Detection method and system, detection device, equipment and storage medium

By detecting the characteristic data and optical imaging data of the sample, the detection model is constructed, and the problem of low detection efficiency and accuracy of multifocal tumor resection in surgical procedures is solved, achieving higher detection precision and efficiency.

CN118941488AActive Publication Date: 2024-11-12FUDAN UNIVERSITY

Patent Information

Application Number
CN202310524936.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-09
Publication Date
2024-11-12
Estimated Expiration
2043-05-09

AI Technical Summary

Technical Problem

The efficiency, precision and accuracy of multifocal tumor resection detection in existing surgical procedures still need to be improved. Especially in local enlarged resection and motype microsurgery, the pathological examination takes a long time and is not very accurate, making it difficult to take into account the scope of resection and Keep normal tissue.

Method used

By obtaining the characteristic data of the detection sample, including the signal optical characteristic curves of multiple measurement points and the optical imaging data of tissue slices, image segmentation is performed to obtain segmented images and block labels, the detection model is constructed, the model is trained using the training data set and performed Prediction, realize the classification judgment of measurement points.

Benefits of technology

It improves the accuracy and precision of detection, simplifies the detection process, reduces detection time, is suitable for detection of multifocal diseases, improves the efficiency of feature data acquisition, and does not require fitting analysis, which improves the detection efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941488A_ABST
    Figure CN118941488A_ABST
Patent Text Reader

Abstract

The invention discloses a detection method and system, a detection device, equipment and a storage medium, and the detection method comprises the steps: carrying out the image segmentation of an optical image of each tissue slice, so as to obtain a plurality of segmented images, and obtaining a block label of each segmented image in the tissue slice; the classification labels corresponding to the measurement points are obtained based on the position corresponding relation between the segmented images and the measurement points and the block labels of the segmented images, compared with the optical image of each tissue slice, the segmented images and the measurement points are smaller in size and higher in fineness, the fine degree of a detection model during classification judgment can be improved, and the detection accuracy is improved. The detection accuracy and fineness are correspondingly improved, the efficiency of obtaining the feature data is improved by obtaining the multiple measurement points on the detection sample and the signal light feature curves of the measurement points, prediction processing can be carried out based on original data through a detection model to obtain corresponding classification labels, and the detection accuracy and fineness are improved. And fitting analysis on a signal light characteristic curve is not needed, so that the detection efficiency is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of optical detection, and in particular, to a detection method and system, a detection device, equipment, and a storage medium. Background Art

[0002] The selection of surgical scope has always been the focus of surgical treatment. Excessive resection will cause excessive wounds and involve important organs, while incomplete resection will lead to a high recurrence rate. How to remove all lesions as much as possible while preserving normal tissue as much as possible is a major challenge.

[0003] At present, surgical operations for multifocal tumor resection, such as melanoma, basal cell carcinoma of the eyelid, fibrosarcoma protuberans, and extramammary Paget's disease, often use local extended resection or Mohs micrographic surgery. The current local extended resection plan is to expand the resection around the tumor by 1-5 cm, but there is much debate about the appropriate range of expansion. Mohs micrographic surgery requires layer-by-layer examination of the tissue pathology of the resection margin, which is time-consuming, has many pathological examination points (up to hundreds or even thousands of points), and a huge workload for pathology. In addition, the quality and accuracy of intraoperative frozen section examinations are not ideal, and positive resection margins cannot be completely avoided. Therefore, it is particularly important to take into account the appropriate resection range and preserve normal tissue as much as possible, and excellent resection margin judgment technology is particularly important.

[0004] However, the detection efficiency, precision and accuracy still need to be improved. Summary of the invention

[0005] The problem solved by the embodiments of the present invention is to provide a detection method and system, a detection device, equipment and a storage medium to improve the accuracy, precision and efficiency of detection.

[0006] To solve the above problems, an embodiment of the present invention provides a detection method, including: obtaining characteristic data of a detection sample, the characteristic data including a plurality of measurement points located on the detection sample and a characteristic curve of a signal light generated by each of the measurement points; obtaining optical imaging data of tissue slices of the detection sample at intervals of several positions; based on the optical imaging data, performing image segmentation on the optical image of each of the tissue slices to obtain a plurality of segmented images; obtaining a block label of each segmented image in the tissue slice; obtaining a classification label corresponding to the measurement point based on the position correspondence between the segmented image and the measurement point, and the block label of the segmented image, the measurement point and the corresponding classification label FDU-PA230002-CN

[0007] The signal light characteristic curve is used to form a first training data set; an initial detection model is constructed; the initial detection model is trained using the first training data set to obtain a trained detection model; and the measurement point and the corresponding signal light characteristic curve are predicted using the trained detection model to obtain a corresponding classification label.

[0008] Correspondingly, an embodiment of the present invention further provides a detection device, which is used to obtain the characteristic data of the detection sample in the detection method described in the embodiment of the present invention; the detection device includes: a light source, which is used to provide detection light; an optical fiber detection end, which is connected to the light source, and the optical fiber detection end is used as a detection end for detecting a measurement point on the detection sample; wherein the detection light is irradiated onto the measurement point through the optical fiber detection end so that the detection sample at the measurement point generates signal light; a detection module, which is connected to the optical fiber detection end, and the detection module is used to detect the signal light generated by the detection sample at the measurement point, and obtain the signal light characteristic curve of each measurement point.

[0009] Correspondingly, an embodiment of the present invention further provides a detection system, including: a data acquisition module, used to obtain characteristic data of a detection sample, the characteristic data including multiple measurement points located on the detection sample and a characteristic curve of a signal light generated by each of the measurement points; an imaging module, used to obtain optical imaging data of tissue slices of the detection sample at intervals of several positions; an image segmentation module, used to perform image segmentation on the optical image of each of the tissue slices based on the optical imaging data to obtain multiple segmented images; a block label acquisition module, used to obtain a block label of each segmented image in the tissue slice; a classification label acquisition module, used to obtain a classification label corresponding to the measurement point based on the position correspondence between the segmented image and the measurement point, and the block label of the segmented image, the measurement point and the corresponding classification label and the signal light characteristic curve are used to constitute a first training data set; a model construction module, used to construct an initial detection model; a training module, used to train the initial detection model using the first training data set to obtain a trained detection model; a prediction module, used to predict the measurement point and the corresponding signal light characteristic curve using the trained detection model to obtain a corresponding classification label.

[0010] Correspondingly, an embodiment of the present invention also provides a device, comprising at least one memory and at least one processor, wherein the memory stores one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the detection method provided by the embodiment of the present invention.

[0011] Correspondingly, an embodiment of the present invention further provides a storage medium, wherein the storage medium stores one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the detection method provided by the embodiment of the present invention.

[0012] FDU-PA230002-CN

[0013] Compared with the prior art, the technical solution of the embodiment of the present invention has the following advantages:

[0014] In the detection method provided by the embodiment of the present invention, by performing image segmentation on the optical image of each tissue slice to obtain multiple segmented images, and obtaining a block label of each segmented image in the tissue slice, and based on the position correspondence between the segmented image and the measurement point and the block label of the segmented image, the classification label corresponding to the measurement point is obtained. Compared with the optical image of each tissue slice, the segmented image and the measurement point are smaller in size and more precise. Therefore, in the step of training the initial detection model using the first training data set to obtain the trained detection model, the precision of the detection model in classification judgment can be improved, which is correspondingly beneficial to improving the accuracy and precision of detection, and is particularly suitable for the detection and judgment of multifocal diseases. Moreover, by obtaining multiple measurement points on the detection sample and the signal light characteristic curves of the measurement points, it is also beneficial to improve the efficiency of obtaining feature data, and the detection model can be used to perform prediction processing based on the original data (i.e., the measurement point and the corresponding signal light characteristic curve) to obtain the corresponding classification label, and there is no need to perform fitting analysis on the signal light characteristic curve, which is beneficial to improving the detection efficiency.

[0015] The detection device provided in the embodiment of the present invention includes an optical fiber detection end connected to the light source, and the optical fiber detection end is used as a detection end for detecting the measurement point on the detection sample; wherein the detection light is irradiated onto the measurement point through the optical fiber detection end so that the detection sample at the measurement point position generates signal light, thereby providing a basis for the detection module to detect the signal light and obtain the signal light characteristic curve of the signal point, and the measurement point is measured through the optical fiber detection end. Compared with staining and optical detection of tissue sections, by detecting each measurement point to obtain the signal light characteristic curve, the process of sampling, fixing, staining and the like of the detection sample is omitted, and the signal light characteristic curve is directly detected at each measurement point on the surface of the detection sample, which is conducive to simplifying the process and improving the convenience of detection, and the time spent on staying at each measurement point for detection is less, which is also conducive to greatly improving the detection speed; and the optical fiber detection end is convenient for detecting detection samples such as cells, tissues, skin on the surface of the human body, and devices in the human body, and also improves the scope of application of the detection sample. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of a flow chart of an embodiment of a detection method of the present invention;

[0017] Figure 2 A schematic diagram showing a detection sample and obtained characteristic data;

[0018] Figure 3 A schematic diagram showing the slicing of the test sample at several intervals of positions is shown;

[0019] Figure 4 Shows Figure 1 A schematic diagram of a flow chart of an embodiment of step S4 in FIG.

[0020] FDU-PA230002-CN

[0021] Figure 5 Two examples of optical images of tissue sections are shown;

[0022] Figure 6 Schematic diagrams of performing prediction processing on the segmented image corresponding to the optical image to obtain the block labels of the corresponding segmented image are respectively shown;

[0023] Figure 7 It is a structural schematic diagram of an embodiment of a detection device of the present invention;

[0024] Figure 8 is a functional block diagram of an embodiment of a system for constructing a detection model of the present invention;

[0025] Fig. 9 It is a hardware structure diagram of an embodiment of the device provided by the present invention. DETAILED DESCRIPTION

[0026] As can be seen from the background technology, the current detection efficiency, precision and accuracy still need to be improved. The following combines several detection methods to analyze the reasons why the detection efficiency, precision and accuracy still need to be improved.

[0027] One detection method is to introduce it through an optical fiber and then introduce it to the lesion site through an endoscope to complete image acquisition, or to perform fluorescence lifetime imaging detection on tissue sections. This detection method determines the surgical margin based on the average value of the fluorescence lifetime of the image. The measured average value of the image is compared with the threshold. If it is higher or lower than the threshold, it is malignant or normal / benign. However, this method requires fluorescence microscopy imaging, the system cost is high, and the judgment speed of large areas is slow, and the judgment accuracy is not high enough.

