Vagina detection and evaluation system and method based on tactile imaging

Abstract

The present invention belongs to the technical field of vaginal detection, and discloses a vaginal detection and evaluation system and method based on tactile imaging; the system and method include: collecting pressure data of the inside of the vagina of a target patient based on a flexible sensor to obtain a corresponding pressure sensing signal; performing signal processing on the collected pressure sensing signal, and constructing a corresponding pressure distribution image based on the pressure sensing signal after signal processing; performing image processing on the obtained pressure distribution image to obtain a corresponding vaginal tactile image; evaluating the abnormality of the vagina of the corresponding patient based on the obtained vaginal tactile image, obtaining a corresponding abnormality evaluation report, and providing feedback; the present invention can improve the accuracy, efficiency and clinical application value of vaginal detection from many aspects, and provide strong support for medical work.

Description

Technical Field

[0001] The present invention relates to the technical field of vagina detection, and more specifically, to a vagina detection and evaluation system and method based on tactile imaging. Background Art

[0002] Traditional vaginal examination methods, such as digital examination by gynecologists, mainly rely on the doctor's experience and subjective judgment, and it is difficult to accurately quantify the pressure distribution and subtle structural changes inside the vagina. For example, some early vaginal muscle dysfunction or minor lesions can be easily overlooked through digital examination, resulting in untimely disease detection and delay in the best treatment time. Some existing imaging detection methods, such as ultrasound examination, have limited resolution for soft tissue inside the vagina and cannot clearly present the pressure changes of the vaginal wall, making it difficult for doctors to obtain comprehensive and accurate information for diagnosis. Due to the limitations of existing detection technologies, early diagnosis and intervention of vaginal diseases in clinical practice face many challenges. Many vaginal diseases do not have obvious symptoms in the early stage and are difficult to detect in time through existing detection methods. When the disease develops to the point where obvious symptoms appear, it is often already quite serious, which increases the difficulty and cost of treatment. At the same time, due to the lack of an effective detection and evaluation system, doctors lack sufficient data support when formulating treatment plans, making it difficult to achieve accurate treatment, affecting the patient's rehabilitation effect and quality of life.

[0003] In view of this, the present invention proposes a vagina detection and evaluation system and method based on tactile imaging to solve the above problems. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions:

[0005] Vaginal inspection and assessment system based on tactile imaging, including:

[0006] A data acquisition module collects pressure data from the vagina of a target patient based on a flexible sensor to obtain a corresponding pressure sensing signal;

[0007] A data processing module, used for processing the collected pressure sensing signals and constructing a corresponding pressure distribution image based on the pressure sensing signals after signal processing;

[0008] An image processing module, used for performing image processing on the obtained pressure distribution image to obtain a corresponding vaginal tactile image;

[0009] The abnormality detection module evaluates the abnormality degree of the corresponding patient's vagina based on the obtained vaginal tactile image, obtains a corresponding abnormality evaluation report, and provides feedback.

[0010] Furthermore, the process of collecting pressure data inside the vagina of the target patient based on the flexible sensor and obtaining the corresponding pressure sensing signal includes:

[0011] The data acquisition module is provided with a flexible probe composed of a plurality of pressure sensor arrays; the flexible probe is input into the vagina of the target patient, and the pressure data of different positions inside the vagina of the target patient are collected by changing the insertion depth of the corresponding flexible probe, and the corresponding pressure sensing signal is obtained and output.

[0012] Furthermore, the process of performing signal processing on the collected pressure sensing signal and constructing a corresponding pressure distribution image based on the pressure sensing signal after signal processing includes:

[0013] Performing frame rate division on the obtained pressure sensing signal, obtaining the corresponding single needle sensing image and normalizing it;

[0014] Adaptively segment the normalized single-frame sensor signal to obtain a number of signal sub-bands and invalid sub-bands;

[0015] Obtaining the frequency band spectrum energy corresponding to each signal sub-band; obtaining the signal spectrum entropy corresponding to the corresponding single-frame sensor signal based on the frequency band spectrum energy corresponding to each signal sub-band;

[0016] A spectral entropy threshold is set, and the obtained signal spectral entropy is compared with the corresponding spectral entropy threshold. If the signal spectral entropy is less than the spectral entropy threshold, the corresponding single-frame sensor signal is marked as the data start end; if the signal spectral entropy is not less than the spectral entropy threshold, the corresponding single-frame sensor signal is marked as the data end end;

[0017] Based on the obtained data start end and data end end, the corresponding pressure sensing signal is divided to obtain a corresponding data signal and a noise signal;

[0018] Performing improved wavelet transform on the obtained data signal to obtain corresponding wavelet coefficients, and performing adaptive threshold processing on the obtained wavelet coefficients;

[0019] The noise signal is zeroed, and wavelet reconstruction is performed based on the noise signal after zeroing and the data signal after adaptive threshold processing to obtain a corresponding enhanced signal sub-band; and the signal is recombined with the invalid sub-band to obtain a corresponding enhanced sensing signal; based on a bilinear interpolation algorithm, the discrete enhanced sensing signal is converted into a grayscale image to obtain a corresponding pressure distribution image.

