Optical detection device for rapidly judging peritoneal dialysis outlet infection

Abstract

The invention discloses an optical detection device for rapidly judging peritoneal dialysis outlet infection, and relates to the technical field of medical instruments, and the device comprises a collected image set obtaining module which is used for calling a light source to carry out light emission according to a preset light type sequence and carrying out image collection of different spectrums by using a high-resolution camera and an optical filter wheel, acquiring a collection image set; the multi-scale feature set obtaining module is used for traversing the collected image set to carry out multi-scale feature collection and obtaining a collected image multi-scale feature group set; the interactive iteration feature set obtaining module is used for carrying out intra-group feature interactive iteration to obtain an interactive iteration feature set; and the mapping analysis result obtaining module is used for performing infection mapping analysis based on the interactive iteration feature set to obtain a target mapping analysis result. The technical problems of low infection image recognition accuracy and poor detection result reliability in the prior art are solved, and the technical effect of improving the optical detection result reliability is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and particularly to an optical detection device for quickly judging peritoneal dialysis exit infection. Background Art

[0002] With the development of medical technology, non-invasive methods based on optical detection have gradually attracted attention. For example, methods such as fluorescence imaging and multispectral analysis have been used in some medical detection fields, but there are still many challenges in their application in peritoneal dialysis infection detection.

[0003] Existing optical detection methods have insufficient ability to distinguish the spectral characteristics of infected tissues from healthy tissues, resulting in a high misjudgment rate and affecting the detection effect. Traditional methods rely on single-spectrum or shallow feature extraction and do not fully utilize multi-scale and cross-spectrum depth information, resulting in poor stability of detection results and possible deviations under different lighting conditions.

[0004] There are technical problems in the prior art such as low accuracy of infected image recognition and poor reliability of detection results. Summary of the Invention

[0005] The present invention provides an optical detection device for quickly judging peritoneal dialysis exit infection to solve the technical problems of low accuracy of infected image recognition and poor reliability of detection results in the prior art.

[0006] In a first aspect, the present invention provides an optical detection device for quickly judging peritoneal dialysis exit infection, wherein the optical detection device for quickly judging peritoneal dialysis exit infection includes:

[0007] An acquisition image set obtaining module, configured to call a light source to emit light according to a preset light type sequence, and use a high-resolution camera and a filter wheel to collect images of different spectra to obtain an acquisition image set;

[0008] A multi-scale feature set obtaining module, configured to traverse the acquisition image set to perform multi-scale feature acquisition to obtain a set of multi-scale feature groups of the acquisition images;

[0009] An interactive iterative feature set obtaining module, configured to perform intra-group feature interactive iteration on the set of multi-scale feature groups of the acquisition images respectively to obtain an interactive iterative feature set;

[0010] A mapping analysis result obtaining module, configured to perform infection mapping analysis based on the interactive iterative feature set to obtain a target mapping analysis result.

[0011] In a second aspect, the present invention also provides an optical detection method for quickly judging peritoneal dialysis exit infection, wherein the optical detection method for quickly judging peritoneal dialysis exit infection includes:

[0012] The light source is called to emit light according to a preset light type sequence, and a high-resolution camera and a filter wheel are used to collect images of different spectra to obtain a set of collected images;

[0013] Traverse the set of collected images to collect multi-scale features, and obtain a set of multi-scale feature groups of the collected images;

[0014] Perform intra-group feature interaction iteration on the set of multi-scale feature groups of the collected images respectively to obtain a set of interaction iteration features;

[0015] Based on the set of interaction iteration features, perform infection mapping analysis to obtain a target mapping analysis result.