[0028] Another detection method is to detect the autofluorescence FLIM (Fluorescence Lifetime Imaging) images of a large number of samples, and then use an unsupervised clustering algorithm to predict cancer. This detection method is easy to predict cancer in a single type of cell, but it is not applicable to complex tissue judgments or in vivo detection.

[0029] Another detection method is applied to Mohs micrographic surgery, the purpose of which is to remove all tumors while retaining as much normal tissue as possible. This method is to first remove the tumor visible to the naked eye, then remove a thin layer of tissue at the edge, map and locate it, and slice and stain the thin layer at the edge piece by piece, and then perform a microscopic examination. If all tissue slices are observed to be negative under the microscope, the operation is over; if a positive margin is observed somewhere, continue to remove thin layers of tissue layer by layer in the area where the residual cancer tissue remains, and repeat the section staining and pathological examination until complete removal. However, such an operation is very time-consuming and may take several hours or even days. There are many pathological examination points, and it is usually necessary to examine thousands or even tens of thousands of tissues. If there is a problem of too large a distance between slices, it may also cause the remaining part of the tumor to be missed, resulting in low detection efficiency and detection accuracy.

[0030] Another detection method is to use a fiber-based point scanning FLIM (Fluorescence Lifetime FDU-PA230002-CN

[0031] Imaging, fluorescence lifetime imaging) system is integrated into the da Vinci surgical system to perform intraoperative FLIM measurements for intraoperative guidance. This method first obtains data points on the patient's tissue, and then uses a supervised machine learning algorithm to predict the probability of cancer. In this way, the labels of the training set come from the experience of doctors and pathology teams, and are only applicable to the labels of large lesions. For multifocal cancerous tissues, the labels may be too rough or wrong or require a very large amount of pathological work, and are subjective and have sampling errors. If there are label errors in small-sized areas, the accuracy of the entire model will be affected.

[0032] In summary, the current detection efficiency, precision and accuracy still need to be improved.

[0033] In order to solve the technical problem, an embodiment of the present invention provides a detection method. Figure 1 It is a schematic diagram of a flow chart of an embodiment of the detection method of the present invention.

[0034] like Figure 1 As shown, in this embodiment, the detection method includes the following basic steps:

[0035] Step S1: obtaining characteristic data of a test sample, wherein the characteristic data includes a plurality of measurement points on the test sample and a characteristic curve of a signal light at each of the measurement points;

[0036] Step S2: obtaining optical imaging data of tissue slices of the test sample at several intervals;

[0037] Step S3: Based on the optical imaging data, performing image segmentation on the optical image of each tissue slice to obtain a plurality of segmented images;

[0038] Step S4: obtaining a block label of each segmented image in the tissue slice;

[0039] Step S5: based on the position correspondence between the segmented image and the measuring point, and the block label of the segmented image, obtaining the classification label corresponding to the measuring point, wherein the measuring point, the corresponding classification label and the signal light characteristic curve are used to form a first training data set;

[0040] Step S6: constructing an initial detection model;

[0041] Step S7: using the first training data set to train the initial detection model to obtain a trained detection model;

[0042] Step S8: using the trained detection model, predict and process the measurement points and the corresponding signal light characteristic curves to obtain corresponding classification labels.

[0043] In the method for constructing the detection model provided by the embodiment of the present invention, the optical image of each tissue slice is segmented to obtain a plurality of segmented images, and a block label of each segmented image in the tissue slice is obtained, and based on the position correspondence between the segmented image and the measurement point and the block label of the segmented image, the classification label corresponding to the measurement point and the FDU-PA230002-CN of each tissue slice are obtained.

[0044] Compared with the optical image, the size of the segmented image and the measurement point is smaller and the precision is higher. Therefore, in the step of using the first training data set to train the initial detection model to obtain the trained detection model, the precision of the detection model in classification judgment can be improved, which is correspondingly beneficial to improving the accuracy and precision of detection, and is particularly suitable for the detection and judgment of multifocal diseases. Moreover, by obtaining multiple measurement points on the test sample and the signal light characteristic curves of the measurement points, it is also beneficial to improve the efficiency of obtaining characteristic data, and can use the detection model to perform predictive processing based on the original data (i.e., the measurement points and the corresponding signal light characteristic curves) to obtain the corresponding classification label, without the need for fitting analysis of the signal light characteristic curve, which is beneficial to improving the detection efficiency.

[0045] In order to make the above-mentioned purposes, features and advantages of the embodiments of the present invention more obvious and understandable, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0046] refer to Figure 1 , step S1: obtaining characteristic data of a detection sample, wherein the characteristic data includes a plurality of measurement points located on the detection sample and a characteristic curve of a signal light generated by each of the measurement points.

[0047] The characteristic data of the detection sample is obtained so that the classification label corresponding to the measurement point can be obtained based on the position correspondence between the segmented image and the measurement point and the block label of the segmented image, so that the measurement point and the corresponding classification label and the signal light characteristic curve can constitute a first training data set, and the first training data set can be used to train the initial detection model.

[0048] The test sample is a sample to be tested to obtain a diagnosis result.

[0049] The detection sample is provided with a plurality of measurement points, each of which is used as a measurement position, so as to facilitate the detection of the signal light characteristic curve generated at each measurement position. Compared with staining and optical detection of tissue sections, this embodiment obtains the signal light characteristic curve by detecting each measurement point, thereby eliminating the processes of sampling, fixing, staining, etc. of the detection sample. Instead, the signal light characteristic curve is obtained by directly detecting each measurement point on the surface of the detection sample, which is conducive to simplifying the process and improving the convenience of detection. In addition, less time is spent on staying at each measurement point for detection, which is also conducive to greatly improving the detection speed.

[0050] Combined with reference Figure 2 , Figure 2 A schematic diagram of a detection sample and the characteristic data obtained is shown. As an example, Figure 2 (a) shows a schematic diagram of a test sample 10 and a plurality of measurement points 11 on the test sample 10 .

[0051] As an example, the test sample is a test sample of extramammary Paget's disease. In other embodiments, the test sample can also be a test sample that can reflect other types of cancer, such as FDU-PA230002-CN

[0052] It can be extended to other multifocal skin cancers, including basal cell carcinoma, squamous cell carcinoma, melanoma, cutaneous fibrosarcoma, etc. The detection of these cancerous changes can directly detect the lesion site through the optical fiber detection end.

[0053] In other embodiments, the test sample can also be a test sample that can reflect other types of cancer, including digestive tract cancer (oral cancer, esophageal cancer, gastric cancer, intestinal cancer, etc.), urinary system cancer (urethral cancer, bladder cancer, ureteral cancer, kidney cancer, etc.), respiratory system cancer (tracheal cancer, bronchial cancer, lung cancer, etc.), reproductive system cancer (cervical cancer, vaginal cancer, vulvar cancer, fallopian tube cancer, ovarian cancer, etc.). The detection of these cancers can be performed by an endoscope combined with an optical fiber detection end.

[0054] In this embodiment, the step of obtaining the characteristic data of the test sample includes: setting the measurement point on the test sample based on the test requirements; and testing the measurement point to obtain the characteristic data of the measurement point.

[0055] It should be noted that Figure 2 The size, position, number, density and arrangement of the measuring points shown in (a) are only used as an example. In a specific implementation, the size, position, number, density and arrangement of the measuring points are determined based on actual detection needs.

[0056] A characteristic curve of signal light generated by the detection sample at each measurement point when excited by the detection light is obtained. The characteristic curve of the signal light can reflect the characteristics of the signal light generated by different substances in the detection sample when excited by the detection light, and then the measurement information and status of different substances in the detection sample can be obtained from the characteristics of the signal light, which is convenient for predicting the block labels of the detection sample at different measurement points.

[0057] Specifically, during actual detection, detection light (for example, laser) is irradiated onto the measurement point of the detection sample through the optical fiber detection end, so that the detection sample at the measurement point generates signal light, and then the signal light generated by the detection sample at the measurement point can be detected, thereby obtaining a characteristic curve of the signal light.

[0058] In this embodiment, in the step of obtaining characteristic data of the detection sample, multi-channel characteristic data of the detection sample is obtained, and the wavelength range of the optical signal obtained by each channel is different. By obtaining multi-channel characteristic data of the detection sample, the received signal light can cover a certain spectral width, which is conducive to increasing the dimension and comprehensiveness of the detection, and correspondingly improving the accuracy of the detection.

[0059] As an example, in this embodiment, characteristic curves of signal light generated by two coenzymes, collagen, protoporphyrin IX (PpIX) and other substances in the test sample are obtained. Among them, in this embodiment, the two coenzymes are nicotinamide adenine dinucleoside (phosphate) (NAD (P) H, reduced form of nicotinamide-adenine dinucleotide (phosphate)) and flavin adenine dinucleotide (Flavin FDU-PA230002-CNAdenine Dinucleotide, FAD).

[0060] Specifically, in this embodiment, for NAD(P)H, a picosecond laser of 330nm to 410nm (for example, a 405nm laser) or a femtosecond laser of 660nm to 820nm can be provided as detection light to receive fluorescence with a wavelength of 420nm to 540nm for the signal light. For FAD, a picosecond laser of 350nm to 490nm (for example, a 488nm laser) or a femtosecond laser of 700nm to 980nm can be provided as detection light to receive fluorescence with a wavelength of 500nm to 620nm for the signal light. For collagen, a picosecond laser of 380nm to 520nm or a femtosecond laser of 760nm to 1040nm can be provided as detection light, and the wavelength of the received signal light can be selected within the range of 350-600nm according to the different wavelengths of the detection light, and the generated fluorescence or the second harmonic light signal generated by the femtosecond light excitation can be received. For PpIX, a picosecond laser of 390nm to 440nm or a femtosecond laser of 780nm to 880nm can be provided as detection light, and the wavelength band of the received signal light is fluorescence of 550nm to 680nm.

[0061] Accordingly, in this embodiment, the signal light includes fluorescence and second harmonic light signals. In other embodiments, based on the type of substance in the actual detection sample and the type of detection light, the signal light may also include other types of light signals. Accordingly, in other embodiments, the number of channels of the obtained characteristic data may also be other numbers.

[0062] That is to say, in this embodiment, in the step of obtaining characteristic data of the detection sample, multiple channels respectively acquire the fluorescence generated by NAD(P)H, the fluorescence generated by FAD, the fluorescence generated by collagen or the second harmonic light signal, and the fluorescence generated by PpIX, and obtain the characteristic curve of the signal light corresponding to the fluorescence or the second harmonic light signal.