[0020] Furthermore, the process of performing adaptive segmentation includes:

[0021] Defining frequency band energy parameters ; In the formula, Indicates The first The frequency band energy corresponding to each sub-band; Indicates The signal power energy corresponding to a single frame of sensing signal; Indicates the total number of divided sub-bands; ; Represents the index of the sub-band in a single frame of sensor signal;

[0022] Get the minimum frequency band energy parameter corresponding to the corresponding single-frame sensor signal , and obtain the number of sub-bands corresponding to the corresponding minimum frequency band energy parameter;

[0023] The frequency band score corresponding to the corresponding sub-frequency band is obtained. If the frequency band score is not less than a preset score threshold, the corresponding sub-frequency band is marked as a signal sub-frequency band. If the frequency band score is less than the score threshold, the corresponding sub-frequency band is marked as an invalid sub-frequency band.

[0024] Furthermore, the signal spectral entropy ; In the formula, Indicates The frequency band logarithmic energy corresponding to a single frame sensor signal; ; represents the regulating factor; Indicates The first The band spectrum energy of the signal sub-band;

[0025] ; Indicates The frequency band spectral entropy of a single frame sensor signal; and represent the average value of the frequency band energy and the average value of the frequency band spectral entropy of the invalid channel sub-band respectively; Indicates The first The band spectrum energy of the signal sub-band;

[0026] The wavelet coefficients ; In the formula, denote the scale and time factors respectively; represents the wavelet basis function, ; Indicates the center frequency of the window during wavelet transform; represents the adjustment factor; represents an imaginary unit; Indicates data signal, is the time index within the data signal.

[0027] Further, the grayscale value corresponding to each pixel point in the corresponding pressure distribution map is obtained, and a corresponding grayscale histogram is constructed;

[0028] Based on the grayscale histogram, the pixel point corresponding to the grayscale value peak is obtained and set as the initial cluster point , , and is an integer;

[0029] Obtain the membership coefficients between each pixel point and the corresponding initial cluster point in the corresponding pressure distribution image respectively; perform multi-scale Gaussian filtering on the obtained pressure distribution image to obtain the corresponding scale space image ;

[0030] Define the clustering objective function In the formula, Represents the initial cluster point Pixels in the neighborhood of With pixels The spatial similarity coefficient between Represents the initial cluster point The neighborhood interval of ; and Represents pixel points and the initial cluster points Gray value of represents the Gaussian kernel function; represents fuzzy weight; Represents pixel and the initial cluster points The membership coefficient between ;

[0031] Based on the maximum membership principle, the independent variable parameters of the corresponding clustering objective function are Performing an optimal solution; and replacing the original fuzzy weights based on the fuzzy weights in the optimal solution result;

[0032] Based on the fuzzy weight after replacement, the membership coefficient Make corrections to obtain new membership coefficients ; In the formula, Represents pixel The neighborhood pixel set of Represents the index of the pixel in the neighborhood pixel set; Represents pixel and the neighborhood pixel set The membership coefficient between pixels; represents the fuzzy weight after replacement; Used to represent the total number of pixels in the neighborhood pixel set;

[0033] Repeat the above membership coefficient updating process until it meets the iteration stop condition, then obtain the membership coefficient between each pixel point and the corresponding initial cluster point, and divide the corresponding pixel point into the area where the initial cluster point corresponding to the maximum membership coefficient is located, obtain the corresponding pixel division interval, and construct the corresponding vaginal tactile image based on the obtained pixel division interval.

[0034] Furthermore, the process of evaluating the abnormality of the corresponding patient's vagina based on the obtained vaginal tactile image, obtaining a corresponding abnormality evaluation report, and providing feedback includes:

[0035] The obtained vaginal tactile images are sequentially input into a pre-constructed vaginal anatomical model to obtain corresponding vaginal identification information;

[0036] Perform three-dimensional reconstruction based on the obtained vaginal identification information to obtain a corresponding three-dimensional structural model; compare the obtained three-dimensional reconstruction model with a pre-established normal vaginal shape reference model to obtain a corresponding comparison result;

[0037] Based on the comparison results, the difference between the corresponding three-dimensional reconstructed model and each anatomical structure in the corresponding normal vaginal reference model is obtained, and an abnormality score is assigned to the corresponding three-dimensional reconstructed model based on the difference to obtain a corresponding abnormality score; the abnormality score corresponding to each vaginal tactile image is obtained, and a weighted sum is performed to obtain an overall abnormality score corresponding to the corresponding patient's vagina;

[0038] Generate a corresponding abnormality assessment result based on the obtained overall abnormality score and the abnormality scores corresponding to each vaginal tactile image; and generate a corresponding feedback suggestion based on the abnormality assessment result.