[0016] The present invention discloses an optical detection device for quickly judging peritoneal dialysis exit infection, including: a collected image set obtaining module, which is used to call a light source to emit light according to a preset light type sequence, and use a high-resolution camera and a filter wheel to collect images of different spectra to obtain a set of collected images; a multi-scale feature set obtaining module, which is used to traverse the set of collected images to collect multi-scale features and obtain a set of multi-scale feature groups of the collected images; an interaction iteration feature set obtaining module, which is used to perform intra-group feature interaction iteration on the set of multi-scale feature groups of the collected images respectively to obtain a set of interaction iteration features; a mapping analysis result obtaining module, which is used to perform infection mapping analysis based on the set of interaction iteration features to obtain a target mapping analysis result. The technical effect of extracting feature information of different scales and improving the accuracy of optical detection is achieved. Description of the Drawings

[0017] Figure 1 It is a schematic structural diagram of the optical detection device for quickly judging peritoneal dialysis exit infection of the present invention;

[0018] Figure 2 It is a schematic flow chart of the optical detection method for quickly judging peritoneal dialysis exit infection of the present invention.

[0019] Description of the reference numerals: the collected image set obtaining module 11, the multi-scale feature set obtaining module 12, the interaction iteration feature set obtaining module 13, the mapping analysis result obtaining module 14. Detailed Embodiments

[0020] The above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and specific embodiments to better understand the above technical solutions. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used to explain the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention. In addition, it should be noted that for the convenience of description, only the parts related to the present invention rather than all are shown in the drawings.

[0021] Embodiment 1, Figure 1 is a schematic structural diagram of an optical detection device for quickly judging peritoneal dialysis exit site infection according to the present invention. Among them, the device includes:

[0022] An acquisition image set obtaining module 11, configured to call a light source to emit light according to a preset light type sequence, and use a high-resolution camera and a filter wheel to collect images of different spectra to obtain an acquisition image set;

[0023] Further, the light source is a multi-spectral LED array.

[0024] In a possible embodiment, the light source refers to a device that emits light and is used to excite specific spectral characteristics in tissues. Here, a multi-spectral LED array is used, which can emit light covering multiple spectral bands from visible light to near-infrared to obtain image data with different reflection or emission characteristics. The multi-spectral LED array is an array composed of multiple LED lights, and these LED lights can emit light of different wavelengths. Each LED emits light of a specific wavelength (such as red, green, blue, near-infrared, etc.). By exciting with light of multiple bands, more tissue information can be obtained, which is crucial for judging whether the peritoneal dialysis exit site is infected. The high-resolution camera can capture more details and thus can obtain clear tissue images. This camera is used to obtain the reflection or emission light information of the peritoneal dialysis exit area under different spectral irradiations. The filter wheel is a wheel installed in the optical system that can quickly switch filters of different wavelengths. By placing different filters in front of the camera, light of a specific wavelength can be selected to pass through the lens, filtering out unnecessary light and obtaining image data at the specified wavelength.

[0025] Through multispectral imaging technology, different spectral information of the peritoneal dialysis exit site is obtained, and this information will help further analyze whether the site is infected. First, a multispectral LED array emits light of different wavelengths to irradiate the peritoneal dialysis exit area. Then, a high-resolution camera captures the reflected or emitted light, and together with a filter wheel installed in front of the camera, images under different spectra are taken respectively. The images of each spectrum reflect different characteristics of the peritoneal dialysis exit area, which may be changes caused by the absorption and reflection of light of different wavelengths by tissues.

[0026] Through the cooperation of the light source, camera and filter, the optical detection device can capture more image data with multi-dimensional information, providing sufficient and diverse basic information for subsequent analysis.

[0027] Furthermore, the image set acquisition module 11 includes:

[0028] The brightness distribution consistency factor set acquisition unit is used to traverse the acquired image set for brightness distribution consistency recognition to obtain the brightness distribution consistency factor set;

[0029] The signal-to-noise ratio factor set acquisition unit is used to traverse the acquired image set for signal-to-noise ratio analysis to obtain the signal-to-noise ratio factor set;

[0030] The image quality factor set acquisition unit is used to perform weighted calculation on the brightness distribution consistency factor set and the signal-to-noise ratio factor set to obtain the image quality factor set;

[0031] The image re-acquisition instruction acquisition unit is used to determine whether the image quality factor set meets the preset image quality factor threshold. If not, an image re-acquisition instruction is obtained.