[0063] Accordingly, in this embodiment, one of the channels is used to obtain the characteristic curve of the second harmonic light signal generated by the test sample. The second harmonic light signal can reflect the content of collagen in the test sample, and the content of collagen can reflect the condition of the pathological changes in the test sample, thereby obtaining more dimensional information for evaluating the pathological changes, and correspondingly improving the comprehensiveness and accuracy of the test.

[0064] As an example, the signal light characteristic curve includes a fluorescence lifetime decay curve or a spectrum curve.

[0065] The fluorescence lifetime decay curve contains the intensity ratio between different channels, which can reflect the content ratio of substances corresponding to different channels in the test sample, and also contains fluorescence lifetime information and fluorescence spectrum intensity. Therefore, by obtaining the fluorescence lifetime decay curve of the measurement point, multi-dimensional characteristic information of signal light can be obtained, which is conducive to improving the accuracy of judgment; in addition, by obtaining the fluorescence lifetime decay curve, there is no need to perform exponential fitting analysis to obtain fluorescence lifetime information, but it can be directly obtained from the fluorescence FDU-PA230002-CN

[0066] The fluorescence lifetime information is obtained from the light lifetime decay curve, and the original data of the fluorescence lifetime decay curve can be directly used to form the first training data set for training the initial detection model, and used as the input of the detection model, which is correspondingly beneficial to further improve the detection speed.

[0067] The spectrum curve can reflect the intensity of the received signal light, and further can reflect the substance content contained in the test sample at the position of the measurement point. Moreover, the speed of obtaining the spectrum curve is fast, which is conducive to improving the detection speed.

[0068] As an example, Figure 2 (b) and Figure 2 (c) shows the spectrum curve obtained at the measurement point 12 and the schematic diagram of the fluorescence lifetime decay curves of multiple channels obtained at the measurement point 13. As an example, in this embodiment, four channels (such as Figure 2 The fluorescence lifetime decay curves corresponding to the signal lights of CH1, CH2, CH3 and CH4 shown in (c) are used as examples for explanation. In other embodiments, the number of fluorescence lifetime decay curves obtained is also different according to the actual number of channels.

[0069] refer to Figure 2 , executing step S2: obtaining optical imaging data of tissue slices of the detection sample at several intervals of positions.

[0070] Optical imaging data of tissue slices of the test sample at several intervals are obtained, so that the optical image of the tissue slice can be segmented based on the optical imaging data to obtain multiple segmented images, and block labels of the segmented images can be obtained, and then the classification labels corresponding to the measurement points can be obtained by using the position correspondence between the segmented images and the measurement points.

[0071] Combined with reference Figure 3 , showing a schematic diagram of slicing the test sample at several intervals. Figure 3 As shown, as an example, optical imaging data of tissue slices of the detection sample at intervals of 8 positions are obtained.

[0072] In other embodiments, based on actual detection requirements, the number, positions, and intervals between tissue slices can be flexibly adjusted.

[0073] In this embodiment, the optical imaging data is taken as fluorescence lifetime imaging data as an example for explanation. In other embodiments, the optical imaging data may also be stimulated Raman scattering microscopy (SRS) data, multi-photon excitation fluorescence imaging (MPE) data, autofluorescence imaging data, histochemical staining image data or second harmonic imaging data.

[0074] Continue to refer Figure 1 , executing step S3: based on the optical imaging data, cutting each tissue FDU-PA230002-CN

[0075] The optical image of the slice is subjected to image segmentation to obtain a plurality of segmented images.

[0076] The optical image of each tissue slice is segmented to obtain multiple segmented images, so as to obtain a block label for each segmented image in the tissue slice, so that the classification label corresponding to the measurement point can be obtained based on the position correspondence between the segmented image and the measurement point.

[0077] Moreover, compared with the optical image of each tissue section, the segmented image and the measurement points are smaller in size and more precise, which is beneficial for improving the precision of the detection model in classification judgment in the subsequent step of using the first training data set to train the initial detection model to obtain a trained detection model, and correspondingly helps to improve the accuracy and precision of the detection.

[0078] In this embodiment, based on the optical imaging data, the step of performing image segmentation on the optical image of each tissue slice to obtain multiple segmented images includes: based on detection requirements, performing image segmentation on the optical image of each tissue slice to obtain multiple segmented images.

[0079] Specifically, based on the detection requirements, the number of image segmentations performed on the optical image of each tissue slice is set to obtain segmented images of corresponding sizes.

[0080] It should be noted that, in the present embodiment, in the step of obtaining characteristic data of the test sample, the measurement point is located on the surface of the test sample, and a characteristic curve of the signal light generated by the surface tissue of the test sample at the position of the measurement point is measured accordingly. Accordingly, in the present embodiment, the segmented image close to the surface of the tissue slice in the optical image of each tissue slice corresponds to the measurement point.

[0081] In a specific implementation, the optical image of each tissue slice can be segmented by OpenCV (Open Source Computer Vision Library) to obtain a plurality of segmented images. More specifically, the optical image can be equally cropped to obtain a plurality of segmented images.

[0082] Continue to refer Figure 1 , execute step S4: obtain the block label of each segmented image in the tissue section.

[0083] The block labels of the segmented image in the tissue slice are obtained, so that the classification labels corresponding to the measurement points can be obtained later based on the position correspondence between the segmented image and the measurement points and the block labels of the segmented image.

[0084] In this embodiment, the segmented image block labels are normal or malignant as an example for description. In other embodiments, the segmented image block labels may also include benign. Alternatively, in other embodiments, based on actual detection requirements and classification of detection results, the segmented image block labels may also include other types.

[0085] FDU-PA230002-CN

[0086] refer to Figure 4 , showing Figure 1 The following is a flow chart of an embodiment of step S4. Figure 4 , the specific step S4 of obtaining the block labels of each segmented image in the tissue slice in this embodiment is described in detail.

[0087] like Figure 4As shown, step S41 is performed: obtaining pathological diagnosis results of a portion of tissue sections.

[0088] Pathological diagnosis results of a portion of tissue slices are obtained, so that a block label of each segmented image in the portion of tissue slices can be obtained based on the pathological diagnosis results.

[0089] In this embodiment, pathological diagnosis results of only a portion of tissue sections are obtained, thereby reducing the number of tissue sections for pathological diagnosis, which is correspondingly beneficial to improving the detection rate.

[0090] Specifically, in this embodiment, the step of obtaining the pathological diagnosis results of a portion of the tissue sections may include: staining and performing pathological examination on the portion of the tissue sections, and having a pathologist make a judgment to obtain the pathological diagnosis results.

[0091] In a specific implementation, the partial number of tissue sections can be subjected to hematoxylin and eosin stain (H&E), followed by a pathological examination and judgment by a pathologist to obtain a pathological diagnosis result.

[0092] like Figure 4 As shown, step S42: based on the pathological diagnosis result, label each segmented image in the partial number of tissue slices to obtain block labels of the segmented image, and the segmented images in the partial number of tissue slices and the corresponding block labels are used to form a second training data set.

[0093] The second training data set is used for subsequent training of the classification model.

[0094] Specifically, based on the pathological diagnosis result of the tissue section, each segmented image of the current tissue section is annotated with a corresponding block label. For example, if the pathological diagnosis result of the current tissue section is malignant, each segmented image of the current tissue section is annotated as malignant; if the pathological diagnosis result of the current tissue section is normal, each segmented image of the current tissue section is annotated as normal. In other words, the segmentation label corresponding to each segmented image of the current tissue section is consistent with the pathological diagnosis result of the tissue section.

[0095] Combined with reference Figure 5 , two examples of optical images 100 of tissue slices are shown. The optical image 110 shows a tissue slice whose pathological diagnosis result is malignant, and the corresponding block labels of the segmented image 130 are all malignant; the optical image 120 shows a tissue slice whose pathological diagnosis result is normal, and the corresponding block labels of the segmented image 130 are all normal.

[0096] like Figure 4As shown, step S43: construct a classification model. Construct a classification model so that the classification model can be used later.

[0097] The second training data set is used to train the classification model.

[0098] As an example, the classification model includes a residual neural network (ResNet) model. The residual neural network sets a "shortcut connection" for degradation, which greatly eliminates the difficulty of training a neural network with too much depth. As an example, the ResNet model is a ResNet 50 model.

[0099] In other embodiments, based on actual needs, the classification model may also be other types of classification models.

[0100] It should be noted that, continue to refer to Figure 4 As an example, step S4 of obtaining the block label of each segmented image in the tissue slice also includes: after obtaining the block label of each segmented image in the partial number of tissue slices based on the pathological diagnosis result, and before using the second training data set to train the classification model, step S46: performing image recognition processing on the segmented images in the second training data set to obtain image recognition results; step S47: based on the image recognition results, removing the segmented images whose background area accounts for more than a preset value from the second training data set.

[0101] Image recognition processing is performed on the segmented images in the second training data set, and based on the image recognition results, the segmented images in which the background area accounts for a greater proportion than a preset value are removed from the second training data set, thereby preventing the segmented images in which the background area accounts for too large a proportion from being mixed in the second training data set, which in turn helps to prevent interference with the training processing of subsequent classification models, and correspondingly helps to reduce the probability of the subsequent trained classification model predicting a false positive or false negative.

[0102] like Figure 4 As shown, execute step S44: use the second training data set to train the classification model to obtain a trained classification model.

[0103] The classification model is trained using the second training data set so that the classification model can analyze and learn the relationship between the segmented image and the block labels, and then the classification model can be used to predict the segmented images of the remaining number of tissue slices to obtain corresponding block labels.

[0104] Combined with reference Figure 5In this embodiment, only tissue slices 110 with pathological diagnosis results of malignant and tissue slices 120 with pathological diagnosis results of normal are used to form the second training data set, and the training process of the classification model is explained as an example. In other embodiments, based on actual detection needs, tissue slices with pathological diagnosis results of malignant, normal and benign can also be used to form the second training data set to train the classification model. FDU-PA230002-CN

[0105] like Figure 4 As shown, step S45 is executed: after obtaining the trained classification model, the classification model is used to perform prediction processing on the segmented images in the remaining number of tissue slices to obtain block labels of the segmented images in the remaining number of tissue slices.

[0106] The classification model is used to predict the segmented images in the remaining number of tissue slices to obtain corresponding block labels. Compared with the optical images of the tissue slices, the size of the segmented images is smaller, which is conducive to obtaining more precise labels. In addition, the classification model is used to predict the segmented images of the remaining number of tissue slices, which not only helps to improve the accuracy of the block labels of the segmented images, but also saves a lot of labeling time and improves the rate of labeling the segmented images.

