Wax block and slice registration system and registration method

Abstract

The invention discloses a wax block and slice registration system and a registration method. The method comprises the following steps: acquiring a facial image of an operator, comparing facial features with data in a database, judging the identity of the operator according to similarity, reading the data based on an ambient light sensor, and automatically adjusting the illumination intensity of a planar light source; a wax block and a slice are placed on the wax block table and the slice table respectively, and voice instructions are collected through a voice recognition module and transmitted back to the control end of the embedded development platform; color images of the wax block table and the slicing table are collected and processed; decoding the processed wax block and slice image, and if the pathological numbers of the wax block and the slice are consistent, performing detection segmentation on the wax block and the slice to complete registration; otherwise, replacing the slice, and re-collecting the image. According to the method, accurate registration of the wax blocks and the slices is automatically carried out in the pathological slide making process, a pathologist can be assisted to carry out registration work of the wax blocks and the slices, and manual operation and subjective errors are reduced.

Description

Technical Field

[0001] The invention relates to the technical field of pathological medicine, and in particular to a wax block and slice registration system and registration method. Background Art

[0002] Pathology is of great significance for clinical diagnosis guidance. Checking whether there are patch errors, slice damage, etc. in the process of making pathological slices is an important part of pathological slice making. At present, the matching detection of wax blocks and slices still requires doctors to manually screen one by one. This work is very time-consuming and there are human subjective errors, which increases the risk of misdiagnosis caused by the inability to screen wrong patches and slice damage. This study can realize a wax block and skin cut automatic matching detection system and method, that is, assisting doctors to perform automated wax block and slice similarity detection, thereby reducing manual labor and reducing the misdiagnosis rate caused by matching errors, which can greatly improve the efficiency and accuracy of medical work. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a wax block and slice registration system and registration method to realize automatic wax block and slice matching detection.

[0004] On the one hand, to achieve the above-mentioned object, the present invention provides a wax block and slice registration system, comprising: a control module and an operation module, wherein the control module is connected to the operation module;

[0005] The control module includes: an embedded development platform, a touch display screen, a first color industrial camera, a second color industrial camera, a voice recognition module, a micro printer, a mobile device, an ambient light sensor, a facial recognition module, and a monocular camera; the embedded development platform is respectively connected to the touch display screen, the first color industrial camera, the second color industrial camera, the voice recognition module, the micro printer, the ambient light sensor, the facial recognition module, and the monocular camera by wire, and the mobile device is wirelessly connected to the embedded development platform;

[0006] The embedded development platform is located at the top of the system, the touch display screen is flatly connected above the embedded development platform, the first color industrial camera and the second color industrial camera are arranged and connected at the bottom of the touch display screen, the facial recognition module is connected behind the touch display screen, and the monocular camera is connected just above the touch display screen; the voice recognition module and the micro printer are arranged on the brackets respectively; the ambient light sensor is located at the lower left of the cabinet;

[0007] The operation module includes: a cabinet, a wax block table, a slicing table, an operation panel, and a planar light source; the cabinet is located at the bottom of the system, the operation panel is installed on the upper surface of the cabinet, the planar light source is installed flat below the operation panel, the wax block table is located on the left side of the operation panel, and the slicing table is located on the right side of the operation panel.

[0008] On the other hand, to achieve the above-mentioned purpose, the present invention also provides a registration method of a wax block and a slice registration system, comprising:

[0009] The monocular camera in the control module collects the operator's facial image, extracts facial features based on the facial recognition module, compares the facial features with the data in the database, and determines the identity of the operator based on the similarity. After the identity verification is passed, the ambient light sensor reads the data and automatically adjusts the light intensity of the plane light source;

[0010] Wax blocks and slices are placed on the wax block table and the slice table respectively, and voice commands are collected through the voice recognition module and transmitted back to the control end of the embedded development platform; after the control end sends out the command signal, the color images of the wax block table and the slice table are collected through the color industrial camera respectively, and the color images are processed, and the processed images are displayed on the touch display screen;

[0011] The processed wax block and slice images are decoded. If the pathological numbers of the wax block and the slice are consistent, the wax block and the slice are detected and segmented, and the detection results are saved to complete the registration; otherwise, the slice is replaced and the image is recaptured.