[0032] In a possible embodiment, the brightness distribution consistency factor set is a set of parameters obtained by analyzing the brightness distribution of the acquired images, and these parameters are used to measure the brightness consistency of the images under different lighting conditions. The signal-to-noise ratio (SNR) factor set is the ratio of the signal (useful information in the image) to the noise (such as noise points of the optical system and electronic sensor noise), and is used to measure the clarity and cleanliness of the image. A high signal-to-noise ratio means better image quality, while a low signal-to-noise ratio may lead to blurred details or misjudgment. The image quality factor set is an overall image quality score obtained by comprehensively weighting the brightness distribution consistency factor set and the signal-to-noise ratio factor set, and is used to measure whether the currently acquired images meet the requirements of subsequent analysis. The image re-acquisition instruction is used to let the optical detection device re-acquire images to ensure the accuracy and reliability of the final analysis.

[0033] Preferably, the acquired image set is grayscaled using the image grayscaling formula, where the image grayscaling formula is: I gray = 0.299R + 0.587G + 0.114B; where I gray is the grayscale value, R is the red channel pixel value, G is the green channel pixel value, and B is the blue channel pixel value. Furthermore, the acquired image set is divided according to the division scale preset by those skilled in the art, and then the mean standard deviation of the grayscale value is calculated for each local area. The mean standard deviation of the grayscale value of each acquired image is divided by the mean grayscale value of the acquired image to obtain the brightness distribution consistency factor set.

[0034] Preferably, the region of interest and the background region of each acquired image are determined, and the mean grayscale value of the region of interest is divided by the mean grayscale value of the background region to obtain the signal-to-noise ratio factor. Furthermore, the brightness distribution consistency factor set and the signal-to-noise ratio factor set are mapped and weighted calculated according to the weight value preset by those skilled in the art to obtain the image quality factor set.

[0035] Furthermore, the preset image quality factor threshold is the lowest image quality factor preset by those skilled in the art for subsequent analysis. It is judged whether the image quality factor set meets the preset image quality factor threshold. If not, an image re-acquisition instruction is obtained; if so, image acquisition does not need to be re-performed. The goal of image quality control is achieved, and the technical effect of ensuring that subsequent feature extraction and analysis are based on high-quality image data is achieved, thereby improving the accuracy and stability of detection and reducing the risk of misjudgment caused by image noise or uneven illumination.

[0036] The multi-scale feature set acquisition module 12 is used to traverse the acquired image set for multi-scale feature acquisition to obtain the acquired image multi-scale feature group set;

[0037] Furthermore, the multi-scale feature set acquisition module 12 includes:

[0038] The image Gaussian smoothing processing unit is used to convolve the acquired image set using the Gaussian kernel function to obtain the preprocessed acquired image set after Gaussian smoothing processing;

[0039] The image group set acquisition unit is used to sample the preprocessed acquired image set according to the preset sampling scale set respectively to obtain the preprocessed sampled image group set, where each preprocessed sampled image group is an image group obtained by sampling the preprocessed acquired image according to different sampling scales;

[0040] The multi-scale feature group set acquisition unit is used to traverse the preprocessed sampled image group set for image feature extraction to obtain the acquired image multi-scale feature group set.

[0041] Further, the Gaussian kernel function is as follows:

[0042]

[0043] where G(x, y) is the Gaussian weight value of the pixel at (x, y) in the acquired image, σ is the standard deviation, and (x, y) is the coordinate position of the pixel relative to the center of the Gaussian kernel.

[0044] In a possible embodiment, the features of an image often have different spatial scale characteristics. For example, large-scale features may contain overall organizational structure information, while small-scale features can reflect local texture or edge information. Therefore, through multi-scale feature extraction, it can be ensured that the detection system can take into account information at different granularities, improving the accuracy and stability of infection recognition. Therefore, by first performing Gaussian smoothing on the acquired image, then downsampling the image at different scales, and finally extracting features from the images at different scales, a multi-scale feature group set of the acquired image is constructed.