[0107] like Figure 6 As shown in FIG. 1 , the segmented images corresponding to the optical images 100a, 100b, and 100c are respectively predicted and processed to obtain the block labels of the corresponding segmented image 130. The optical image 100a shows a tissue slice of all malignant tissues, the optical image 100b shows a tissue slice of a mixed area, and the optical image 100c shows a tissue slice of all normal tissues. Figure 6 In the segmented image, the color depth of the segmented image is used to distinguish between malignant and normal.

[0108] Among them, all benign, all malignant or all normal respectively refer to that the block labels of all segmented images of the current tissue slice are benign, malignant or normal.

[0109] It should be noted that, in this embodiment, the method of obtaining the block labels of the segmented images in the remaining number of tissue slices by training the classification model is used as an example for explanation. The method of obtaining the block labels of the segmented images in the remaining number of tissue slices is not limited to this.

[0110] In other embodiments, the step of obtaining the block labels of the segmented images in the remaining number of tissue slices may further include: based on the pathological diagnosis results of the remaining number of tissue slices, labeling each segmented image in the remaining number of tissue slices with a corresponding block label. Specifically, the segmentation label of each segmented image is consistent with the pathological diagnosis result of the corresponding tissue slice.

[0111] It should also be noted that, in some other embodiments, the step of obtaining the block labels of each segmented image in the tissue slice may also include: after obtaining the block labels of the segmented images in the remaining number of tissue slices, performing cleaning and correction processing on the second training data set; after performing cleaning and correction processing on the second training data set, retraining the classification model using the second training data set; and performing prediction processing on the segmented images using the retrained classification model to obtain the block labels of each segmented image.

[0112] Clean and correct the second training data set, and then retrain the classification model. FDU-PA230002-CN

[0113] This improves the accuracy of the training samples, avoids the interference of erroneous samples on the model training as much as possible, and further improves the reliability and robustness of the classification model obtained by training, and correspondingly improves the accuracy of the block labels obtained by the prediction processing.

[0114] In a specific implementation, the second training data set may be cleaned and corrected based on cross-validation combined with confidence learning, fidelity weighted learning, or sampling detection.

[0115] In some embodiments, based on cross-validation combined with confidence learning, the step of cleaning and correcting the second training data set may include: obtaining the joint distribution of the pathological diagnosis results and the block labels of the segmented image in the second training data set; based on the joint distribution, according to the mean of the probability of each category, deleting from the second training data set samples that are lower than the current category mean and whose probability of another category is higher than the mean of another category.

[0116] By obtaining the joint distribution of the pathological diagnosis results and the block labels of the segmented images in the second training data set, and based on the joint distribution, based on the joint distribution, according to the mean of the probability of each category, the samples whose probability is lower than the current category mean and whose probability of another category is higher than the mean of another category are deleted from the second training data set, thereby achieving the purpose of data cleaning, and readjusting the category weights in the second training data set, and then retraining the classification model, thereby improving the accuracy of the classification model.

[0117] For example, in a specific implementation, based on the joint distribution of pathological diagnosis results and block labels, samples with a probability lower than the normal class mean and a probability higher than the malignant class mean are deleted from the second training data set according to the mean of the normal and malignant class probabilities.

[0118] In some embodiments, based on fidelity weighted learning, the step of cleaning and correcting the second training data set may include: using the classification model as the first classification model; training a student network using the optical image of the tissue section and the corresponding pathological diagnosis result; training a teacher network using the segmented image and the corresponding block label; and generating soft labels for all the segmented images and the corresponding block labels using the teacher network to obtain a soft data set.

[0119] Correspondingly, the soft data set is used to perform adversarial training on the student network to obtain a second classification model; and the second classification model is used to predict all segmented images to obtain block labels of the segmented images in all tissue sections.

[0120] By training the student network and the teacher network, and using the teacher network to obtain a soft data set, and using the soft data set to perform adversarial training on the student network, an FDU-PA230002-CN that approximates the first classification model is obtained.

[0121] The second classification model can also retain the noise data in the optical image and the corresponding pathological diagnosis result during the adversarial training process, and can perform fidelity weighted learning accordingly, thereby improving the accuracy and prediction authenticity of the second classification model.

[0122] In some embodiments, based on sampling detection, the step of cleaning and correcting the second training data set may include: dividing the second training data set into multiple sub-data, and using a classification model to perform sampling prediction on each sub-data; after sampling prediction on each sub-data, cleaning the second training data set to re-label or delete from the second training data set data whose prediction results are inconsistent with the classification results obtained based on the pathological diagnosis results; retraining the classification model using the cleaned second training data set; after retraining the classification model, using the classification model to perform prediction processing on the segmented images in the remaining number of tissue slices to obtain block labels for the segmented images in the remaining number of tissue slices.

[0123] By cleaning the second training data set, the data that are inconsistent with the prediction results and the block labels obtained based on the pathological diagnosis results are re-labeled or deleted from the second training data set, and the classification model is retrained, thereby preventing erroneous sample data from interfering with model training, and correspondingly improving the accuracy of classification model prediction.

[0124] It should be noted that, in this embodiment, after obtaining the block labels of the segmented image corresponding to the optical image of each tissue slice, based on the proportion of different block labels of the segmented image in each tissue slice, the positive rate can be given for pictures of a large size range. As an example, the accuracy of the classification model finally trained can reach 95%, and the positive rate can be given for slices of malignant areas, mixed areas, and normal areas for pictures of a large size range. The positive rate of the malignant area is close to 100%, the positive rate of the normal area is 0%, and the positive rate of the mixed area varies due to different specific cancer types. Taking extramammary Paget's disease as an example, the positive rate of the mixed area measured in this embodiment is about 30% to 60%. Therefore, in this embodiment, the prediction accuracy of the classification model is high.

[0125] Continue to refer Figure 1 , execute step S5: based on the position correspondence between the segmented image and the measurement point, and the block label of the segmented image, obtain the classification label corresponding to the measurement point, and the measurement point and the corresponding classification label and the signal light characteristic curve are used to form a first training data set.

[0126] The classification labels corresponding to the measurement points are obtained so as to construct the first training data set, thereby facilitating subsequent training processing of the initial detection model using the first training data set.

[0127] Compared with the optical image of each tissue slice, the size of the segmented image and the measurement point is smaller and more precise.

[0128] Higher, so that in the step of training the initial detection model with the first training data set to obtain the trained detection model, the precision of the detection model in classification judgment can be improved, which is correspondingly beneficial to improving the accuracy and precision of detection, and is particularly suitable for the detection and judgment of multifocal diseases. Moreover, by obtaining multiple measurement points on the test sample and the signal light characteristic curves of the measurement points, it is also beneficial to improve the efficiency of obtaining characteristic data, and can obtain the corresponding classification label based on the original data (i.e., the measurement points and the corresponding signal light characteristic curves) using the detection model, without the need for fitting analysis of the signal light characteristic curves, which is beneficial to improving the detection efficiency.

[0129] In this embodiment, the segmented image close to the surface of the tissue slice in the optical image of each tissue slice corresponds to the measurement point, and accordingly, the block label of the segmented image close to the surface of the tissue slice is used as the classification label of the measurement point at the corresponding position.

[0130] Continue to refer Figure 1 , execute step S6: construct an initial detection model, so as to subsequently use the first training data set to train the initial detection model to obtain a trained detection model.

[0131] As an example, the step of constructing the initial detection model includes: constructing the initial detection model based on logistic regression, decision tree, random forest or support vector machines (SVM).

[0132] In this embodiment, the construction of the initial detection model based on the support vector machine is used as an example for explanation. The support vector machine is a novel small sample learning method with a solid theoretical basis, which basically does not involve probability measurement and the law of large numbers, and its final decision function is determined by only a few support vectors. The complexity of the calculation depends on the number of support vectors, rather than the dimension of the sample space, which in a sense avoids the "dimensionality disaster".

[0133] In other embodiments, the initial detection model may also be constructed based on other suitable classification models.

[0134] It should be noted that, in this embodiment, the method further includes: after obtaining the classification label corresponding to the measurement point based on the positional correspondence between the segmented image and the measurement point, and the block label of the segmented image, and before using the first training data set to train the initial detection model, executing step S9: performing data expansion processing on the first training data set.

[0135] The first training data set is subjected to data expansion processing to increase the sample size.

[0136] As an example, the step of performing data expansion processing on the first training data set includes any one or more of the following situations: linearly superimposing several signal light characteristic curves of the same category using random proportional coefficients whose sum is 1; randomly translating FDU-PA230002-CN along the horizontal axis within a limited spectral line displacement fluctuation range;

[0137] Signal light characteristic curve; adding random Gaussian white noise of varying intensities to the signal light characteristic curve.

[0138] Continue to refer Figure 1, execute step S7: use the first training data set to train the initial detection model to obtain a trained detection model.

[0139] In this embodiment, by performing image segmentation on the optical image of each of the tissue slices to obtain a plurality of segmented images, and obtaining a block label for each segmented image in the tissue slice, and based on the positional correspondence between the segmented image and the measurement point and the block label of the segmented image, a classification label corresponding to the measurement point is obtained. Compared with the optical image of each tissue slice, the segmented image and the measurement point are smaller in size and more precise. Therefore, in the step of training the initial detection model using the first training data set to obtain a trained detection model, the precision of the detection model in classification judgment can be improved, which is correspondingly beneficial to improving the accuracy and precision of detection, and is particularly suitable for the detection and judgment of multifocal diseases. Moreover, by obtaining a plurality of measurement points on the detection sample and the signal light characteristic curves of the measurement points, it is also beneficial to improve the efficiency of obtaining feature data, and the detection model can be used to perform prediction processing based on the original data (i.e., the measurement point and the corresponding signal light characteristic curve) to obtain the corresponding classification label, and there is no need to perform fitting analysis on the signal light characteristic curve, which is beneficial to improving the detection efficiency.

[0140] Continue to refer Figure 1 , execute step S8: use the trained detection model to predict the characteristic curves of the measurement points and signal light to obtain corresponding classification labels.

[0141] By using the trained detection model, the characteristic curves of the measurement points and the signal light are predicted to obtain corresponding classification labels, thereby improving the accuracy and precision of the detection. This is particularly suitable for the detection and judgment of multifocal diseases. Moreover, by obtaining multiple measurement points on the test sample and the signal light characteristic curves of the measurement points, it is also beneficial to improve the efficiency of obtaining characteristic data, and the detection model can be used to obtain the corresponding classification labels based on the original data (i.e., the measurement points and the corresponding signal light characteristic curves), without the need to perform fitting analysis on the signal light characteristic curves, which is beneficial to improving the detection efficiency.