[0012] Preferably, judging the identity of the operator according to the similarity includes:

[0013] The collected facial image is preprocessed by grayscale, denoising and light compensation, and then the facial feature vector is extracted by the facial recognition module, the facial feature vector is compared with the feature vector in the database, and the cosine similarity is calculated. If the cosine similarity exceeds a preset threshold, it is judged that the identity authentication is passed, otherwise the identity authentication fails.

[0014] Preferably, automatically adjusting the illumination intensity of the planar light source based on reading data from an ambient light sensor includes:

[0015] The current light intensity data is transmitted back to the control end of the embedded development platform through the ambient light sensor. The control end uses a PID algorithm based on feedback control to process the light intensity data and automatically adjusts the light intensity of the planar light source according to the output value of the PID. If the output value is large, the light brightness is increased, and if the output value is small, the light brightness is reduced.

[0016] Preferably, the voice command is collected by the voice recognition module and transmitted back to the control end of the embedded development platform, including:

[0017] Wax blocks and slices with barcodes are placed on the wax block table and the slice table respectively. The speech recognition module collects the user's voice input through a microphone device, pre-processes the collected voice signal, extracts Mel frequency cepstral coefficients MFCC from the pre-processed voice signal, inputs the Mel frequency cepstral coefficients MFCC into the speech recognition model and converts them into text instructions, parses the text instructions through the control end of the embedded development platform, and executes corresponding operations according to the parsed instructions;

[0018] The method for preprocessing the collected speech signal is as follows: removing background noise by using a spectral subtraction noise reduction technique, and adjusting the strength of the speech signal based on an automatic gain control algorithm.

[0019] Preferably, the color images of the wax block stage and the slicing stage are collected by the color industrial camera respectively, and the color images are processed, including:

[0020] The color industrial camera collects color images of the wax block stage and the slice stage respectively within a time period t, and performs color adjustment on the collected images through an automatic color balancing algorithm based on Retinex and local adaptive filtering;

[0021] The color-adjusted RGB image is converted to the LAB color space, shadow pixels are detected according to the mean of channel A and channel B, the color of the shadow area is adjusted, and the shadow edge is corrected using median filtering to obtain the processed image.

[0022] Preferably, the processed wax block and slice image are decoded, and if the pathological numbers of the wax block and the slice are consistent, the wax block and the slice are detected and segmented, including:

[0023] The pre-processed wax block and slice image are decoded by a method based on OpenCV, the barcode in black and white stripe pattern is read by the preset position coordinates, and the barcode is parsed by the BarcodeDetector method based on OpenCV to obtain the corresponding pathology number;

[0024] If the pathological numbers of the wax block and the slice are consistent, the wax block and the slice are detected and segmented based on Maskformer.

[0025] Preferably, the wax block and the slice are detected and segmented based on Maskformer, including:

[0026] The wax block and slice are subjected to feature map extraction by Image Encoder, and the feature map is passed through the Pixel Decoder module to generate a high-resolution feature map. Subsequently, the Transformer decoder is used to convert the high-resolution feature map into a mask and category label, and the segmentation result is obtained by corresponding to the corresponding image. The Hu moment vector of the segmentation result is calculated, and the contour features of the wax block and slice are characterized by the Hu moment vector.

[0027] Preferably, the detection results are saved and the registration is completed, including:

[0028] The segmentation results of the wax block and the slice are spliced ​​together, and the spliced ​​feature vector is input into the trained multi-layer perceptron MLP to output the similarity;

[0029] A similarity detection record of the wax block and the slice is generated with the pathology number as the file name, the similarity detection record is transmitted to a predetermined storage location, and the matching result record is printed by a printer; wherein the similarity detection record includes a color image, a segmentation result, and a similarity result that need to be archived.

[0030] Compared with the prior art, the present invention has the following advantages and technical effects:

[0031] The present invention automatically performs accurate registration of wax blocks and slices during the pathological slice making process, which can assist doctors in automatically performing accurate matching detection of wax blocks and slices when making pathological slices, thereby reducing manual labor and human errors in pasting slices, and reducing the misdiagnosis rate caused by subjective matching errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0033] Figure 1 A structural framework diagram of a wax block and slice registration system according to an embodiment of the present invention;

[0034] Figure 2 is a flow chart of execution functions of an embodiment of the present invention;

[0035] Figure 3 is a flow chart of a registration method according to an embodiment of the present invention;