[0045] In one embodiment, Gaussian smoothing is used to preprocess the acquired image set respectively to reduce noise and smooth details while retaining the main features. By using the Gaussian kernel function to smooth the acquired image set, the preprocessed acquired image set is obtained. The preset sampling scales are sampling scales preset by those skilled in the art, including 1 / 2, 1 / 4, 1 / 8, etc. The preprocessed sampled image group refers to the image group scaled according to different sampling scales, and each group contains multiple resolution versions generated from the same image. Exemplarily, if the original image size is 1024×1024, then after sampling at the sampling scales of 1 / 2, 1 / 4, and 1 / 8, the obtained image group is 512×512, 256×256, and 128×128. Among them, high-resolution images are suitable for global structure analysis, and low-resolution images can ignore detail noise. Using images at different scales to provide different levels of infection features helps to improve the recognition ability of machine learning models or algorithms.

[0046] Preferably, the Sobel operator is used to extract the edge morphology of the infected area, and the gray-scale change degrees of the image in the horizontal and vertical directions are calculated respectively, that is, the gradient information of the image. Further, the gradient amplitude of each pixel in the image is calculated. The gradient amplitude represents the gray-scale change degree of the pixel. That is to say, the larger the value, the more likely the pixel is at the tissue boundary or the boundary of the infected area, and the smaller the value, the more likely the pixel is in the internal area of the tissue, so as to obtain the edge image features. Preferably, the gray-level co-occurrence matrix (GLCM) is calculated to extract features such as contrast, entropy value, and correlation of the image, analyze the roughness and pattern of the infected area, and thus obtain the overall image features. The edge image features and the overall image features are summarized to obtain the multi-scale features of the acquisition image of a preprocessed sampling image. Based on the above principle, image feature extraction is performed on the set of preprocessed sampling image groups to obtain the set of multi-scale features of the acquisition image.

[0047] The interactive iterative feature set obtaining module 13 is used to perform intra-group feature interactive iteration on the set of multi-scale features of the acquisition image respectively to obtain an interactive iterative feature set;

[0048] Furthermore, the interactive iterative feature set obtaining module 13 includes:

[0049] The multi-scale feature group set extraction unit of the acquisition image is used to extract the first multi-scale feature group of the acquisition image from the set of multi-scale features of the acquisition image;

[0050] The first enumeration combination obtaining unit is used to perform pairwise enumeration on the first multi-scale feature group of the acquisition image to obtain a first enumeration combination set;

[0051] The interactive iterative feature combination obtaining unit is used to traverse the first enumeration combination set to perform feature interactive iteration to obtain an interactive iterative feature set, where each interactive iterative feature corresponds to a first enumeration combination;

[0052] The first interactive iterative feature obtaining unit is used to perform feature integration on the interactive iterative feature set to obtain a first interactive iterative feature;

[0053] The interactive iterative feature set obtaining unit is used to perform intra-group feature interactive iteration on the remaining multi-scale feature groups of the acquisition image in the set of multi-scale features of the acquisition image to obtain an interactive iterative feature set.

[0054] In a possible embodiment, the in-group feature interaction iteration refers to the combined analysis of the features expressed by differently scaled sampled images within the same feature group to find the correlations between the features and continuously optimize the feature expression. The enumeration combination means combining the features in the same feature group pairwise to prepare for subsequent exploration of the correlations between the features. The convolution calculator is a functional module that uses the convolution operation in deep learning to further extract and fuse the features, thereby generating a new feature set.

[0055] In one embodiment, the first multi-scale feature group of the acquired images is extracted from the set of multi-scale feature groups of the acquired images. Then, the features in the first multi-scale feature group of the acquired images are combined pairwise to form the first enumeration combination set. This can ensure interactive calculations between different features and explore potential patterns. For example: combining the edge features of low-frequency information with the texture features of high-frequency information may help to more clearly distinguish the infected area and healthy tissue. Combining visible light features with infrared features may help to enhance the contrast of the tissue.

[0056] Furthermore, a feature union is obtained for the interactive iteration feature set to obtain the first interactive iteration feature. Based on the same principle of obtaining the first interactive iteration feature, in-group feature interaction iteration is performed on the remaining multi-scale feature groups of the acquired images in the set of multi-scale feature groups of the acquired images to obtain the interactive iteration feature set.