[0142] In addition, in this embodiment, the classification labels corresponding to the measurement points are first obtained, and the measurement points and the corresponding classification labels and the signal light characteristic curves constitute a first training data set, and the initial detection model is trained using the first training data set to obtain a trained detection model. That is, this embodiment is a supervised clustering algorithm, and the number of clusters is relatively clear (2 or 3), so that the detection model can easily achieve high accuracy, thereby facilitating the detection of complex tissues.

[0143] In other embodiments, the detection method may further include: obtaining the positive rate of the detection sample based on the classification labels of all measurement points on the same detection sample.

[0144] The proportion of measured points with different classification labels is used to obtain the positive rate of the detected samples.

[0145] Correspondingly, the present invention also provides a detection device. Figure 7 It is a schematic structural diagram of an embodiment of a detection device of the present invention.

[0146] In this embodiment, the detection device is used to obtain the characteristic data of the detection sample in the detection method of the above-mentioned embodiment.

[0147] In this embodiment, the detection device includes: a light source for providing detection light; an optical fiber detection end 300, connected to the light source, and used as a detection end for detecting a measurement point on the detection sample; wherein the detection light is irradiated onto the measurement point through the optical fiber detection end 300 so that the detection sample at the measurement point generates signal light; a detection module 310, connected to the optical fiber detection end 300, and used to detect the signal light generated by the detection sample at the measurement point, and obtain a signal light characteristic curve for each measurement point.

[0148] The detection device includes an optical fiber detection end 300 connected to the light source, and the optical fiber detection end 300 is used as a detection end for detecting the measurement point on the detection sample; wherein the detection light is irradiated onto the measurement point through the optical fiber detection end so that the detection sample at the measurement point generates signal light, thereby providing a basis for the detection module 310 to detect the signal light and obtain the signal light characteristic curve of the signal point, and the measurement point is measured through the optical fiber detection end. Compared with staining and optical detection of tissue sections, by detecting each measurement point to obtain the signal light characteristic curve, the process of sampling, fixing, staining and the like of the detection sample is omitted, and the signal light characteristic curve is directly detected at each measurement point on the surface of the detection sample, which is conducive to simplifying the process and improving the convenience of detection, and the time spent on staying at each measurement point for detection is less, which is also conducive to greatly improving the detection speed; and the optical fiber detection end 300 is convenient for detecting detection samples such as cells, tissues, skin on the surface of the human body, and devices in the human body, and also improves the scope of application of the detection sample.

[0149] In one embodiment, the detection module 311 includes: multiple groups of detection units (not shown), each group of detection units detects a different wavelength range of optical signals; each group of detection units includes a filter and a detector connected to each other, and the filter is connected to the optical fiber detection end.

[0150] The detection module 311 includes multiple detection units, and each detection unit detects a different wavelength range of the optical signal, thereby obtaining multi-channel characteristic data. More specifically, in this embodiment, each detection unit is used to obtain a signal light characteristic curve of the current channel.

[0151] In this embodiment, the filter is used to filter the optical signals of different channels so that each channel can obtain FDU-PA230002-CN

[0152] The detector is used to collect data of optical signals from different channels.

[0153] In this embodiment, by providing a plurality of detection units, the characteristic data of the channels can be detected simultaneously, which is beneficial to improving the detection efficiency.

[0154] In another embodiment, the detection module 312 includes: a filter turntable connected to the optical fiber detection end, and a plurality of filters are arranged on the filter turntable; and a detector connected to the filter turntable.

[0155] By setting the filter turntable, different filters on the filter turntable are selected to match the detector, thereby obtaining multi-channel characteristic data.

[0156] In yet another embodiment, the detection module 313 includes: a spectrometer connected to the optical fiber detection end, for obtaining a spectrum curve.

[0157] Correspondingly, the present invention also provides a system for constructing a detection model. Figure 8 It is a functional block diagram of an embodiment of a system for constructing a detection model of the present invention.

[0158] refer to Figure 8In this embodiment, the detection model construction system 400 includes: a data acquisition module 401, used to obtain characteristic data of the detection sample, the characteristic data including multiple measurement points located on the detection sample and characteristic curves of signal light generated by each of the measurement points; an imaging module 402, used to obtain optical imaging data of tissue slices of the detection sample at intervals of several positions; an image segmentation module 403, used to perform image segmentation on the optical image of each of the tissue slices based on the optical imaging data to obtain multiple segmented images; a block label acquisition module 404, used to obtain a block label for each segmented image in the tissue slice; a classification label acquisition module 405, used to obtain a classification label for each segmented image in the tissue slice; Module 405 is used to obtain the classification label corresponding to the measurement point based on the position correspondence between the segmented image and the measurement point, and the block label of the segmented image, and the measurement point, the corresponding classification label and the signal light characteristic curve are used to constitute a first training data set; model construction module 406 is used to build an initial detection model; training module 407 is used to train the initial detection model with the first training data set to obtain a trained detection model; prediction module 408 is used to predict the measurement point and the corresponding signal light characteristic curve with the trained detection model to obtain the corresponding classification label.

[0159] The data acquisition module 401 obtains the characteristic data of the detection sample so that the classification label corresponding to the measurement point can be obtained based on the position correspondence between the segmented image and the measurement point and the block label of the segmented image, so that the measurement point and the corresponding classification label and the signal light characteristic curve can constitute a first training data set, and then the first training data set can be used to train the initial detection model.

[0160] The test sample is a sample to be tested to obtain a diagnosis result.

[0161] FDU-PA230002-CN

[0162] The detection sample is provided with a plurality of measurement points, each of which is used as a measurement position, so as to facilitate the detection of the signal light characteristic curve generated at each measurement position. Compared with staining and optical detection of tissue sections, this embodiment obtains the signal light characteristic curve by detecting each measurement point, thereby eliminating the processes of sampling, fixing, staining, etc. of the detection sample. Instead, the signal light characteristic curve is obtained by directly detecting each measurement point on the surface of the detection sample, which is conducive to simplifying the process and improving the convenience of detection. In addition, less time is spent on staying at each measurement point for detection, which is also conducive to greatly improving the detection speed.

[0163] As an example, Figure 2(a) shows a schematic diagram of a test sample 10 and a plurality of measurement points 11 on the test sample 10 .

[0164] As an example, the detection sample is an extramammary Paget's disease detection sample. In other embodiments, the detection sample can also be a detection sample that can reflect other types of canceration, for example, it can be extended to other multifocal skin cancers, including: basal cell carcinoma, squamous cell carcinoma, melanoma, cutaneous fibrosarcoma, etc. The detection of these cancerations can directly detect the lesion site through the optical fiber detection end.

[0165] In other embodiments, the test sample can also be a test sample that can reflect other types of cancer, including digestive tract cancer (oral cancer, esophageal cancer, gastric cancer, intestinal cancer, etc.), urinary system cancer (urethral cancer, bladder cancer, ureteral cancer, kidney cancer, etc.), respiratory system cancer (tracheal cancer, bronchial cancer, lung cancer, etc.), reproductive system cancer (cervical cancer, vaginal cancer, vulvar cancer, fallopian tube cancer, ovarian cancer, etc.). The detection of these cancers can be performed by an endoscope combined with an optical fiber detection end.

[0166] In this embodiment, obtaining the characteristic data of the test sample includes: setting the measurement point on the test sample based on the test requirements; and testing the measurement point to obtain the characteristic data of the measurement point.

[0167] It should be noted that Figure 2 The size, position, number, density and arrangement of the measuring points shown in (a) are only used as an example. In a specific implementation, the size, position, number, density and arrangement of the measuring points are determined based on actual detection needs.

[0168] A characteristic curve of signal light generated by the detection sample at each measurement point when excited by the detection light is obtained. The characteristic curve of the signal light can reflect the characteristics of the signal light generated by different substances in the detection sample when excited by the detection light, and then the measurement information and status of different substances in the detection sample can be obtained from the characteristics of the signal light, which is convenient for predicting the block labels of the detection sample at different measurement points.

[0169] Specifically, during actual detection, the detection light (for example, laser) is irradiated to the measurement point of the detection sample through the optical fiber detection end, so that the detection sample at the measurement point generates signal light, and then FDU-PA230002-CN

[0170] The signal light generated by the detection sample at the measurement point position can be detected to obtain a characteristic curve of the signal light.

[0171] In this embodiment, in the step of obtaining characteristic data of the detection sample, multi-channel characteristic data of the detection sample is obtained, and the wavelength range of the optical signal obtained by each channel is different. By obtaining multi-channel characteristic data of the detection sample, the received signal light can cover a certain spectral width, which is conducive to increasing the dimension and comprehensiveness of the detection, and correspondingly improving the accuracy of the detection.

[0172] As an example, in this embodiment, characteristic curves of signal light generated by two coenzymes, collagen, protoporphyrin IX (PpIX) and other substances in the test sample are obtained. In this embodiment, the two coenzymes are nicotinamide adenine dinucleotide (phosphate) (NAD (P) H, reduced form of nicotinamide-adenine dinucleotide (phosphate)) and flavin adenine dinucleotide (Flavin Adenine Dinucleotide, FAD).

[0173] Specifically, in this embodiment, for NAD(P)H, a picosecond laser of 330nm to 410nm (for example, a 405nm laser) or a femtosecond laser of 660nm to 820nm can be provided as detection light to receive fluorescence with a wavelength of 420nm to 540nm for the signal light. For FAD, a picosecond laser of 350nm to 490nm (for example, a 488nm laser) or a femtosecond laser of 700nm to 980nm can be provided as detection light to receive fluorescence with a wavelength of 500nm to 620nm for the signal light. For collagen, a picosecond laser of 380nm to 520nm or a femtosecond laser of 760nm to 1040nm can be provided as detection light, and the wavelength of the received signal light can be selected within the range of 350-600nm according to the different wavelengths of the detection light, and the generated fluorescence or the second harmonic light signal generated by the femtosecond light excitation can be received. For PpIX, a picosecond laser of 390nm to 440nm or a femtosecond laser of 780nm to 880nm can be provided as detection light, and the wavelength band of the received signal light is fluorescence of 550nm to 680nm.

[0174] Accordingly, in this embodiment, the signal light includes fluorescence and second harmonic light signals. In other embodiments, based on the type of substance in the actual detection sample and the type of detection light, the signal light may also include other types of light signals. Accordingly, in other embodiments, the number of channels of the obtained characteristic data may also be other numbers.

[0175] That is to say, in this embodiment, in the step of obtaining characteristic data of the detection sample, multiple channels respectively acquire the fluorescence generated by NAD(P)H, the fluorescence generated by FAD, the fluorescence generated by collagen or the second harmonic light signal, and the fluorescence generated by PpIX, and obtain the characteristic curve of the signal light corresponding to the fluorescence or the second harmonic light signal.