[0036] Among them, 1-embedded development platform, 2-touch display screen, 3-first color industrial camera, 4-second color industrial camera, 5-voice recognition module, 6-micro printer, 7-cabinet, 8-wax block table, 9-slicing table, 10-operation panel, 11-planar light source, 12-mobile device, 13-ambient light sensor, 14-facial recognition module, 15-monocular camera. DETAILED DESCRIPTION

[0037] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0038] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] The present invention proposes a wax block and slice registration system, such as Figure 1 , including: a control module and an operation module, the control module is connected with the operation module;

[0040] Wherein, the control module includes: an embedded development platform 1, a touch display screen 2, a first color industrial camera 3, a second color industrial camera 4, a voice recognition module 5, a micro printer 6, a mobile device end 12, an ambient light sensor 13, a facial recognition module 14, and a monocular camera 15; the embedded development platform 1 is respectively connected to the touch display screen 2, the first color industrial camera 3, the second color industrial camera 4, the voice recognition module 5, the micro printer 6, the ambient light sensor 13, the facial recognition module 14, and the monocular camera 15 by wire, and the mobile device end 12 is wirelessly connected to the embedded development platform 1;

[0041] The embedded development platform 1 is located at the top of the system, the touch display screen 2 is connected flatly above the embedded development platform 1, the first color industrial camera 3 and the second color industrial camera 4 are connected at the bottom of the touch display screen 2, the facial recognition module 14 is connected behind the touch display screen 2, and the monocular camera 15 is connected just above the touch display screen 2; the voice recognition module 5 and the micro printer 6 are arranged on the brackets respectively; the ambient light sensor 13 is located at the lower left of the cabinet 7;

[0042] The operation module includes: a cabinet 7, a wax block table 8, a slicing table 9, an operation panel 10, and a planar light source 11; the cabinet 7 is located at the bottom of the system, the operation panel 10 is installed on the upper surface of the cabinet 7, the planar light source 11 is installed flatly below the operation panel 10, the wax block table 8 is located on the left side of the operation panel 10, and the slicing table 9 is located on the right side of the operation panel 10.

[0043] This embodiment also provides a registration method for a wax block and a slice registration system, such as Figure 2-Figure 3 ,include:

[0044] The monocular camera in the control module collects the operator's facial image, extracts facial features based on the facial recognition module, compares the facial features with the data in the database, and determines the identity of the operator based on the similarity. After the identity verification is passed, the ambient light sensor reads the data and automatically adjusts the light intensity of the plane light source;

[0045] Wax blocks and slices are placed on the wax block table and the slice table respectively, and voice commands are collected through the voice recognition module and transmitted back to the control end of the embedded development platform; after the control end sends out the command signal, the color images of the wax block table and the slice table are collected through the color industrial camera respectively, and the color images are processed, and the processed images are displayed on the touch display screen;

[0046] The processed wax block and slice images are decoded. If the pathological numbers of the wax block and the slice are consistent, the wax block and the slice are detected and segmented, and the detection results are saved to complete the registration; otherwise, the slice is replaced and the image is recaptured.

[0047] The present invention automatically performs accurate registration of wax blocks and slices during the pathological slice making process, which can assist doctors in automatically performing accurate matching detection of wax blocks and slices when making pathological slices, thereby reducing manual labor and human errors in pasting slices, and reducing the misdiagnosis rate caused by subjective matching errors.

[0048] Furthermore, judging the identity of the operator according to the similarity includes:

[0049] The collected facial images are preprocessed by grayscale, denoising and light compensation, and then the facial feature vector is extracted through the facial recognition module. The facial feature vector is compared with the feature vector in the database, and the cosine similarity is calculated. If the cosine similarity exceeds the preset threshold, the identity authentication is judged to be successful, otherwise the identity authentication fails.

[0050] Use a monocular camera to collect facial images, pre-process the images, extract facial features through the facial recognition module, compare with the database, and determine the identity based on the similarity. The specific implementation process of this step is as follows:

[0051] After the facial image is captured by a monocular camera, preprocessing such as grayscale, denoising and light compensation is performed, and then the facial feature vector is extracted by the facial recognition module, compared with the feature vector in the database, and the cosine similarity is calculated. In this embodiment, if the similarity exceeds the preset threshold of 0.65, it is judged that the identity authentication is passed, otherwise the identity authentication fails.