[0057] Further, the interactive iteration feature combination obtaining unit includes:

[0058] The feature similarity cluster obtaining subunit is used to calculate the feature similarities between the internal features of the combinations in the first enumeration combination set respectively to obtain the first enumeration combination feature similarity cluster;

[0059] The iterative matrix set obtaining subunit is used to perform normalization processing on the first enumeration combination feature similarity cluster and fill it into an initially empty matrix respectively to obtain the first interactive iteration matrix set;

[0060] The interactive iteration feature set obtaining subunit is used to perform convolution calculations on the first interactive iteration matrix set and the corresponding first enumeration combinations respectively to obtain the interactive iteration feature set.

[0061] Further, the interactive iteration feature set obtaining subunit includes:

[0062] The training data setting micro-unit is used to obtain a plurality of sample interactive iteration matrices and a plurality of sample multi-scale features of the acquired images as training data;

[0063] The convolution calculator obtains micro-units for evenly dividing the training data into n groups, uses the n groups of data to perform supervised training on a framework constructed based on a feedforward neural network, and updates the network parameters during the training to obtain a trained convolution calculator.

[0064] The convolution calculation micro-unit is used to perform convolution calculations on the two first acquisition image multi-scale features in the first interaction iteration matrix set and the corresponding first enumeration combination respectively by using the convolution calculator, so as to obtain an interaction iteration feature set.

[0065] In a possible embodiment, the first enumeration combination feature similarity cluster is used to measure the similarity between the features of a first enumeration combination, and is usually calculated based on statistical methods (such as cosine similarity, Euclidean distance) or machine learning methods (such as feature embedding). The calculated feature similarity cluster is normalized so that all data falls within the same numerical range (such as 0 to 1). The normalized data is filled into an initially empty matrix to form the first interaction iteration matrix set. Each first interaction iteration matrix is used to store the similarity information of different feature combinations and serves as the input for subsequent feature calculations.

[0066] Furthermore, the trained convolution calculator is used to perform convolution calculations on the first interaction iteration matrix set and the corresponding first enumeration combination, so as to enhance the association of features and obtain the interaction iteration feature combination. The goal of discovering deeper feature patterns through convolution calculations and improving the detection accuracy is achieved.

[0067] Preferably, the feature interaction process is optimized through deep learning methods to improve the recognition ability of the optical detection device for the infected area. First, optical images of multiple peritoneal dialysis infection samples are collected, multi-scale features are extracted, and the feature interaction matrix between samples is calculated to form a training data set. Subsequently, the data is evenly divided into multiple subsets, and supervised training is performed using a feedforward neural network. The network parameters are continuously optimized through backpropagation to enable it to learn the optimal combination patterns of different spectral features. After training is completed, the trained convolution calculator is used to process the two first acquisition image multi-scale features in the newly input first interaction iteration matrix set and the corresponding first enumeration combination, and an interaction iteration feature set is generated through convolution calculations, thereby enhancing the discrimination ability of the features, improving the detection accuracy and stability, and ensuring that the model can maintain a high detection reliability in different lighting conditions and different patient samples.

[0068] The mapping analysis result obtaining module 14 is used to perform infection mapping analysis based on the interaction iteration feature set to obtain a target mapping analysis result.

[0069] Furthermore, the mapping analysis result obtaining module 14 includes:

[0070] An infection mapping database acquisition unit for acquiring an infection mapping database;

[0071] A target mapping analysis result acquisition unit for retrieving the infection mapping database by using the interactive iterative feature set as an index to obtain a target mapping analysis result.

[0072] In a possible embodiment, the infection mapping database is a database storing different infection categories and their optical characteristics, containing a large number of labeled samples and being available for comparison and analysis. Preferably, those skilled in the art label the infection samples and obtain the spectral characteristics, morphological information, and classification labels of the infection samples. Furthermore, the database is searched and matched by using the interactive iterative feature set as an index. Specifically, the cosine similarity calculation method is used to calculate the similarity between the interactive iterative feature set and the infection characteristics stored in the database, so as to identify the most suitable infection type. Finally, the system outputs the target mapping analysis result, which can be the infection type and the infection probability. The technical effect of improving the accuracy and reliability of detection is achieved.