[0176] FDU-PA230002-CN

[0177] Accordingly, in this embodiment, one of the channels is used to obtain the characteristic curve of the second harmonic light signal generated by the test sample. The second harmonic light signal can reflect the content of collagen in the test sample, and the content of collagen can reflect the condition of the pathological changes in the test sample, thereby obtaining more dimensional information for evaluating the pathological changes, and correspondingly improving the comprehensiveness and accuracy of the test.

[0178] As an example, the signal light characteristic curve includes a fluorescence lifetime decay curve or a spectrum curve.

[0179] Among them, the fluorescence lifetime decay curve contains the intensity ratio between different channels, which can reflect the content ratio of substances corresponding to different channels in the test sample, and also contains fluorescence lifetime information and fluorescence spectrum intensity. Therefore, by obtaining the fluorescence lifetime decay curve of the measuring point, multi-dimensional characteristic information of the signal light can be obtained, which is beneficial to improve the accuracy of judgment; in addition, by obtaining the fluorescence lifetime decay curve, there is no need to perform exponential fitting analysis to obtain the fluorescence lifetime information, but the fluorescence lifetime information can be obtained directly from the fluorescence lifetime decay curve, and then the original data of the fluorescence lifetime decay curve can be directly used to form the first training data set for training the initial detection model, and used as the input of the detection model, which is correspondingly beneficial to further improve the detection speed.

[0180] The spectral curve can reflect the intensity of the received signal light, and further can reflect the substance content contained in the test sample at the position of the measurement point. Moreover, the speed of obtaining the spectral curve is fast, which is also conducive to improving the detection speed.

[0181] As an example, Figure 2 (b) and Figure 2 (c) shows the spectrum curve obtained at the measurement point 12 and the schematic diagram of the fluorescence lifetime decay curves of multiple channels obtained at the measurement point 13. As an example, to obtain four channels (such as Figure 2 The fluorescence lifetime decay curves corresponding to the signal lights of CH1, CH2, CH3 and CH4 shown in (c) are used as examples for explanation. In other embodiments, the number of fluorescence lifetime decay curves obtained is also different according to the actual number of channels.

[0182] In this embodiment, when the signal light characteristic curve is a fluorescence lifetime decay curve, obtaining characteristic data of the detection sample includes: obtaining the original fluorescence lifetime decay curve corresponding to each measurement point; and preprocessing the original fluorescence lifetime decay curve to remove the curve before the peak of the fluorescence lifetime decay curve.

[0183] The imaging module 402 is used to obtain optical imaging data of tissue slices of the test sample at several intervals, so as to subsequently perform image segmentation on the optical image of the tissue slice based on the optical imaging data to obtain multiple segmented images, and obtain block labels of the segmented images, and then use the position correspondence between the segmented images and the measurement points to obtain the classification labels corresponding to the measurement points.

[0184] FDU-PA230002-CN

[0185] Combined with reference Figure 3 , showing a schematic diagram of slicing the test sample at several intervals. Figure 3 As shown, as an example, optical imaging data of tissue slices of the detection sample at intervals of 8 positions are obtained.

[0186] In other embodiments, based on actual detection requirements, the number, positions, and intervals between tissue slices can be flexibly adjusted.

[0187] In this embodiment, the optical imaging data is taken as fluorescence lifetime imaging data as an example for explanation. In other embodiments, the optical imaging data may also be stimulated Raman scattering microscopy (SRS) data, multi-photon excitation fluorescence imaging (MPE) data, autofluorescence imaging data, histochemical staining image data or second harmonic imaging data.

[0188] The image segmentation module 403 is used to perform image segmentation on the optical image of each tissue slice based on the optical imaging data to obtain a plurality of segmented images.

[0189] The optical image of each tissue slice is segmented to obtain multiple segmented images, so as to obtain a block label for each segmented image in the tissue slice, so that the classification label corresponding to the measurement point can be obtained based on the position correspondence between the segmented image and the measurement point.

[0190] Moreover, compared with the optical image of each tissue section, the segmented image and the measurement points are smaller in size and more precise, which is beneficial in the subsequent step of using the first training data set to train the initial detection model to obtain a trained detection model, and can improve the precision of the detection model in classification judgment, which is correspondingly beneficial to improving the accuracy and precision of the detection.

[0191] In this embodiment, based on the optical imaging data, the step of performing image segmentation on the optical image of each tissue slice to obtain multiple segmented images includes: based on detection requirements, performing image segmentation on the optical image of each tissue slice to obtain multiple segmented images.

[0192] Specifically, based on the detection requirements, the number of image segmentations performed on the optical image of each tissue slice is set to obtain segmented images of corresponding sizes.

[0193] It should be noted that, in this embodiment, in the step of obtaining the characteristic data of the test sample, the measurement point is located on the surface of the test sample, and the characteristic curve of the signal light generated by the surface tissue of the test sample at the position of the measurement point is measured accordingly. Accordingly, in this embodiment, the segmented image close to the surface of the tissue slice in the optical image of each tissue slice corresponds to the measurement point.

[0194] In a specific implementation, the optical image of each tissue slice can be segmented by OpenCV (Open Source Computer Vision Library) to obtain a plurality of segmented images. More specifically, the optical image can be equally cropped to obtain a plurality of segmented images.

[0195] The block label acquisition module 404 is used to obtain the block label of each segmented image in the tissue slice, so as to obtain the classification label corresponding to the measurement point based on the position correspondence between the segmented image and the measurement point and the block label of the segmented image.

[0196] In this embodiment, the segmented image block labels are normal or malignant as an example for description. In other embodiments, the segmented image block labels may also include benign. Alternatively, in other embodiments, based on actual detection requirements and classification of detection results, the segmented image block labels may also include other types.

[0197] As an example, the block label acquisition module 404 includes: a pathological diagnosis unit (not shown in the figure), which is used to obtain the pathological diagnosis results of a partial number of tissue slices; a data set acquisition unit (not shown in the figure), which is used to label each segmented image in the partial number of tissue slices based on the pathological diagnosis results to obtain the block labels of the segmented images, and the segmented images and corresponding block labels in the partial number of tissue slices are used to constitute a second training data set; a classification model construction unit (not shown in the figure), which is used to construct a classification model; a classification model training unit (not shown in the figure), which is used to train the classification model using the second training data set to obtain a trained classification model; a prediction unit (not shown in the figure), which is used to predict the segmented images in the remaining number of tissue slices using the classification model to obtain the block labels of the segmented images in the remaining number of tissue slices.

[0198] The pathological diagnosis unit obtains the pathological diagnosis results of a portion of the tissue slices, so as to subsequently obtain the block labels of each segmented image in the portion of the tissue slices based on the pathological diagnosis results.

[0199] In this embodiment, pathological diagnosis results of only a portion of tissue sections are obtained, thereby reducing the number of tissue sections for pathological diagnosis, which is correspondingly beneficial to improving the detection rate.

[0200] Specifically, in this embodiment, the step of obtaining the pathological diagnosis results of a portion of the tissue sections may include: staining and performing pathological examination on the portion of the tissue sections, and having a pathologist make a judgment to obtain the pathological diagnosis results.

[0201] In a specific implementation, the partial number of tissue sections can be subjected to hematoxylin and eosin stain (H&E), followed by a pathological examination and judgment by a pathologist to obtain a pathological diagnosis result.

[0202] FDU-PA230002-CN

[0203] The data set acquisition unit obtains the block labels of each segmented image in a certain number of tissue slices.

[0204] Specifically, based on the pathological diagnosis result of the tissue section, each segmented image of the current tissue section is annotated with a corresponding block label. For example, if the pathological diagnosis result of the current tissue section is malignant, each segmented image of the current tissue section is annotated as malignant; if the pathological diagnosis result of the current tissue section is normal, each segmented image of the current tissue section is annotated as normal. In other words, the segmentation label corresponding to each segmented image of the current tissue section is consistent with the pathological diagnosis result of the tissue section.

[0205] Combined with reference Figure 5 , two examples of optical images 100 of tissue slices are shown. The optical image 110 shows a tissue slice whose pathological diagnosis result is malignant, and the corresponding block labels of the segmented image 130 are all malignant; the optical image 120 shows a tissue slice whose pathological diagnosis result is normal, and the corresponding block labels of the segmented image 130 are all normal.

[0206] The classification model building unit builds a classification model so that the classification model training unit can use the second training data set to train the classification model.

[0207] As an example, the classification model includes a residual neural network (ResNet) model. The residual neural network sets a "shortcut connection" for degradation, which greatly eliminates the difficulty of training a neural network with too much depth. As an example, the ResNet model is a ResNet 50 model.

[0208] In other embodiments, based on actual needs, the classification model may also be other types of classification models.

[0209] As an example, the block label acquisition module 404 may also include: an identification unit (not shown), which is used to perform image recognition processing on the segmented images in the second training data set to obtain image recognition results after obtaining the block labels of each segmented image in the partial number of tissue sections based on the pathological diagnosis results, and before using the second training data set to train the classification model, and remove the segmented images in which the background area accounts for more than a preset value from the second training data set based on the image recognition results.

[0210] Image recognition processing is performed on the segmented images in the second training data set, and based on the image recognition results, the segmented images in which the background area accounts for a greater proportion than a preset value are removed from the second training data set, thereby preventing the segmented images in which the background area accounts for too large a proportion from being mixed in the second training data set, which in turn helps to prevent interference with the training processing of subsequent classification models, and correspondingly helps to reduce the probability of the subsequent trained classification model predicting a false positive or false negative.

[0211] FDU-PA230002-CN

[0212] The classification model training unit uses the second training data set to train the classification model so that the classification model can analyze and learn the relationship between the segmented image and the block label, and then can subsequently use the classification model to predict the segmented images of the remaining number of tissue slices to obtain the corresponding block labels.

[0213] Combined with reference Figure 5 In this embodiment, only tissue slices 110 with a pathological diagnosis result of malignant and tissue slices 120 with a pathological diagnosis result of normal are used to form a second training data set, and the training process of the classification model is described as an example. In other embodiments, based on actual detection requirements, tissue slices with a pathological diagnosis result of malignant, tissue slices with a pathological diagnosis result of normal, and tissue slices with a pathological diagnosis result of benign can also be used to form a second training data set to train the classification model.

[0214] The prediction unit is used to use the classification model to perform prediction processing on the segmented images in the remaining number of tissue slices to obtain corresponding block labels. Compared with the optical image of the tissue slice, the size of the segmented image is smaller, which is conducive to obtaining more precise labels. Moreover, the prediction processing of the segmented images of the remaining number of tissue slices by the classification model is not only conducive to improving the accuracy of obtaining the block labels of the segmented images, but also saves a lot of annotation time, thereby improving the rate of obtaining the annotation of the segmented images.