[0052] Furthermore, based on the ambient light sensor reading data and automatically adjusting the light intensity of the planar light source, the method includes:

[0053] The current light intensity data is transmitted back to the control end of the embedded development platform through the ambient light sensor. The control end uses the PID algorithm based on feedback control to process the light intensity data and automatically adjusts the light intensity of the planar light source according to the output value of the PID. If the output value is large, the light brightness is increased, and if the output value is small, the light brightness is reduced.

[0054] After the identity authentication is passed, the ambient light sensor data is read and the light intensity of the planar light source is automatically adjusted based on the feedback control algorithm. The specific implementation process of this step is as follows:

[0055] After the identity authentication is passed, the ambient light sensor then transmits the current light intensity back to the control end. The control end uses the PID algorithm based on feedback control to process the light data, and then automatically adjusts the light intensity of the plane light source according to the output of the PID. The larger the output value, the greater the light brightness, and the smaller the output value, the lower the light brightness, ensuring that the system works in a good lighting environment.

[0056] Furthermore, the voice commands are collected through the voice recognition module and transmitted back to the control end of the embedded development platform, including:

[0057] Wax blocks and slices with barcodes are placed on the wax block table and the slice table respectively. The speech recognition module collects the user's voice input through a microphone device, pre-processes the collected voice signal, extracts Mel frequency cepstral coefficients MFCC from the pre-processed voice signal, inputs the Mel frequency cepstral coefficients MFCC into the speech recognition model and converts them into text instructions. The text instructions are parsed through the control end of the embedded development platform, and corresponding operations are performed according to the parsed instructions;

[0058] The method for preprocessing the collected speech signal is as follows: removing background noise by using a spectral subtraction noise reduction technique, and adjusting the strength of the speech signal based on an automatic gain control algorithm.

[0059] Place the wax block and slice with barcode on the wax block table and slice table respectively. The voice recognition module collects voice input, pre-processes, extracts features and recognizes the collected voice signals, and parses them into execution instructions and transmits them back to the control end. The specific implementation process of this step is as follows:

[0060] Wax blocks and slices with barcodes are placed on the wax block table and the slice table respectively. The voice recognition module collects the user's voice input through the microphone device, pre-processes the collected voice signal, removes background noise through the spectral subtraction noise reduction technology, and adjusts the strength of the voice signal through the automatic gain control algorithm; extracts the Mel-frequency cepstral coefficient (MFCC) from the pre-processed voice signal, inputs the Mel-frequency cepstral coefficient (MFCC) into the voice recognition model to convert it into text instructions, and the control end parses the text instructions and performs corresponding operations according to the parsed instructions.

[0061] Furthermore, the color images of the wax block stage and the slice stage are collected respectively by a color industrial camera, and the color images are processed, including:

[0062] The color industrial camera collects color images of the wax block stage and the slice stage respectively within a time period t, and the collected images are color adjusted through an automatic color equalization algorithm based on Retinex and local adaptive filtering; the RGB image after color adjustment is converted to the LAB color space, and the shadow pixels are detected according to the mean of channel A and channel B. The color of the shadow area is adjusted, and the median filter is used to correct the shadow edge to obtain the processed image.

[0063] After the control end sends a signal, the color industrial cameras on both sides collect color images of the wax block stage and the slice stage respectively, perform color balancing, shadow removal and over-illumination processing on the images, and the processed images are displayed on the touch screen. The specific implementation process of this step is as follows:

[0064] After the control end sends a signal, the color industrial cameras on both sides collect color images of the wax block stage and the slice stage respectively within the t time period. The collected images are adjusted through local adaptive filtering based on the Retinex automatic color balancing algorithm; the RGB image after color adjustment is converted to the LAB color space, the shadow pixels are detected according to the mean of channel A and channel B, and the color of the shadow area is adjusted. The median filter is used to correct the shadow edge to prevent excessive lighting; the processed image will be displayed on the touch screen for the operator to view.

[0065] Furthermore, the processed wax block and slice images are decoded. If the pathological numbers of the wax block and the slice are consistent, the wax block and the slice are detected and segmented, including:

[0066] The preprocessed wax block and slice images are barcode decoded using an OpenCV-based method. The black and white stripe pattern barcode is read using the preset position coordinates, and the barcode is parsed using the OpenCV-based BarcodeDetector method to obtain the corresponding pathology number. If the pathology numbers of the wax block and the slice are consistent, the wax block and the slice are detected and segmented separately based on Maskformer.