[0073] In summary, the optical detection device for quickly judging peritoneal dialysis exit infection provided by the present invention has the following technical effects:

[0074] This application calls a light source to emit light according to a preset light type sequence, uses a high-resolution camera and a filter wheel to collect images of different spectra to obtain a set of collected images, then traverses the set of collected images to perform multi-scale feature collection to obtain a set of multi-scale feature groups of the collected images, and then performs intra-group feature interaction iteration on the set of multi-scale feature groups of the collected images respectively to obtain an interactive iterative feature set, and performs infection mapping analysis based on the interactive iterative feature set to obtain a target mapping analysis result. The technical effect of extracting feature information of different scales and improving the accuracy of optical detection is achieved.

[0075] Embodiment 2 Figure 2 is a schematic flowchart of the optical detection method for quickly judging peritoneal dialysis exit infection of the present invention. For example, Figure 1 the schematic structural diagram of the optical detection device for quickly judging peritoneal dialysis exit infection in the present invention can be used to implement the process as Figure 2 shown.

[0076] Based on the same concept as the optical detection device for quickly judging peritoneal dialysis exit infection in the above embodiment, the optical detection method for quickly judging peritoneal dialysis exit infection provided by the present invention further includes:

[0077] Calling a light source to emit light according to a preset light type sequence, and using a high-resolution camera and a filter wheel to collect images of different spectra to obtain a set of collected images;

[0078] Traverse the set of collected images to perform multi-scale feature collection, and obtain a set of multi-scale feature groups of the collected images;

[0079] Perform intra-group feature interaction iteration on the set of multi-scale feature groups of the collected images respectively, and obtain a set of interaction iteration features;

[0080] Perform infection mapping analysis based on the set of interaction iteration features, and obtain the target mapping analysis result.

[0081] Further, the method further includes:

[0082] Traverse the set of collected images to perform brightness distribution consistency recognition, and obtain a set of brightness distribution consistency factors;

[0083] Traverse the set of collected images to perform signal-to-noise ratio analysis, and obtain a set of signal-to-noise ratio factors;

[0084] Perform weighted calculation on the set of brightness distribution consistency factors and the set of signal-to-noise ratio factors, and obtain a set of image quality factors;

[0085] Judge whether the set of image quality factors meets the preset image quality factor threshold. If not, obtain an image re-collection instruction.

[0086] Further, the method further includes:

[0087] Perform convolution on the set of collected images respectively by using a Gaussian kernel function, and obtain a set of preprocessed collected images after Gaussian smoothing processing;

[0088] Perform sampling on the set of preprocessed collected images respectively according to a preset sampling scale set, and obtain a set of preprocessed sampling image groups, where each preprocessed sampling image group is an image group obtained by sampling the preprocessed collected images according to different sampling scales;

[0089] Traverse the set of preprocessed sampling image groups to perform image feature extraction, and obtain a set of multi-scale feature groups of the collected images.

[0090] Further, the Gaussian kernel function is:

[0091]

[0092] Wherein, G(x, y) is the Gaussian weight value of the pixel point located at (x, y) in the collected image, σ is the standard deviation, and (x, y) is the coordinate position of the pixel point relative to the center of the Gaussian kernel.

[0093] Further, the method further includes:

[0094] Extract the first multi-scale feature group of the collected images from the set of multi-scale feature groups of the collected images;

[0095] Enumerate the multi-scale feature groups of the first acquired image pairwise to obtain a first enumeration combination set;

[0096] Traverse the first enumeration combination set for feature interaction iteration to obtain an interaction iteration feature set, where each interaction iteration feature corresponds to a first enumeration combination;

[0097] Integrate the features of the interaction iteration feature set to obtain a first interaction iteration feature;

[0098] Perform intra-group feature interaction iteration on the remaining multi-scale feature groups of the acquired image multi-scale feature group set to obtain an interaction iteration feature set.