[0215] like Figure 6 As shown in FIG. 1 , the segmented images corresponding to the optical images 100a, 100b, and 100c are respectively predicted and processed to obtain the block labels of the corresponding segmented image 130. The optical image 100a shows a tissue slice of all malignant tissues, the optical image 100b shows a tissue slice of a mixed area, and the optical image 100c shows a tissue slice of all normal tissues. Figure 6 In the segmented image, the color depth of the segmented image is used to distinguish between malignant and normal.

[0216] Among them, all benign, all malignant or all normal means that the block labels of all segmented images of the current tissue slice are benign, malignant or normal.

[0217] It should be noted that in this embodiment, the classification model is trained by the classification model training unit, and the prediction unit is used to obtain the block labels of the segmented images in the remaining number of tissue slices. The method of obtaining the block labels of the segmented images in the remaining number of tissue slices is not limited to this.

[0218] In other embodiments, each segmented image in the remaining number of tissue slices may be labeled with a corresponding block label based on the pathological diagnosis result of the remaining number of tissue slices. Specifically, the segmentation label of each segmented image is consistent with the pathological diagnosis result of the corresponding tissue slice.

[0219] It should also be noted that, in some other embodiments, the block label acquisition module may also include:

[0220] A cleaning and correction unit (not shown) is used to perform cleaning and correction processing on the second training data set; a retraining unit (not shown) is used to retrain the classification model using the second training data set after the cleaning and correction processing on the second training data set; a re-prediction unit (not shown) is used to perform prediction processing on the segmented image using the retrained classification model to obtain a block label for each segmented image.

[0221] The data cleaning and correction unit performs cleaning and correction processing on the second training data set, and the retraining unit retrains the classification model, thereby improving the accuracy of the training samples to avoid interference of erroneous samples on the model training as much as possible, thereby improving the reliability and robustness of the classification model obtained by training, and correspondingly improving the accuracy of the block labels obtained by the prediction processing.

[0222] In a specific implementation, the second training data set may be cleaned and corrected based on cross-validation combined with confidence learning, fidelity weighted learning, or sampling detection.

[0223] In some embodiments, based on cross-validation combined with confidence learning, cleaning and correction processing of the second training data set may include: obtaining the joint distribution of the pathological diagnosis results and the block labels of the segmented images in the second training data set; based on the joint distribution, according to the mean of the probability of each category, deleting from the second training data set samples that are lower than the current category mean and whose probability of another category is higher than the mean of another category.

[0224] By obtaining the joint distribution of the pathological diagnosis results and the block labels of the segmented images in the second training data set, and based on the joint distribution, based on the joint distribution, according to the mean of the probability of each category, the samples whose probability is lower than the current category mean and whose probability of another category is higher than the mean of another category are deleted from the second training data set, thereby achieving the purpose of data cleaning, and readjusting the category weights in the second training data set, and then retraining the classification model, thereby improving the accuracy of the classification model.

[0225] For example, in a specific implementation, based on the joint distribution of pathological diagnosis results and block labels, samples with a probability lower than the normal class mean and a probability higher than the malignant class mean are deleted from the second training data set according to the mean of the normal and malignant class probabilities.

[0226] In some embodiments, based on fidelity weighted learning, cleaning and correcting the second training data set may include: using the classification model as the first classification model; training a student network using the optical image of the tissue section and the corresponding pathological diagnosis result; training a teacher network using the segmented image and the corresponding block label; and generating soft labels for all the segmented images and the corresponding block labels using the teacher network to obtain a soft data set (Soft FDU-PA230002-CNDataset).

[0227] Correspondingly, the soft data set is used to perform adversarial training on the student network to obtain a second classification model; and the second classification model is used to predict all segmented images to obtain block labels of the segmented images in all tissue sections.

[0228] By training the student network and the teacher network, and using the teacher network to obtain a soft data set, and using the soft data set to perform adversarial training on the student network, a second classification model that approximates the first classification model is obtained, and during the adversarial training process, the noise data in the optical image and the corresponding pathological diagnosis result can be retained, and corresponding fidelity weighted learning can be performed, thereby improving the accuracy and prediction authenticity of the second classification model.

[0229] In some embodiments, based on sampling detection, cleaning and correcting the second training data set may include: dividing the second training data set into multiple sub-data, and using a classification model to perform sampling prediction on each sub-data; after sampling prediction on each sub-data, cleaning the second training data set to re-label or delete from the second training data set data whose prediction results are inconsistent with the classification results obtained based on the pathological diagnosis results; retraining the classification model using the cleaned second training data set; after retraining the classification model, using the classification model to perform prediction processing on the segmented images in the remaining number of tissue slices to obtain block labels for the segmented images in the remaining number of tissue slices.

[0230] By cleaning the second training data set, the data that are inconsistent with the prediction results and the block labels obtained based on the pathological diagnosis results are re-labeled or deleted from the second training data set, and the classification model is retrained, thereby preventing erroneous sample data from interfering with model training, and correspondingly improving the accuracy of classification model prediction.

[0231] It should be noted that, in this embodiment, after obtaining the block labels of the segmented image corresponding to the optical image of each tissue slice, based on the proportion of different block labels of the segmented image in each tissue slice, the positive rate can be given for pictures of a large size range. As an example, the accuracy of the classification model finally trained can reach 95%, and the positive rate can be given for slices of malignant areas, mixed areas, and normal areas. The positive rate of the malignant area is close to 100%, the positive rate of the normal area is 0%, and the positive rate of the mixed area is different due to different specific cancer types. Taking extramammary Paget's disease as an example, the positive rate of the mixed area measured in this embodiment is about 30% to 60%. Therefore, in this embodiment, the prediction accuracy of the classification model is high.

[0232] FDU-PA230002-CN

[0233] The classification label acquisition module 405 obtains the classification label corresponding to the measurement point based on the position correspondence between the segmented image and the measurement point, and the block label of the segmented image. The measurement point and the corresponding classification label and the signal light characteristic curve are used to form a first training data set, so as to facilitate the subsequent training of the initial detection model using the first training data set.

[0234] Compared with the optical image of each tissue slice, the segmented image and the measurement point are smaller in size and more precise, so that in the step of training the initial detection model using the first training data set to obtain a trained detection model, the precision of the detection model in classification judgment can be improved, which is correspondingly beneficial to improving the accuracy and precision of detection, and is particularly suitable for the detection and judgment of multifocal diseases. Moreover, by obtaining multiple measurement points on the test sample and the signal light characteristic curves of the measurement points, it is also beneficial to improve the efficiency of obtaining characteristic data, and the detection model can be used to obtain the corresponding classification label based on the original data (i.e., the measurement points and the corresponding signal light characteristic curves), without the need to perform fitting analysis on the signal light characteristic curves, which is beneficial to improving detection efficiency.

[0235] In this embodiment, the segmented image close to the surface of the tissue slice in the optical image of each tissue slice corresponds to the measurement point, and accordingly, the block label of the segmented image close to the surface of the tissue slice is used as the classification label of the measurement point at the corresponding position.

[0236] The model building module 406 is used to build an initial detection model so that the training module 407 can use the first training data set to train the initial detection model to obtain a trained detection model.

[0237] As an example, the model building module 406 builds the initial detection model based on logistic regression, decision tree, random forest or support vector machines (SVM).

[0238] In this embodiment, the model building module 406 constructs the initial detection model based on the support vector machine as an example for explanation. The support vector machine is a novel small sample learning method with a solid theoretical basis, which basically does not involve probability measurement and the law of large numbers, and its final decision function is determined by only a few support vectors. The complexity of the calculation depends on the number of support vectors, rather than the dimension of the sample space, which in a sense avoids the "dimensionality disaster".

[0239] In other embodiments, the model building module may also build the initial detection model based on other suitable classification models.

[0240] It should be noted that, in this embodiment, the detection model construction system 400 may further include: a data expansion module (not shown), which is used to obtain the classification label corresponding to the measurement point based on the position correspondence between the segmented image and the measurement point, and the block label of the segmented image.

[0241] Before using the first training data set to train the initial detection model, data expansion processing is performed on the first training data set.

[0242] The data expansion module performs data expansion processing on the first training data set to increase the sample size.

[0243] As an example, the data expansion module includes any one or more of the following situations: a superposition unit (not shown in the figure), which is used to linearly superimpose several signal light characteristic curves of the same category using random proportional coefficients whose sum is 1; a translation unit (not shown in the figure), which is used to randomly translate the signal light characteristic curve along the horizontal axis within a limited spectral line displacement fluctuation range; a noise adding unit (not shown in the figure), which is used to add random Gaussian white noise of varying intensities to the signal light characteristic curve.

[0244] The training module 407 is used to train the initial detection model using the first training data set to obtain a trained detection model.

[0245] The detection model is used to obtain corresponding classification labels based on the measurement points and the corresponding signal light characteristic curves.

[0246] In this embodiment, by performing image segmentation on the optical image of each of the tissue slices to obtain a plurality of segmented images, and obtaining a block label for each segmented image in the tissue slice, and based on the positional correspondence between the segmented image and the measurement point and the block label of the segmented image, a classification label corresponding to the measurement point is obtained. Compared with the optical image of each tissue slice, the segmented image and the measurement point are smaller in size and more precise. Therefore, in the step of training the initial detection model using the first training data set to obtain a trained detection model, the precision of the detection model in classification judgment can be improved, which is correspondingly beneficial to improving the accuracy and precision of detection, and is particularly suitable for the detection and judgment of multifocal diseases. Moreover, by obtaining a plurality of measurement points on the detection sample and the signal light characteristic curves of the measurement points, it is also beneficial to improve the efficiency of obtaining feature data, and the detection model can be used to perform prediction processing based on the original data (i.e., the measurement point and the corresponding signal light characteristic curve) to obtain the corresponding classification label, and there is no need to perform fitting analysis on the signal light characteristic curve, which is beneficial to improving the detection efficiency.

[0247] The prediction module 408 is used to use the trained detection model to perform prediction processing on the characteristic curves of the measurement points and signal light to obtain corresponding classification labels.

[0248] By using the trained detection model, the characteristic curves of the measurement points and the signal light are predicted to obtain the corresponding classification labels, thereby improving the accuracy and precision of the detection, which is particularly suitable for the detection and judgment of multifocal diseases. Moreover, by obtaining multiple measurement points on the test sample and the characteristic curves of the signal light of the measurement points, it is also beneficial to improve the efficiency of obtaining feature data, and can be based on the original data FDU-PA230002-CN

[0249] The corresponding classification label is obtained by using the detection model based on the data (i.e., the measurement point and the corresponding signal light characteristic curve), and there is no need to perform fitting analysis on the signal light characteristic curve, which is beneficial to improving the detection efficiency.

