;

[0094] An intelligent data analysis platform equipped with an affected side judgment model to determine the affected side of horizontal canal benign paroxysmal positional vertigo.

[0095] This embodiment also relates to a computer device, including a processor and a memory. The processor is used to execute a computer program stored in the memory to implement the method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo.

[0096] This embodiment also relates to a computer-readable storage medium. The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, it implements the method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo.

[0097] In the roll test involved in this embodiment, it should be ensured that the turning or head-turning speed and angle of both sides of the patient are as consistent as possible. When the patient is able to move, the turning operation should be preferentially performed.

[0098] This embodiment involves the process of reducing the patient according to the judgment result of the affected side. Barbecue and Gufoni reduction are performed according to the affected side. After the reduction is completed, a 180° RT verification is performed again after an interval of about 20 minutes. If the patient has no obvious vertigo and characteristic nystagmus, the reduction is consolidated once. If the patient still has vertigo within two days, a judgment should be made again. Those who do not return for a follow-up visit and those who have no characteristic nystagmus during the 180° RT follow-up visit are regarded as successful reduction; during the verification after the initial reduction, if the patient's vertigo is not significantly reduced and there is characteristic nystagmus, the affected side is determined according to the latter examination result for reduction. If the two judgment results are the same, the reduction is performed again; if the patient's vertigo is still not significantly reduced and there is characteristic nystagmus during the re-verification, the reduction is performed in the opposite direction.

[0099] Embodiment 2:

[0100] This embodiment relates to the method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo described in Embodiment 1, which is used to judge 91 patients with horizontal canal stone BPPV. As Figure 3 shown, the specific data are as follows:

[0101] The area under the ROC curve (AUC) of the LON group was 0.984 (0.949 - 1.000), the AUC of the RT group was 0.945 (0.883 - 1.000), and the AUC of the BST group was 0.929 (0.859 - 0.998). The AUCs were all greater than 0.9, proving that this method has high accuracy in judging the affected side in the roll test, the prone-supine test, and the nystagmus latency judgment method. There were statistically significant differences in the comparison of the ROC curves of the LON, RT, and BST groups with the reference line (P < 0.001). There was no difference in the comparison of the ROC curves between the LON group and the RT group (P = 0.157), no difference in the comparison of the ROC curves between the BST group and the LON group (P = 0.078), and no difference in the ROC curves between the BST group and the RT group (P = 0.661), indicating that there was no difference in the area under the curves of the LON, RT, and BST groups. For the 91 patients with horizontal canal stone BPPV involved in this embodiment, the nystagmus latencies of turning the head to the affected side and the healthy side were 3.071 ± 1.786 s and 1.7882 ± 0.8305 s respectively; Tianjin Medical University and Tianjin Institute of Otorhinolaryngology reported that for 179 patients with horizontal canal stone BPPV undergoing a 90° roll test, the nystagmus latencies of turning the head to the affected side and the healthy side were 1.922 ± 1.501 s and 1.447 ± 0.855 s respectively; Hebei University reported that for 93 patients with horizontal canal stone BPPV undergoing a 90° roll test, the nystagmus latencies of turning the head to the affected side and the healthy side were 2.01 ± 0.89 s and 1.85 ± 0.76 s respectively; in contrast, using the method involved in this embodiment, the difference in the nystagmus latencies on both sides was significant when the patients underwent a 180° roll test. The method involved in this embodiment can accurately judge the affected side of horizontal canal stone benign paroxysmal positional vertigo that is difficult to distinguish from multiple angles, is convenient and fast, and is easy to promote and use.

[0102] This embodiment also relates to the operating principle of the method described in Embodiment 1: As Figure 3The figure shows the short arm a, forearm b, posterior arm c, ampulla d, unipedal e, and utricle f. Excluding horizontal forearm bBPPV, when the patient performs 180°RT and initially lies supine, the otoliths are most likely located at the lowest point of gravity 1. If the patient turns to the affected side first, the otoliths flow from point 1 to point 3, and then to the healthy side when turning to the healthy side, the otoliths flow from point 3 to point 2, resulting in a LON that turns to the healthy side. When turning to the affected side again, the otoliths flow from point 2 to point 3, and then to the affected side. Since point 3 is closer to the ampulla d than point 2, the ampulla d ridge is stimulated first when turning to the healthy side than to the affected side, and the LON on the healthy side is shorter. If the patient turns to the healthy side first during 180°RT, the otoliths flow from point 1 to point 2, and then to the affected side when turning to the affected side, the otoliths flow from point 2 to point 3, and then to the healthy side when turning to the healthy side, the LON on the healthy side is still shorter. If 90°RT is used and the patient turns over from the supine mid-position each time, the otolith flow will be from 1 to 2 when turning to the healthy side, and from 1 to 3 when turning to the affected side, and there may be no significant difference in the LON on both sides.