[0039] Furthermore, the construction process of the vaginal anatomical model includes:

[0040] The backbone network of the vaginal anatomical model is an improved ResNet network, and the basic framework of the improved ResNet network includes an input layer, a feature classification layer, and an output layer;

[0041] The input layer is used to receive the vaginal tactile images in the input training samples and perform image preprocessing on them to obtain the corresponding input images, wherein the image preprocessing includes normalization processing and image scaling and other processes;

[0042] The feature classification layer is used to receive the input image and perform hole convolution processing on it to obtain the corresponding high-dimensional feature vector and low-dimensional feature vector ; Perform global average pooling and maximum pooling operations on the corresponding high-dimensional feature vector to obtain a corresponding feature mapping vector, and perform a secondary global average pooling operation on the obtained feature mapping vector to obtain two dimensional feature vectors of different dimensions;

[0043] Perform image processing on the feature image of the corresponding dimension based on a single-layer perception layer to obtain a corresponding feature mapping vector;

[0044] Performing image fusion and normalization processing on the obtained perception feature vector, feature mapping vector and high-dimensional mapping vector to obtain a corresponding fused feature vector;

[0045] The output layer is used to map the obtained fusion feature vector into a preset parameter space to obtain a corresponding model output result;

[0046] Construct a training data set; define AdaGrad as the optimizer to continuously optimize the parameters of the vaginal anatomical model during the training process, input the constructed training data set into the vaginal anatomical model in batches, record the value of the corresponding loss function, and when the value of the loss function of consecutive L batches no longer decreases or changes, save the parameters of the vaginal anatomical model at this time, and complete the training of the vaginal anatomical model.

[0047] Furthermore, the corresponding feature mapping vector is obtained:

[0048] ; In the formula, represents the feature map vector; and Represent average pooling and maximum pooling operations respectively; represents the outer product operation; represents the least squares operation; represents the adjustment factor;

[0049] Perceptual feature vector ; and Represents the dimensional feature vectors of two different dimensions corresponding to the corresponding feature mapping vectors; Represents the forward propagation operation within a single perception layer; Vector fusion operation representing dimensional feature vectors;

[0050] Defining the loss function for the vaginal anatomy model ; In the formula, , and Respectively represent the total number and index of training samples; Indicates the actual label corresponding to the training sample; Indicates that the prediction result of the corresponding training sample belongs to the category label The probability distribution of Represents the class labels of known anatomical structures.

[0051] Tactile imaging-based vaginal inspection and assessment methods, including:

[0052] Step 1: Collect pressure data inside the vagina of the target patient based on the flexible sensor to obtain a corresponding pressure sensing signal;

[0053] Step 2: performing signal processing on the collected pressure sensing signal, and constructing a corresponding pressure distribution image based on the pressure sensing signal after signal processing;

[0054] Step 3: performing image processing on the obtained pressure distribution image to obtain a corresponding vaginal tactile image;

[0055] Step 4: Based on the obtained vaginal tactile images, the abnormality degree of the corresponding patient's vagina is evaluated, a corresponding abnormality evaluation report is obtained, and feedback is provided.

[0056] The technical effects and advantages of the vagina detection and evaluation system and method based on tactile imaging of the present invention are as follows:

[0057] 1. The flexible sensor collects the internal vaginal pressure data, and can accurately obtain the pressure information at different positions; in data processing, the adaptive segmentation, spectral entropy analysis and other methods can be used to effectively distinguish the data signal from the noise signal, enhance the effective signal, and make the constructed pressure distribution image more accurately reflect the internal vaginal pressure condition; accurately extract the vaginal anatomical structure characteristics, and generate a clear vaginal tactile image. In terms of abnormality detection, the vaginal tactile image is input into the pre-constructed anatomical model for three-dimensional reconstruction, and compared with the normal reference model, the degree of vaginal abnormality can be accurately assessed, and the accuracy of diagnosis can be improved.

[0058] 2. By combining the three-dimensional reconstructed model, doctors can observe the overall contour, curvature and relative position of the various parts of the vagina from multiple angles, and obtain comprehensive and intuitive information, which is convenient for more accurate judgment of the health status of the vagina and helps doctors develop targeted treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A schematic diagram of a vagina detection and evaluation system based on tactile imaging according to the present invention;

[0060] Figure 2 It is a schematic diagram of the vagina detection and evaluation method based on tactile imaging of the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Example 1

[0063] See also Figure 1 As shown, the vagina detection and evaluation system based on tactile imaging described in this embodiment includes:

[0064] A data acquisition module collects pressure data from the vagina of a target patient based on a flexible sensor to obtain a corresponding pressure sensing signal;

[0065] A data processing module, used for processing the collected pressure sensing signals and constructing a corresponding pressure distribution image based on the pressure sensing signals after signal processing;

[0066] An image processing module, used for performing image processing on the obtained pressure distribution image to obtain a corresponding vaginal tactile image;

[0067] The abnormality detection module evaluates the abnormality degree of the corresponding patient's vagina based on the obtained vaginal tactile image, obtains a corresponding abnormality evaluation report, and provides feedback;

[0068] The modules are connected to each other via wired and / or wireless means to achieve data transmission between modules.

[0069] It should be further explained that, in the specific implementation process, the process of collecting pressure data inside the vagina of the target patient based on the flexible sensor and obtaining the corresponding pressure sensing signal includes:

[0070] The data acquisition module is provided with a flexible detector composed of a plurality of pressure sensor arrays, which is used to collect pressure data inside the patient's vagina;

[0071] The flexible probe is inserted into the vagina of a target patient, and pressure data at different positions inside the vagina of the target patient are collected by changing the insertion depth of the corresponding flexible probe to obtain and output corresponding pressure sensing signals.