[0250] In addition, in this embodiment, the classification labels corresponding to the measurement points are first obtained, and the measurement points and the corresponding classification labels and the signal light characteristic curves constitute a first training data set, and the initial detection model is trained using the first training data set to obtain a trained detection model. That is, this embodiment is a supervised clustering algorithm, and the number of clusters is relatively clear (2 or 3), so that the detection model can easily achieve high accuracy, thereby facilitating the detection of complex tissues.

[0251] In other embodiments, the detection method may further include: obtaining the positive rate of the detection sample based on the classification labels of all measurement points on the same detection sample. Specifically, the positive rate of the detection sample is obtained based on the ratio of measurement points with different classification labels on the same detection sample.

[0252] The embodiment of the present invention further provides a device, which can implement the detection method provided by the embodiment of the present invention by loading the above-mentioned detection model construction method in the form of a program.

[0253] An optional hardware structure of the terminal device provided by the embodiment of the present invention can be as follows Fig. 9 As shown, it includes: at least one processor 01, at least one communication interface 02, at least one memory 03 and at least one communication bus 04.

[0254] In the embodiment of the present invention, the number of processor 01, communication interface 02, memory 03, and communication bus 04 is at least one, and the processor 01, communication interface 02, and memory 03 communicate with each other through the communication bus 04; optionally, the communication interface 02 may be an interface of a communication module for network communication, such as an interface of a GSM module; the processor 01 may be a central processing unit CPU, or an application specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiment of the present invention.

[0255] The memory 03 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0256] The memory 03 stores one or more computer instructions, and the one or more computer instructions are executed by the processor 01 to implement the detection method provided in the embodiment of the present invention.

[0257] It should be noted that the above-mentioned terminal device may also include other devices (not shown) that may not be necessary for understanding the contents disclosed in the embodiments of the present invention; given that these other devices may not be necessary for understanding the contents disclosed in the embodiments of the present invention, the embodiments of the present invention will not introduce them one by one.

[0258] FDU-PA230002-CN

[0259] Correspondingly, an embodiment of the present invention further provides a storage medium, wherein the storage medium stores one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the detection method provided by the embodiment of the present invention.

[0260] The above-mentioned embodiments of the present invention are combinations of elements and features of the present invention. Unless otherwise mentioned, the elements or features may be considered as optional. Each element or feature may be put into practice without being combined with other elements or features. In addition, the embodiments of the present invention may be constructed by combining some elements and / or features. The order of operations described in the embodiments of the present invention may be rearranged. Some constructions of any one embodiment may be included in another embodiment, and may be replaced by the corresponding construction of another embodiment. It is obvious to those skilled in the art that claims that do not have a clear reference relationship to each other in the attached claims may be combined into embodiments of the present invention, or may be included as new claims in the amendment after submitting this application.

[0261] The embodiments of the present invention can be implemented by various means such as hardware, firmware, software or a combination thereof. In a hardware configuration, the method according to the exemplary embodiment of the present invention can be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, etc.

[0262] In a firmware or software configuration, the embodiments of the present invention may be implemented in the form of modules, procedures, functions, etc. The software code may be stored in a memory unit and executed by a processor. The memory unit is located inside or outside the processor and may send data to and receive data from the processor via various known means.

[0263] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0264] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.

Claims

1. A detection method, characterized in that: include: Obtaining characteristic data of a test sample, the characteristic data comprising a plurality of measurement points located on the test sample and a characteristic curve of a signal light generated by each of the measurement points; Obtaining optical imaging data of tissue sections of the test sample at several intervals; Based on the optical imaging data, performing image segmentation on the optical image of each of the tissue slices to obtain a plurality of segmented images; Obtaining a block label for each segmented image in the tissue slice; Based on the position correspondence between the segmented image and the measurement point, and the block label of the segmented image, the classification label corresponding to the measurement point is obtained, and the measurement point, the corresponding classification label and the signal light characteristic curve are used to form a first training data set; Build an initial detection model; Using the first training data set to train the initial detection model to obtain a trained detection model; The trained detection model is used to predict the characteristic curves of the measurement points and the corresponding signal light to obtain the corresponding classification labels.

2. The detection method according to claim 1, characterized in that The step of obtaining a block label of each segmented image in the tissue slice comprises: Obtaining pathological diagnosis results of a portion of the tissue sections; Based on the pathological diagnosis result, labeling processing is performed on each segmented image in the partial number of tissue sections to obtain block labels of the segmented images, and the segmented images in the partial number of tissue sections and the corresponding block labels are used to form a second training data set; Build a classification model; Using the second training data set, training the classification model to obtain a trained classification model; The trained classification model is used to perform prediction processing on the segmented images in the remaining number of tissue slices to obtain block labels of the segmented images in the remaining number of tissue slices.

3. The detection method according to claim 2, characterized in that The step of obtaining the block label of each segmented image in the tissue section also includes: obtaining the block label of each segmented image in the partial number of tissue sections based on the pathological diagnosis result, and before using the second training data set to train the classification model, performing image recognition processing on the segmented images in the second training data set to obtain image recognition results; based on the image recognition results, removing the segmented images whose background area accounts for more than a preset value from the second training data set; and / or, The step of obtaining the block labels of each segmented image in the tissue slice also includes: after obtaining the block labels of the segmented images in the remaining number of tissue slices, based on cross-validation combined with confidence learning, fidelity weighted learning, or sampling detection, cleaning and correcting the second training data set; after cleaning and correcting the second training data set, retraining the classification model using the second training data set; and predicting the segmented images using the retrained classification model to obtain the block labels of each segmented image.

4. The detection method according to claim 1, characterized in that The characteristic curve of the signal light includes a fluorescence lifetime decay curve or a spectrum curve.

5. The detection method according to claim 1, characterized in that In the step of obtaining characteristic data of the detection sample, multi-channel characteristic data of the detection sample is obtained, and the wavelength range of the optical signal obtained by each channel is different.

6. The detection method according to claim 1, characterized in that The step of obtaining characteristic data of the test sample includes: setting the measurement point on the test sample based on the test requirements; testing the measurement point to obtain the characteristic data of the measurement point; Based on the optical imaging data, the step of performing image segmentation on the optical image of each tissue slice to obtain a plurality of segmented images comprises: performing image segmentation on the optical image of each tissue slice to obtain a plurality of segmented images based on detection requirements; The segmented image close to the surface layer of the tissue slice in the optical image of each tissue slice corresponds to the measurement point.

7. The detection method according to claim 1, characterized in that The method further includes: after obtaining the classification labels corresponding to the measurement points based on the position correspondence between the segmented image and the measurement points, and the block labels of the segmented image, and before training the initial detection model using the first training data set, performing data expansion processing on the first training data set; The step of performing data expansion processing on the first training data set includes any one or more of the following situations: linearly superimposing several signal light characteristic curves of the same category using random proportional coefficients whose sum is 1; randomly translating the signal light characteristic curve along the horizontal axis within a limited spectral line displacement fluctuation range; adding random Gaussian white noise of varying intensities to the signal light characteristic curve.

8. The detection method according to claim 1, characterized in that The optical imaging data includes fluorescence lifetime imaging data, stimulated Raman scattering microscopy imaging data, multiphoton excitation fluorescence imaging data, autofluorescence imaging data, histochemical staining image data or second harmonic imaging data.

9. The detection method according to claim 1, characterized in that: The detection method further includes: obtaining the positive rate of the detection sample based on the classification labels of all measurement points on the same detection sample.

10. A detection device, characterized in that: The detection device is used to obtain the characteristic data of the detection sample in the detection method according to any one of claims 1 to 9; the detection device comprises: A light source, used for providing detection light; An optical fiber detection end connected to the light source, the optical fiber detection end is used as a detection end for detecting a measurement point on the detection sample; wherein the detection light is irradiated onto the measurement point through the optical fiber detection end so that the detection sample at the measurement point generates signal light; A detection module is connected to the optical fiber detection end, and is used to detect the signal light generated by the detection sample at the measurement point position and obtain a signal light characteristic curve of each measurement point.

11. The detection device according to claim 10, characterized in that: The detection module comprises: A plurality of detection units, each detection unit group detects a different wavelength range of optical signals; each detection unit group comprises a filter and a detector connected to each other, the filter is connected to the optical fiber detection end; or, A filter turntable is connected to the optical fiber detection end, and a plurality of filters are arranged on the filter turntable; and a detector is connected to the filter turntable.

12. The detection device according to claim 10 or 11, characterized in that: The detection module includes: a spectrometer connected to the optical fiber detection end and used to obtain a spectrum curve.

13. A detection system, characterized in that: include: A data acquisition module, used to obtain characteristic data of a test sample, wherein the characteristic data includes a plurality of measurement points located on the test sample and a characteristic curve of a signal light generated by each of the measurement points; An imaging module, used to obtain optical imaging data of tissue slices of the test sample at several intervals; An image segmentation module, configured to perform image segmentation on the optical image of each tissue slice based on the optical imaging data to obtain a plurality of segmented images; A block label acquisition module, used to obtain a block label of each segmented image in the tissue section; A classification label acquisition module, used for obtaining the classification label corresponding to the measurement point based on the position correspondence between the segmented image and the measurement point, and the block label of the segmented image, wherein the measurement point, the corresponding classification label and the signal light characteristic curve are used to form a first training data set; Model building module, used to build the initial detection model; A training module, used to train the initial detection model using the first training data set to obtain a trained detection model; The prediction module is used to use the trained detection model to predict the characteristic curve of the measurement point and the corresponding signal light to obtain the corresponding classification label.

14. A device, characterized in that The method comprises at least one memory and at least one processor, wherein the memory stores one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the detection method according to any one of claims 1 to 9.

15. A storage medium, characterized in that: The storage medium stores one or more computer instructions, and the one or more computer instructions are used to implement the detection method according to any one of claims 1-9.

Citation Information

Patent Citations

  • Cancer diagnosis system based on autofluorescence lifetime

    CN106092996A

  • Weakly supervised gland instance segmentation method based on point labeling

    CN112258480A

  • Cancer pathological image automatic detection method and system

    CN112767325A

  • Image data processing method and device based on multi-label segmentation model

    CN113628226A

  • PDL1 expression level prediction method based on multi-instance knowledge distillation model

    CN114970862A

Cited By

  • Bile duct cancer detection method

    CN120275364A