[0103] This embodiment also involves the reason why "the difference in nystagmus latency is greater than or equal to 1s" is used as the criterion for the affected side in embodiment 1. According to previous studies, the time for the ampulla d crest to feel stimulation when the otolith flows from 1 (close to 3 / 4 of the semicircular canal) is close to 2s. Therefore, when the otolith is carried out at 180°RT in this embodiment, the otolith flows between 2 and 3, which is about the distance of half a semicircular canal. This time should be about 1s. Considering that the judgment of LON is affected by factors such as the patient's turning over or turning the head to both sides, the patient's blinking, the sliding of the video goggles when rolling, and the error of the doctor's judgment, it is finally verified through preliminary experiments that it is reasonable to use "the difference in nystagmus latency is greater than or equal to 1s" as the judgment criterion.

[0104] With this method, the LON of the affected side is more than 1s longer than that of the healthy side, and the correct rate of judging the affected side is 67%. The difference in LON between the two sides is significant, with little interference and strong practicality. Many patients without spontaneous nystagmus, and RT and BST cannot determine the affected side, can all be judged according to LON to determine the affected side for successful reposition. Most of the cases where RT cannot determine the affected side or make an incorrect judgment are due to the presence of physiological stenosis or structural abnormalities causing otolith incarceration. When the incarceration point is between 2 and 3, the otolith flows from 2 to 3 when turning to the affected side during RT, and the otolith flow is blocked. The time for the ampulla d crest to feel stimulation is prolonged, that is, the LON is prolonged, and the amount of otolith particle stimulation per unit time is reduced, that is, the intensity of nystagmus is weakened (the intensity of nystagmus on the affected side is weakened). When turning to the healthy side, the otolith moves from 3 to 2, and the LON and nystagmus intensity are not significantly affected. For patients whose RT cannot determine the affected side or makes an incorrect judgment due to otolith incarceration, determining the affected side according to LON highlights the advantage.

Claims

1. A method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo, characterized in that, It includes the following steps: S1: Create a diseased side judgment model: Create a diseased side judgment model for judging the diseased side through the preprocessed nystagmus data; S2: Experimentally collect nystagmus data: Conduct an experiment on patients with horizontal canalithiasis benign paroxysmal positional vertigo to trigger nystagmus, and use a collection device to record the nystagmus data generated by the patients and save it in a local storage device; S3: Preprocess the nystagmus data: Read the nystagmus data of the patients saved in the local storage device, and preprocess the nystagmus data to obtain the preprocessed nystagmus data; S4: Conduct diseased side judgment: Import the preprocessed nystagmus data into the diseased side judgment model to conduct diseased side judgment.

2. The method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo according to claim 1, wherein: The specific steps for creating the diseased side judgment model described in step S1 specifically include: S1-1: Construct a strength ratio judgment module, calculate the nystagmus strength ratio, and set the threshold trigger condition as the strength ratio ≥ 1.5; S1-2: Establish a BST direction comparison matrix, set the association between the body position relationship and the expected direction relationship, and set a direction conflict detection algorithm; S1-3: Set the nystagmus latency calculation formula and set a dynamic threshold for determining the diseased side; S1-4: Fuse multi-modal features of time domain features, frequency domain features, and spatial features to construct a feature cross layer; S1-5: Adopt a rule-neural network hybrid architecture, use the Drools rule engine to process explicit threshold judgments, design a 3-layer LSTM network to process temporal features, and construct an attention mechanism to process spatial features; S1-6: Collect clinically diagnosed samples as the training data set and conduct segmented training on the overall model.

3. The method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo according to claim 1, characterized in that: The experiment for triggering nystagmus in patients described in step S2 specifically includes the following steps: S2-1: Conduct a flexion-supine test on the patient and collect the nystagmus direction of the patient; S2-2: After the nystagmus completely disappears, conduct a 180° roll test on the patient and collect the nystagmus intensity and nystagmus latency of the patient.