[0072] It should be further explained that, in a specific implementation process, the process of performing signal processing on the collected pressure sensing signal and constructing a corresponding pressure distribution image based on the pressure sensing signal after signal processing includes:

[0073] Adaptively segment the normalized single-frame sensor signal to obtain a number of signal sub-bands and invalid sub-bands;

[0074] The process of performing adaptive segmentation includes:

[0075] Defining frequency band energy parameters ; In the formula, Indicates The first The frequency band energy corresponding to each sub-band; Indicates The signal power energy corresponding to a single frame of sensing signal; Indicates the total number of divided sub-bands; ; Represents the index of the sub-band in a single frame of sensor signal;

[0076] Then, the minimum frequency band energy parameter corresponding to the corresponding single-frame sensor signal is obtained. , and obtain the number of sub-bands corresponding to the corresponding minimum frequency band energy parameter;

[0077] Then, the frequency band score corresponding to the corresponding sub-frequency band is obtained. If the frequency band score is not less than a preset score threshold, the corresponding sub-frequency band is marked as a signal sub-frequency band. If the frequency band score is less than the score threshold, the corresponding sub-frequency band is marked as an invalid sub-frequency band. The scoring process of the corresponding sub-frequency band needs to be determined according to the degree of noise interference in the corresponding single-frame sensor signal.

[0078] Get the frequency band spectrum power energy corresponding to each signal sub-band ; In the formula, represents the intermediate variable, ; is an integer and ; Indicates The total number of signal sub-bands corresponding to a single-frame sensor signal; Indicates A signal amplitude spectrum of a single frame sensor signal; An index representing a signal subband; a signal bandwidth representing a signal sub-band;

[0079] Then, based on the frequency band spectrum energy corresponding to each signal sub-band, the signal spectrum entropy corresponding to the corresponding single-frame sensor signal is obtained. ; In the formula, Indicates The frequency band logarithmic energy corresponding to a single frame sensor signal; ; represents the regulating factor; ; and Respectively represent the average value of the frequency band spectrum energy and the average value of the frequency band spectrum entropy of the invalid channel sub-band; Indicates The frequency band spectral entropy of a single frame sensor signal; Indicates The first The band spectrum energy of the signal sub-band;

[0080] A spectral entropy threshold is set, and the obtained signal spectral entropy is compared with the corresponding spectral entropy threshold. If the signal spectral entropy is less than the spectral entropy threshold, the corresponding single-frame sensor signal is marked as the data start end; if the signal spectral entropy is not less than the spectral entropy threshold, the corresponding single-frame sensor signal is marked as the data end end;

[0081] Furthermore, based on the obtained data start end and data end end, the corresponding pressure sensing signal is divided into signals to obtain corresponding data signals and noise signals, for example: the signal segments within the data start end and the data end end are processed as data signals, and the signal segments outside the data start end and the data end end are processed as noise signals; wherein the data signal refers to a signal segment containing pressure data and noise, and the noise signal refers to a pure noise signal segment;

[0082] Then, the obtained data signal is subjected to improved wavelet transform to obtain the corresponding wavelet coefficients ; In the formula, denote the scale and time factors respectively; represents the wavelet basis function, ; represents the center frequency; represents the adjustment factor; represents an imaginary unit; Indicates data signal, is the time index within the data signal;

[0083] Performing adaptive threshold processing on the obtained wavelet coefficients;

[0084] The formula for adaptive threshold processing is:

[0085] ; represents symbolic function operation, represents the wavelet threshold, where ; In the formula, Indicates the signal length of the data signal, represents the noise standard deviation; represents the wavelet coefficients after threshold processing; Indicates the number of decomposition layers in the wavelet transform process;

[0086] Furthermore, the noise signal is returned to zero, and wavelet reconstruction is performed based on the noise signal after the zeroing process and the data signal after the adaptive threshold process to obtain the corresponding enhanced signal sub-band; and the signal is recombined with the invalid sub-band to obtain the corresponding enhanced sensing signal; and then, the discrete enhanced sensing signal is converted into a grayscale image based on a bilinear interpolation algorithm to obtain a corresponding pressure distribution image GF, wherein the grayscale value of the pixel point in the pressure distribution image represents the pressure data size.