[0067] The pre-processed wax block and slice images are decoded by barcode using an opencv-based method. If the pathological numbers of the wax block and the slice are consistent, the wax block and the slice are detected and segmented based on Maskformer, and the Hu moment vector of the segmentation results of the two is calculated to characterize the contour feature vectors of the wax block and the slice; otherwise, the slice is replaced and the image is re-collected. The specific implementation process of this step is as follows:

[0068] The processed image is barcode decoded, and the barcode with black and white stripes is read through the preset position coordinates. The barcode is parsed using the BarcodeDetector method based on OpenCV to obtain the corresponding pathology number. If the pathology numbers are consistent, Maskformer is used for detection and segmentation. Specifically, the wax block and slice are passed through ImageEncoder to extract image features, and the feature map is passed through the Pixel Decoder module to generate a high-resolution feature map. The high-resolution feature map is then converted into a mask and category label using the Transformer decoder, and the segmentation result is obtained by corresponding to the image. The Hu moment vector of the segmentation result is calculated to characterize the contour features of the wax block and slice. Otherwise, the user is prompted that the wax block and slice pathology number do not match, and the slice is replaced to re-acquire the image for registration.

[0069] Furthermore, the detection results are saved and the registration is completed, including:

[0070] The segmentation results of the wax block and the slice are spliced, the spliced ​​feature vector is input into the trained multi-layer perceptron MLP, and the similarity is output; the pathology number is used as the file name, and a similarity detection record of the wax block and the slice is generated, the similarity detection record is transmitted to a predetermined storage location, and the matching result record is printed by a printer; wherein the similarity detection record contains the color image to be archived, the segmentation result and the similarity result.

[0071] Calculate the similarity of the segmentation results of the wax block and the slice, use the pathology number as the file name, generate the similarity detection record of the wax block and the slice, and transfer the images, segmentation results, similarity results and generated matching records to be archived to the designated storage location. The specific implementation process of this step is as follows:

[0072] The feature vectors of the wax block and the slice are spliced, and these spliced ​​feature vectors are input into the trained multi-layer perceptron MLP, and the similarity is output through the output layer; the similarity detection record of the wax block and the slice is generated with the pathology number as the file name. The record contains the color image, segmentation result and similarity result that need to be archived, and is transmitted to a predetermined storage location. At the same time, the matching result record is printed through a printer.

[0073] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A wax block and slice registration system, characterized in that: include: A control module and an operation module, wherein the control module is connected to the operation module; The control module comprises: an embedded development platform (1), a touch display screen (2), a first color industrial camera (3), a second color industrial camera (4), a voice recognition module (5), a micro printer (6), a mobile device (12), an ambient light sensor (13), a facial recognition module (14), and a monocular camera (15); the embedded development platform (1) is respectively connected to the touch display screen (2), the first color industrial camera (3), the second color industrial camera (4), the voice recognition module (5), the micro printer (6), the ambient light sensor (13), the facial recognition module (14), and the monocular camera (15) by wire, and the mobile device (12) is wirelessly connected to the embedded development platform (1); The embedded development platform (1) is located at the top of the system, the touch display screen (2) is flatly connected above the embedded development platform (1), the first color industrial camera (3) and the second color industrial camera (4) are arranged and connected at the bottom of the touch display screen (2), the facial recognition module (14) is connected behind the touch display screen (2), and the monocular camera (15) is connected directly above the touch display screen (2); the voice recognition module (5) and the micro printer (6) are arranged on the bracket respectively; the ambient light sensor (13) is located at the lower left of the cabinet (7); The operation module comprises: a cabinet (7), a wax block table (8), a slicing table (9), an operation panel (10), and a planar light source (11); the cabinet (7) is located at the bottom of the system, the operation panel (10) is installed on the upper surface of the cabinet (7), the planar light source (11) is installed flat below the operation panel (10), the wax block table (8) is located on the left side of the operation panel (10), and the slicing table (9) is located on the right side of the operation panel (10).