4. The method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo according to claim 3, wherein: The flexion-supine test described in step S2-1 specifically includes the following steps: S2-1-1: Conduct a neck flexion test on the patient, make the patient sit on the examination bed, and quickly change the head from the upright position to the flexed position of 90°; S2-1-2: After the nystagmus completely disappears, conduct a sitting-supine test on the patient, and make the patient change from the sitting position to lying supine on a 30° inclined pillow.

5. The method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo according to claim 3, characterized in that: The 180° roll test described in step S2-2 specifically includes the following steps: S2-2-1: Make the patient quickly turn over or turn the head 90° to one side from the supine position at the end of the flexion-supine test, and use the flipped position as the starting side; S2-2-2: After the nystagmus completely disappears, make the patient rotate 180° from the starting side through the supine midline position to the opposite side, and collect and record the nystagmus latency, direction, intensity, and duration; S2-2-3: After the nystagmus completely disappears, make the patient rotate 180° again through the supine midline position and return to the starting side, and collect and record the nystagmus latency, direction, intensity, and duration; S2-2-4: If the nystagmus is not obvious or the collection is incomplete, repeat steps S2-2-2 to S2-2-3 to confirm that the nystagmus data collection is complete and correct.

6. The method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo according to claim 1, characterized in that: The preprocessing of the nystagmus data described in step S3 specifically includes the following steps: S3-1: Denoise the nystagmus signal, apply independent component analysis to separate environmental light interference and blink artifacts, and remove baseline drift through wavelet transform; S3-2: Extract parameters from the time domain, frequency domain, and direction features of nystagmus. Calculate the time difference from the body position change to the first occurrence of nystagmus as the nystagmus latency, and calculate the time span from the start to the complete disappearance of nystagmus as the nystagmus duration. Use the sliding window method to calculate the average velocity, extract the dominant frequency component through FFT, establish a three-dimensional coordinate system, and use the Hough transform to identify the dominant plane of nystagmus; S3-3: Align the nystagmus signal with the video recording time axis, and mark the key body position time points at the moment when the body position starts to change, the moment when the body position is stable, and the start and end moments of nystagmus; S3-4: Segment the data into the flexion-supine stage and the roll test stage according to the body position change signal; S3-5: Verify the nystagmus data to ensure that each test stage contains valid nystagmus cycles.

7. The method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo according to claim 1, wherein: The determination of the affected side using the affected side determination model described in step S4 specifically includes the following steps: S4-1: If the ratio of the nystagmus intensities on both sides collected in the roll test is greater than or equal to 1.5, then use the side with stronger nystagmus in the nystagmus data collected in the roll test as the affected side; S4-2: If the ratio of the nystagmus intensities on both sides collected in the roll test is less than 1.5, then make a judgment based on the nystagmus data collected in the flexion-supine test. The direction of the affected side is the same as the direction of the nystagmus in the head-down position and opposite to the direction of the nystagmus in the supine position in the nystagmus data collected in the flexion-supine test; S4-3: If the direction of the nystagmus in the head-down position and the direction of the nystagmus in the supine position in the nystagmus data collected in the flexion-supine test are the same, or no nystagmus is elicited, and the affected side cannot be determined, then make a judgment based on the nystagmus latency collected in the roll test. When the difference in nystagmus latency is greater than or equal to 1 s, use the side with the longer nystagmus latency as the affected side.

8. A system for determining the affected side of horizontal canal benign paroxysmal positional vertigo, characterized in that, Including: A data preprocessing module for performing relevant preprocessing on the nystagmus data to obtain preprocessed nystagmus data; An image processing module for constructing a time-slow phase velocity (SPV) graph according to the preprocessed nystagmus data; A feature extraction module for feature extraction, effectively screening the time-slow phase velocity (SPV) graph, and obtaining the nystagmus latency, intensity, direction, and duration; An intelligent data analysis platform equipped with an affected side determination model to determine the affected side of geotropic horizontal canal benign paroxysmal positional vertigo.

9. A computer device, characterized in that, Including a processor and a memory, the processor is used to execute the computer program stored in the memory to implement the method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo as described in claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, it implements the method for determining the affected side of geotropic horizontal canal benign paroxysmal positional vertigo as described in claims 1-7.

Images