[0087] It should be further explained that, in the specific implementation process, the process of performing image processing on the obtained pressure distribution image to obtain the corresponding vaginal tactile image includes:

[0088] Obtain the grayscale value corresponding to each pixel in the corresponding pressure distribution map, and construct a corresponding grayscale histogram;

[0089] Based on the grayscale histogram, the pixel point corresponding to the grayscale value peak is obtained and set as the initial cluster point , , and is an integer;

[0090] Obtain the membership coefficients between each pixel point in the corresponding pressure distribution image and the corresponding initial cluster point respectively In the formula, represents the fuzzy weight, which is determined by the number of peak gray values; and Respectively represent the indexes of the initial cluster point and other pixel points except the initial cluster point; ; In the formula, and Represents pixel points and the initial cluster points The gray value of the corresponding pixel; represents the standard deviation of Gaussian noise;

[0091] The obtained pressure distribution image is processed by multi-scale Gaussian filtering to obtain the corresponding scale space image ; In the formula, and Respectively expressed in scale and The scale space image output after Gaussian filtering; represents the pressure distribution diagram;

[0092] Define the clustering objective function In the formula, Represents the initial cluster point Pixels in the neighborhood of With pixels The spatial similarity coefficient between Represents the initial cluster point The neighborhood interval of ; represents the Gaussian kernel function;

[0093] Among them, the spatial similarity coefficient ; In the formula; and Represents pixel points and pixels With the corresponding initial cluster point The membership coefficient between ; Represents pixel and pixels The Euclidean distance between ; is the evaluation coefficient, which is used to measure the pixel points in the neighborhood With the initial cluster point The reliability between them is higher, indicating that the corresponding pixel With the initial cluster point The deeper the mutual influence between them; ; represents the coefficient of variation, and They represent the preset maximum coefficient of variation and minimum coefficient of variation respectively; Represents pixel The total number of neighborhood pixels; Represents the initial cluster point The average grayscale value of all pixels in the neighborhood interval of the neighborhood pixel; and Represents pixel points and the initial cluster points Gray value of ; and is an integer; and Represent the initial cluster points The total number of pixel points indexed by the pixel points in the neighborhood interval;

[0094] Furthermore, based on the maximum membership principle, the independent variable parameters of the corresponding clustering objective function are Find the optimal solution;

[0095] Then, the original fuzzy weights are adjusted based on the fuzzy weights in the optimal solution. Make a replacement;

[0096] Then, based on the replaced fuzzy weights, the corresponding membership coefficients Make corrections to obtain new membership coefficients ; In the formula, Represents pixel The neighborhood pixel set of ; Represents the index of the neighborhood pixel; Represents pixel and neighborhood pixels The membership coefficient between ; Used to represent the total number of pixels in the neighborhood pixel set; represents the fuzzy weight after replacement;

[0097] Repeat the updating process of the membership coefficients until the iteration stop condition is met, then obtain the membership coefficients between each pixel point and the corresponding initial cluster point, and divide the corresponding pixel point into the area where the initial cluster point corresponding to the maximum membership coefficient is located, and obtain the corresponding pixel division interval; obtain the grayscale standard deviation corresponding to the pixel points in each pixel division interval, and compare it with the preset standard deviation threshold. If the grayscale standard deviation is less than the standard deviation threshold, then based on the corresponding pixel division interval, perform image interception on the corresponding pressure distribution image to obtain the corresponding pressure area image; if the grayscale standard deviation is not less than the standard deviation threshold, then based on the above pixel division interval acquisition process, perform secondary division to obtain a new pixel division interval, and so on;

[0098] All the pressure area images obtained are counted, and the overlap between different pressure area images is obtained based on the generalized union-intersection ratio algorithm. If the overlap is less than the preset overlap threshold, no other operation is performed; if the overlap is not less than the preset overlap threshold, indicating that the corresponding pressure area images belong to the same anatomical structure area, the overlapping parts are combined to obtain the corresponding merged area image;

[0099] Repeat the above-mentioned process of acquiring the merged area image to acquire several vaginal tactile images that meet the conditions, wherein the overlap between the vaginal tactile image and other vaginal tactile images is less than a preset overlap threshold, and the grayscale standard deviation of the corresponding vaginal tactile image is less than the standard deviation threshold.

[0100] It should be further explained that, in the specific implementation process, the process of evaluating the abnormality of the corresponding patient's vagina based on the obtained vaginal tactile image, obtaining the corresponding abnormality evaluation report, and providing feedback includes:

[0101] The obtained vaginal tactile images are sequentially input into a pre-constructed vaginal anatomical model to obtain corresponding vaginal identification information; the vaginal identification information refers to parameters such as the anatomical structure type, structure size, and position corresponding to the corresponding vaginal tactile images;

[0102] Then, based on the obtained vaginal identification information, a three-dimensional reconstruction is performed to obtain a corresponding three-dimensional structural model, including its overall contour, curvature, and relative positional relationship of each part;

[0103] Then, the obtained three-dimensional reconstructed model is compared with a pre-established normal vaginal shape reference model to obtain a corresponding comparison result; wherein the normal reference model may be an average value model or a typical model obtained by statistically analyzing the vaginal morphology of a large number of healthy women;

[0104] Based on the comparison result, the difference between the corresponding three-dimensional reconstructed model and each anatomical structure in the corresponding normal vaginal reference model is obtained, and an abnormality score is assigned to the corresponding three-dimensional reconstructed model based on the difference, so as to obtain a corresponding abnormality score; wherein the higher the difference, the higher the corresponding abnormality score;

[0105] Then, the abnormality scores corresponding to each vaginal tactile image are obtained, and weighted summation is performed to obtain the overall abnormality score corresponding to the vagina of the corresponding patient; wherein the weight coefficient of the anatomical structure corresponding to each vaginal tactile image is determined by the specific vaginal disease type corresponding to the corresponding patient;