[0099] Furthermore, the method further includes:

[0100] Calculate the feature similarity between the internal features of the combinations in the first enumeration combination set respectively to obtain a first enumeration combination feature similarity cluster;

[0101] Normalize the first enumeration combination feature similarity cluster and fill it into an initially empty matrix respectively to obtain a first interaction iteration matrix set;

[0102] Perform convolution calculation on the first interaction iteration matrix set and the corresponding first enumeration combination respectively to obtain an interaction iteration feature set.

[0103] Furthermore, the method further includes:

[0104] Obtain multiple sample interaction iteration matrices and multiple sample acquired image multi-scale features as training data;

[0105] Evenly divide the training data into n groups, use the n groups of data to perform supervised training on a framework constructed based on a feedforward neural network, and update the network parameters during training to obtain a trained convolution calculator;

[0106] Use the convolution calculator to perform convolution calculation on the first interaction iteration matrix set and two first acquired image multi-scale features in the corresponding first enumeration combination respectively to obtain an interaction iteration feature set.

[0107] Furthermore, the light source is a multi-spectral LED array.

[0108] Furthermore, the method further includes:

[0109] Obtain an infection mapping database;

[0110] Retrieve the infection mapping database with the interaction iteration feature set as the index to obtain a target mapping analysis result.

[0111] It should be understood that the disclosed embodiments of the present invention and the above descriptions enable those skilled in the art to implement the present invention using the present invention. At the same time, the present invention is not limited to the above-mentioned part of the embodiments. It should be understood that those of ordinary skill in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention and should all be included within the protection scope of the present invention.

Claims

1. An optical detection device for quickly judging peritoneal dialysis exit site infection, characterized in that, The device includes: An acquisition image set obtaining module, configured to call a light source to emit light according to a preset light type sequence, and use a high-resolution camera and a filter wheel to collect images of different spectra, so as to obtain an acquisition image set; A multi-scale feature set obtaining module, configured to traverse the acquisition image set for multi-scale feature acquisition, so as to obtain a set of acquisition image multi-scale feature groups; An interactive iterative feature set obtaining module, configured to perform intra-group feature interactive iteration on the set of acquisition image multi-scale feature groups respectively, so as to obtain an interactive iterative feature set; A mapping analysis result obtaining module, configured to perform infection mapping analysis based on the interactive iterative feature set, so as to obtain a target mapping analysis result.

2. The optical detection device for rapidly judging peritoneal dialysis exit-site infection according to claim 1, wherein The acquisition image set obtaining module includes: A brightness distribution consistency factor set obtaining unit, configured to traverse the acquisition image set for brightness distribution consistency recognition, so as to obtain a brightness distribution consistency factor set; A signal-to-noise ratio factor set obtaining unit, configured to traverse the acquisition image set for signal-to-noise ratio analysis, so as to obtain a signal-to-noise ratio factor set; An image quality factor set obtaining unit, configured to perform weighted calculation on the brightness distribution consistency factor set and the signal-to-noise ratio factor set, so as to obtain an image quality factor set; An image re-acquisition instruction obtaining unit, configured to determine whether the image quality factor set meets a preset image quality factor threshold, and if not, obtain an image re-acquisition instruction.

3. The optical detection device for quickly judging peritoneal dialysis exit infection according to claim 1, wherein, The multi-scale feature set obtaining module includes: An image Gaussian smoothing processing unit, configured to perform convolution on the acquisition image set respectively by using a Gaussian kernel function, so as to obtain a preprocessed acquisition image set after Gaussian smoothing processing; An image group set obtaining unit, configured to sample the preprocessed acquisition image set respectively according to a preset sampling scale set, so as to obtain a set of preprocessed sampled image groups, where each preprocessed sampled image group is an image group obtained by sampling the preprocessed acquisition image according to different sampling scales; A multi-scale feature group set obtaining unit, configured to traverse the set of preprocessed sampled image groups for image feature extraction, so as to obtain a set of acquisition image multi-scale feature groups.