2. A registration method applied to the wax block and slice registration system according to claim 1, characterized in that: include: The monocular camera in the control module collects the operator's facial image, extracts facial features based on the facial recognition module, compares the facial features with the data in the database, and determines the identity of the operator based on the similarity. After the identity verification is passed, the ambient light sensor reads the data and automatically adjusts the light intensity of the plane light source; Wax blocks and slices are placed on the wax block table and the slice table respectively, and voice commands are collected through the voice recognition module and transmitted back to the control end of the embedded development platform; after the control end sends out the command signal, the color images of the wax block table and the slice table are collected through the color industrial camera respectively, and the color images are processed, and the processed images are displayed on the touch display screen; Decode the processed wax block and slice images. If the pathological numbers of the wax block and the slice are consistent, detect and segment the wax block and the slice, save the detection results, and complete the registration. Otherwise, replace the slice and re-acquire the image.

3. The registration method according to claim 2, characterized in that: Judging the identity of the operator according to the similarity includes: The collected facial image is preprocessed by grayscale, denoising and light compensation, and then the facial feature vector is extracted by the facial recognition module, the facial feature vector is compared with the feature vector in the database, and the cosine similarity is calculated. If the cosine similarity exceeds a preset threshold, it is judged that the identity authentication is passed, otherwise the identity authentication fails.

4. The registration method according to claim 3, characterized in that: Based on the data read by the ambient light sensor, the light intensity of the planar light source is automatically adjusted, including: The current light intensity data is transmitted back to the control end of the embedded development platform through the ambient light sensor. The control end uses a PID algorithm based on feedback control to process the light intensity data and automatically adjusts the light intensity of the planar light source according to the output value of the PID. If the output value is large, the light brightness is increased, and if the output value is small, the light brightness is reduced.

5. The registration method according to claim 2, characterized in that: The voice command is collected by the voice recognition module and transmitted back to the control end of the embedded development platform, including: Wax blocks and slices with barcodes are placed on the wax block table and the slice table respectively. The speech recognition module collects the user's voice input through a microphone device, pre-processes the collected voice signal, extracts Mel frequency cepstral coefficients MFCC from the pre-processed voice signal, inputs the Mel frequency cepstral coefficients MFCC into the speech recognition model and converts them into text instructions, parses the text instructions through the control end of the embedded development platform, and executes corresponding operations according to the parsed instructions; The method for preprocessing the collected speech signal is as follows: removing background noise by using a spectral subtraction noise reduction technique, and adjusting the strength of the speech signal based on an automatic gain control algorithm.

6. The registration method according to claim 2, characterized in that: The color industrial camera is used to collect color images of the wax block stage and the slice stage respectively, and the color images are processed, including: The color industrial camera collects color images of the wax block stage and the slice stage respectively within a time period t, and performs color adjustment on the collected images through an automatic color balancing algorithm based on Retinex and local adaptive filtering; The color-adjusted RGB image is converted to the LAB color space, shadow pixels are detected according to the mean of channel A and channel B, the color of the shadow area is adjusted, and the shadow edge is corrected using median filtering to obtain the processed image.

7. The registration method according to claim 2, characterized in that: Decode the processed wax block and slice images. If the pathological numbers of the wax block and slice are consistent, detect and segment the wax block and slice, including: The pre-processed wax block and slice image are decoded by a method based on OpenCV, the barcode in black and white stripe pattern is read by the preset position coordinates, and the barcode is parsed by the BarcodeDetector method based on OpenCV to obtain the corresponding pathology number; If the pathological numbers of the wax block and the slice are consistent, the wax block and the slice are detected and segmented based on Maskformer.

8. The registration method according to claim 7, characterized in that: Based on Maskformer, the wax blocks and slices are detected and segmented respectively, including: The wax block and slice are subjected to feature map extraction by Image Encoder, and the feature map is passed through the Pixel Decoder module to generate a high-resolution feature map. Subsequently, the Transformer decoder is used to convert the high-resolution feature map into a mask and category label, and the segmentation result is obtained by corresponding to the corresponding image. The Hu moment vector of the segmentation result is calculated, and the contour features of the wax block and slice are characterized by the Hu moment vector.

9. The registration method according to claim 8, characterized in that: The detection results are saved to complete the registration, including: The segmentation results of the wax block and the slice are spliced ​​together, and the spliced ​​feature vector is input into the trained multi-layer perceptron MLP to output the similarity; A similarity detection record of the wax block and the slice is generated with the pathology number as the file name, the similarity detection record is transmitted to a predetermined storage location, and the matching result record is printed by a printer; wherein the similarity detection record includes a color image, a segmentation result, and a similarity result that need to be archived.

Images