[0106] Furthermore, a corresponding abnormality assessment result is generated based on the obtained overall abnormality score and the abnormality scores corresponding to each vaginal tactile image; and a corresponding feedback suggestion is generated based on the obtained abnormality assessment result; for example, if the overall abnormality score of the patient's vagina or the abnormality score of a certain anatomical structure is higher than a preset score threshold, medical advice such as immediate medical treatment is fed back to the corresponding patient;

[0107] The construction process of the vaginal anatomical model includes:

[0108] The backbone network of the vaginal anatomical model is an improved ResNet network, and the basic framework of the improved ResNet network includes an input layer, a feature classification layer, and an output layer;

[0109] The input layer is used to receive the vaginal tactile images in the input training samples and perform image preprocessing on them to obtain the corresponding input images, wherein the image preprocessing includes normalization processing and image scaling and other processes;

[0110] The feature classification layer is used to receive the input image and perform hole convolution processing on it to obtain the corresponding high-dimensional feature vector and low-dimensional feature vector ; And perform global average pooling and maximum pooling operations on the corresponding high-dimensional feature vector to obtain the corresponding feature mapping vector:

[0111] ; In the formula, represents the feature map vector; and Represent average pooling and maximum pooling operations respectively; represents the outer product operation; represents the least squares operation; Represents the adjustment coefficient, which is a fixed constant;

[0112] And perform a secondary global average pooling operation on the obtained feature map vector to obtain two dimensional feature vectors of different dimensions;

[0113] Then, based on a single-layer perception layer, the corresponding dimension feature image is processed to obtain the corresponding perception feature vector ; and Represents the dimensional feature vectors of two different dimensions corresponding to the corresponding feature mapping vectors; Represents the forward propagation operation within a single perception layer; Vector fusion operation representing dimensional feature vectors;

[0114] Then, the obtained perceptual feature vector, feature mapping vector and high-dimensional mapping vector are subjected to image fusion and normalization processing to obtain a corresponding fused feature vector;

[0115] The output layer is used to map the obtained fusion feature vector into a preset parameter space to obtain a corresponding model output result (i.e., vagina recognition information); wherein the parameter space is obtained by learning the corresponding output layer during the training process based on the nonlinear mapping relationship between the training samples and the corresponding sample labels;

[0116] Defining the loss function for the vaginal anatomy model ; In the formula, , and Respectively represent the total number and index of training samples, both are integers; Indicates the actual label corresponding to the training sample; Indicates that the prediction result of the corresponding training sample belongs to the category label The probability distribution of Category labels representing known anatomical structures;

[0117] Acquire several tactile images corresponding to different vaginal anatomical structures respectively, and manually annotate them to obtain corresponding category labels, wherein the category labels include information such as the category, structure size, shape, etc. of the corresponding vaginal anatomical structure; and construct a corresponding training data set based on the manually annotated tactile images;

[0118] Define AdaGrad as the optimizer to continuously optimize the parameters of the vaginal anatomical model during the training process, input the constructed training data set into the vaginal anatomical model in batches, and record the value of the corresponding loss function. When the value of the loss function of consecutive L batches no longer decreases or changes, save the parameters of the vaginal anatomical model at this time, and the training of the vaginal anatomical model is completed.

[0119] The present invention realizes comprehensive detection and evaluation of the inside of the patient's vagina through modules such as data acquisition, processing, image generation and abnormality detection; at the same time, by constructing and training a vaginal anatomical model, the recognition precision and accuracy of the system are improved.

[0120] Example 2

[0121] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a vaginal detection and evaluation method based on tactile imaging is provided, including:

[0122] Step 1: Collect pressure data inside the vagina of the target patient based on the flexible sensor to obtain a corresponding pressure sensing signal;

[0123] Step 2: performing signal processing on the collected pressure sensing signal, and constructing a corresponding pressure distribution image based on the pressure sensing signal after signal processing;

[0124] Step 3: performing image processing on the obtained pressure distribution image to obtain a corresponding vaginal tactile image;

[0125] Step 4: Based on the obtained vaginal tactile images, the abnormality degree of the corresponding patient's vagina is evaluated, a corresponding abnormality evaluation report is obtained, and feedback is provided.

[0126] Example 3

[0127] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the vaginal detection and evaluation system and method based on tactile imaging provided above is implemented.

[0128] Since the electronic device introduced in this embodiment is an electronic device used to implement the vagina detection and evaluation system and method based on tactile imaging in the embodiments of this application, based on the vagina detection and evaluation system and method based on tactile imaging introduced in the embodiments of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiments of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used by the vagina detection and evaluation system and method based on tactile imaging in the embodiments of this application, they all fall within the scope of protection of this application.