4. The optical detection device for rapidly judging peritoneal dialysis exit-site infection according to claim 3, wherein, The Gaussian kernel function is: where G(x, y) is the Gaussian weight value of the pixel point at (x, y) in the acquisition image, σ is the standard deviation, and (x, y) is the coordinate position of the pixel point relative to the center of the Gaussian kernel.

5. The optical detection device for rapidly judging peritoneal dialysis exit-site infection according to claim 1, wherein, The interactive iterative feature set obtaining module includes: An acquisition image multi-scale feature group set extraction unit, configured to extract a first acquisition image multi-scale feature group from the set of acquisition image multi-scale feature groups; A first enumeration combination obtaining unit, configured to perform pairwise enumeration on the first acquisition image multi-scale feature group, so as to obtain a first enumeration combination set; An interactive iterative feature combination obtaining unit, configured to traverse the first enumeration combination set for feature interactive iteration, so as to obtain an interactive iterative feature set, where each interactive iterative feature corresponds to a first enumeration combination; A first interactive iterative feature obtaining unit, configured to perform feature integration on the interactive iterative feature set, so as to obtain a first interactive iterative feature; An interactive iterative feature set obtaining unit is configured to perform in-group feature interactive iteration on the remaining collected image multi-scale feature groups in the collected image multi-scale feature group set, so as to obtain an interactive iterative feature set.

6. The optical detection device for quickly judging peritoneal dialysis exit infection according to claim 5, wherein, The interactive iterative feature combination obtaining unit includes: A feature similarity cluster obtaining subunit is configured to calculate the feature similarities between the internal features of the combinations in the first enumerated combination set respectively, so as to obtain a first enumerated combination feature similarity cluster; An iterative matrix set obtaining subunit is configured to perform normalization processing on the first enumerated combination feature similarity cluster and fill it into an initially empty matrix respectively, so as to obtain a first interactive iterative matrix set; An interactive iterative feature set obtaining subunit is configured to perform convolution calculation on the first interactive iterative matrix set and the corresponding first enumerated combination respectively, so as to obtain an interactive iterative feature set.

7. The optical detection device for rapidly judging peritoneal dialysis exit-site infection according to claim 6, wherein, The interactive iterative feature set obtaining subunit includes: A training data setting micro-unit is configured to obtain a plurality of sample interactive iterative matrices and a plurality of sample collected image multi-scale features as training data; A convolution calculator obtaining micro-unit is configured to equally divide the training data into n groups, use the n groups of data to perform supervised training on a framework constructed based on a feedforward neural network, and update network parameters during the training, so as to obtain a trained convolution calculator; A convolution calculation micro-unit is configured to use the convolution calculator to perform convolution calculation on two first collected image multi-scale features in the first interactive iterative matrix set and the corresponding first enumerated combination respectively, so as to obtain an interactive iterative feature set.

8. The optical detection device for rapidly judging peritoneal dialysis exit-site infection according to claim 1, wherein, The light source is a multi-spectral LED array.

9. The optical detection device for quickly judging peritoneal dialysis exit infection according to claim 1, wherein, The mapping analysis result obtaining module includes: An infection mapping database obtaining unit is configured to obtain an infection mapping database; A target mapping analysis result obtaining unit is configured to retrieve the infection mapping database with the interactive iterative feature set as an index, so as to obtain a target mapping analysis result.

10. An optical detection method for quickly judging peritoneal dialysis exit-site infection, characterized in that, including: Call the light source to emit light according to a preset light type sequence, and use a high-resolution camera and a filter wheel to collect images of different spectra, so as to obtain a set of collected images; Traverse the set of collected images to perform multi-scale feature collection, so as to obtain a set of collected image multi-scale feature groups; Perform in-group feature interactive iteration on the set of collected image multi-scale feature groups respectively, so as to obtain an interactive iterative feature set; Perform infection mapping analysis based on the interactive iterative feature set, so as to obtain a target mapping analysis result.

Images