[0129] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0130] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A vaginal detection and assessment system based on tactile imaging, characterized in that: include: A data acquisition module collects pressure data from the vagina of a target patient based on a flexible sensor to obtain a corresponding pressure sensing signal; A data processing module, used for processing the collected pressure sensing signals and constructing a corresponding pressure distribution image based on the pressure sensing signals after signal processing; An image processing module, used for performing image processing on the obtained pressure distribution image to obtain a corresponding vaginal tactile image; The abnormality detection module evaluates the abnormality degree of the corresponding patient's vagina based on the obtained vaginal tactile image, obtains a corresponding abnormality evaluation report, and provides feedback; The process of acquiring vaginal tactile images includes: Obtain the grayscale value corresponding to each pixel in the corresponding pressure distribution map, and construct a corresponding grayscale histogram; Based on the grayscale histogram, the pixel point corresponding to the grayscale value peak is obtained and set as the initial cluster point , , and is an integer; Obtain the membership coefficients between each pixel point and the corresponding initial cluster point in the corresponding pressure distribution image respectively; perform multi-scale Gaussian filtering on the obtained pressure distribution image to obtain the corresponding scale space image ; Define the clustering objective function In the formula, Represents the initial cluster point Pixels in the neighborhood of With pixels The spatial similarity coefficient between Represents the initial cluster point The neighborhood interval of ; and Represents pixel points and the initial cluster points Gray value of represents the Gaussian kernel function; represents fuzzy weight; Represents pixel and the initial cluster points The membership coefficient between ; Based on the maximum membership principle, the independent variable parameters of the corresponding clustering objective function are Performing an optimal solution; and replacing the original fuzzy weights based on the fuzzy weights in the optimal solution result; Based on the fuzzy weight after replacement, the membership coefficient Make corrections to obtain new membership coefficients ; In the formula, Represents pixel The neighborhood pixel set of ; Represents the index of the pixel in the neighborhood pixel set; Represents pixel and the neighborhood pixel set The membership coefficient between pixels; represents the fuzzy weight after replacement; Used to represent the total number of pixels in the neighborhood pixel set; Repeat the above membership coefficient updating process until it meets the iteration stop condition, then obtain the membership coefficient between each pixel point and the corresponding initial cluster point, and divide the corresponding pixel point into the area where the initial cluster point corresponding to the maximum membership coefficient is located, obtain the corresponding pixel division interval, and construct the corresponding vaginal tactile image based on the obtained pixel division interval.

2. The vaginal detection and evaluation system based on tactile imaging according to claim 1, characterized in that: The process of collecting pressure data inside the vagina of the target patient based on the flexible sensor and obtaining the corresponding pressure sensing signal includes: The data acquisition module is provided with a flexible probe composed of a plurality of pressure sensor arrays; the flexible probe is input into the vagina of the target patient, and the pressure data of different positions inside the vagina of the target patient are collected by changing the insertion depth of the corresponding flexible probe, and the corresponding pressure sensing signal is obtained and output.

3. The vaginal detection and evaluation system based on tactile imaging according to claim 2, characterized in that: The process of processing the collected pressure sensing signal and constructing a corresponding pressure distribution image based on the pressure sensing signal after the signal processing includes: Performing frame rate division on the obtained pressure sensing signal, obtaining the corresponding single needle sensing image and normalizing it; Adaptively segment the normalized single-frame sensor signal to obtain a number of signal sub-bands and invalid sub-bands; Obtaining the frequency band spectrum energy corresponding to each signal sub-band; obtaining the signal spectrum entropy corresponding to the corresponding single-frame sensor signal based on the frequency band spectrum energy corresponding to each signal sub-band; A spectral entropy threshold is set, and the obtained signal spectral entropy is compared with the corresponding spectral entropy threshold. If the signal spectral entropy is less than the spectral entropy threshold, the corresponding single-frame sensor signal is marked as the data start end; if the signal spectral entropy is not less than the spectral entropy threshold, the corresponding single-frame sensor signal is marked as the data end end; Based on the obtained data start end and data end end, the corresponding pressure sensing signal is divided to obtain a corresponding data signal and a noise signal; Performing improved wavelet transform on the obtained data signal to obtain corresponding wavelet coefficients, and performing adaptive threshold processing on the obtained wavelet coefficients; The noise signal is zeroed, and wavelet reconstruction is performed based on the noise signal after zeroing and the data signal after adaptive threshold processing to obtain a corresponding enhanced signal sub-band; and the signal is recombined with the invalid sub-band to obtain a corresponding enhanced sensing signal; based on a bilinear interpolation algorithm, the discrete enhanced sensing signal is converted into a grayscale image to obtain a corresponding pressure distribution image.

4. The vaginal detection and evaluation system based on tactile imaging according to claim 3, characterized in that: The process of performing adaptive segmentation includes: Defining frequency band energy parameters ; In the formula, Indicates The first The frequency band energy corresponding to each sub-band; Indicates The signal power energy corresponding to a single frame of sensing signal; Indicates the total number of divided sub-bands; ; Represents the index of the sub-band in a single frame of sensor signal; Get the minimum energy parameter corresponding to the corresponding single-frame sensor signal , and obtain the number of sub-bands corresponding to the corresponding minimum band energy parameter; The frequency band score corresponding to the corresponding sub-frequency band is obtained. If the frequency band score is not less than a preset score threshold, the corresponding sub-frequency band is marked as a signal sub-frequency band. If the frequency band score is less than the score threshold, the corresponding sub-frequency band is marked as an invalid sub-frequency band.

5. The vaginal detection and evaluation system based on tactile imaging according to claim 4, characterized in that: The signal spectral entropy ; In the formula, Indicates The frequency band logarithmic energy corresponding to a single frame sensor signal; ; represents the regulating factor; Indicates The first The band spectrum energy of the signal sub-band; ; Indicates The frequency band spectral entropy of a single frame sensor signal; and represent the average value of the frequency band energy and the average value of the frequency band spectral entropy of the invalid channel sub-band respectively; The wavelet coefficients ; In the formula, denote the scale and time factors respectively; represents the wavelet basis function, ; Indicates the center frequency of the window during wavelet transform; represents the adjustment factor; represents an imaginary unit; Represents the data signal, is the time index within the data signal.

6. The vagina detection and evaluation system based on tactile imaging according to claim 5, characterized in that: The process of evaluating the abnormality of the corresponding patient's vagina based on the obtained vaginal tactile image, obtaining a corresponding abnormality evaluation report, and providing feedback includes: The obtained vaginal tactile images are sequentially input into a pre-constructed vaginal anatomical model to obtain corresponding vaginal identification information; Perform three-dimensional reconstruction based on the obtained vaginal identification information to obtain a corresponding three-dimensional structural model; compare the obtained three-dimensional reconstruction model with a pre-established normal vaginal shape reference model to obtain a corresponding comparison result; Based on the comparison results, the difference between the corresponding three-dimensional reconstructed model and each anatomical structure in the corresponding normal vaginal reference model is obtained, and an abnormality score is assigned to the corresponding three-dimensional reconstructed model based on the difference to obtain a corresponding abnormality score; the abnormality score corresponding to each vaginal tactile image is obtained, and a weighted sum is performed to obtain an overall abnormality score corresponding to the corresponding patient's vagina; Generate a corresponding abnormality assessment result based on the obtained overall abnormality score and the abnormality scores corresponding to each vaginal tactile image; and generate a corresponding feedback suggestion based on the obtained abnormality assessment result.

7. The vagina detection and evaluation system based on tactile imaging according to claim 6, characterized in that: The construction process of the vaginal anatomy model includes: The backbone network of the vaginal anatomical model is an improved ResNet network, and the basic framework of the improved ResNet network includes an input layer, a feature classification layer, and an output layer; The input layer is used to receive the vaginal tactile images in the input training samples and perform image preprocessing on them to obtain corresponding input images; the image preprocessing includes normalization processing and image scaling; The feature classification layer is used to receive the input image and perform hole convolution processing on it to obtain the corresponding high-dimensional feature vector and low-dimensional feature vector ; Perform global average pooling and maximum pooling operations on the corresponding high-dimensional feature vector to obtain a corresponding feature mapping vector, and perform a secondary global average pooling operation on the obtained feature mapping vector to obtain two dimensional feature vectors of different dimensions; Perform image processing on the feature image of the corresponding dimension based on a single-layer perception layer to obtain a corresponding feature mapping vector; Performing image fusion and normalization processing on the obtained perception feature vector, feature mapping vector and high-dimensional mapping vector to obtain a corresponding fused feature vector; The output layer is used to map the obtained fusion feature vector into a preset parameter space to obtain a corresponding model output result; Construct a training data set; define AdaGrad as the optimizer to continuously optimize the parameters of the vaginal anatomical model during the training process, input the constructed training data set into the vaginal anatomical model in batches, record the value of the corresponding loss function, and when the value of the loss function of consecutive L batches no longer decreases or changes, save the parameters of the vaginal anatomical model at this time, and complete the training of the vaginal anatomical model.

8. The vagina detection and evaluation system based on tactile imaging according to claim 7, characterized in that: The formula for obtaining the corresponding feature mapping vector is: ; In the formula, represents the feature map vector; and Represent average pooling and maximum pooling operations respectively; represents the outer product operation; represents the least squares operation; represents the adjustment factor; Perceptual feature vector ; and Represents the dimensional feature vectors of two different dimensions corresponding to the corresponding feature mapping vectors; Represents the forward propagation operation within a single perception layer; Vector fusion operation representing dimensional feature vectors; Defining the loss function for the vaginal anatomy model ; In the formula, , and Respectively represent the total number and index of training samples; Indicates the actual label corresponding to the training sample; Indicates that the prediction result of the corresponding training sample belongs to the category label The probability distribution of Represents the class labels of known anatomical structures.

9. A vaginal detection and evaluation method based on tactile imaging, which is based on the vaginal detection and evaluation system based on tactile imaging according to any one of claims 1 to 8, characterized in that: include: Step 1: Collect pressure data inside the vagina of the target patient based on the flexible sensor to obtain a corresponding pressure sensing signal; Step 2: performing signal processing on the collected pressure sensing signal, and constructing a corresponding pressure distribution image based on the pressure sensing signal after signal processing; Step 3: performing image processing on the obtained pressure distribution image to obtain a corresponding vaginal tactile image; Step 4: Based on the obtained vaginal tactile images, the abnormality degree of the corresponding patient's vagina is evaluated, a corresponding abnormality evaluation report is obtained, and feedback is provided.

Images