CN113822861A - Method and device for judging eye surface swelling - Google Patents
Method and device for judging eye surface swelling Download PDFInfo
- Publication number
- CN113822861A CN113822861A CN202111044291.0A CN202111044291A CN113822861A CN 113822861 A CN113822861 A CN 113822861A CN 202111044291 A CN202111044291 A CN 202111044291A CN 113822861 A CN113822861 A CN 113822861A
- Authority
- CN
- China
- Prior art keywords
- eye surface
- candidate region
- map
- surface swelling
- swelling
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000008961 swelling Effects 0.000 title claims abstract description 122
- 238000000034 method Methods 0.000 title claims abstract description 50
- 206010028980 Neoplasm Diseases 0.000 claims abstract description 84
- 230000003211 malignant effect Effects 0.000 claims abstract description 41
- 230000003902 lesion Effects 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 20
- 238000000605 extraction Methods 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 7
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 claims description 2
- 210000001508 eye Anatomy 0.000 description 106
- 238000003745 diagnosis Methods 0.000 description 13
- 201000011510 cancer Diseases 0.000 description 11
- 238000011282 treatment Methods 0.000 description 11
- 238000012800 visualization Methods 0.000 description 11
- 208000009077 Pigmented Nevus Diseases 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 210000000744 eyelid Anatomy 0.000 description 5
- 206010027145 Melanocytic naevus Diseases 0.000 description 4
- 206010004146 Basal cell carcinoma Diseases 0.000 description 3
- 210000000795 conjunctiva Anatomy 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 210000005252 bulbus oculi Anatomy 0.000 description 2
- 230000007159 enucleation Effects 0.000 description 2
- 230000004807 localization Effects 0.000 description 2
- 230000036210 malignancy Effects 0.000 description 2
- 201000001441 melanoma Diseases 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 206010061619 Deformity Diseases 0.000 description 1
- 206010020649 Hyperkeratosis Diseases 0.000 description 1
- 208000001126 Keratosis Diseases 0.000 description 1
- 206010030113 Oedema Diseases 0.000 description 1
- 241000934136 Verruca Species 0.000 description 1
- 208000000260 Warts Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000001574 biopsy Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 210000004087 cornea Anatomy 0.000 description 1
- 230000001066 destructive effect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 230000001771 impaired effect Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000013532 laser treatment Methods 0.000 description 1
- 201000002576 malignant conjunctival melanoma Diseases 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000002271 resection Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 201000010153 skin papilloma Diseases 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 230000000451 tissue damage Effects 0.000 description 1
- 231100000827 tissue damage Toxicity 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Pathology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Epidemiology (AREA)
- Evolutionary Biology (AREA)
- Primary Health Care (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method and a device for judging eye surface swelling, wherein the method comprises the following steps: firstly, obtaining an eye surface swelling map; inputting the eye surface swelling map into the eye surface swelling judgment model so as to judge whether the eye surface swelling is benign or malignant by the eye surface swelling judgment model to obtain a judgment result; the eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to an eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is good or not after setting the dimensionality of the plurality of second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region. The embodiment of the invention can improve the accuracy of judging whether the ocular surface tumor is benign or malignant.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a device for judging eye surface tumors.
Background
The ocular surface tumor refers to a tumor occurring on the surface structure of the eye such as eyelid, conjunctiva, cornea, and eye appendage, and includes benign tumor such as pigmented nevus and verruca senilis, and various malignant tumors such as basal cell carcinoma and melanoma. Since the tumor is located on the surface of the eye and the tumors with different properties have respective specific appearance, the tumor can be directly seen by optical examination. In the prior art, the method is mainly based on the morphological characteristics of the tumor directly observed by an ophthalmologist under a slit lamp microscope before an operation, and preliminarily judging whether the ocular surface tumor is benign or malignant by combining the medical history, so as to guide the next treatment. However, the method for manually judging whether the ocular surface tumor is benign or malignant by an ophthalmologist in the prior art has the following disadvantages:
1. the doctor is lack of resources, and the traditional manual screening of the benign and malignant tumors can not meet the medical requirements. The diagnosis of the ocular edema has particularity, and needs to have cross comprehensive knowledge of ophthalmology, dermatology and oncology. The doctor culture period is long, the human capital is high, and only a few specialist doctors can master the characteristics of different tumors under the slit lamp microscope. 2. Preoperative judgment of benign and malignant diseases lacks objective standards and has limited accuracy: the method mainly depends on morphological manifestation under the slit lamp, combines with medical history to carry out empirical diagnosis, lacks a standardized diagnosis process, and has the accuracy rate of professor fluctuation of 75-85 percent. And the malignant tumor is easy to miss diagnosis and misdiagnosis: malignant pigmented tumors often disguise as common "pigmented nevi" threatening the life and health of the patient.
In summary, the accuracy of the conventional method for manually judging whether the ocular surface tumor is benign or malignant by an ophthalmologist is not high.
Disclosure of Invention
The embodiment of the invention provides a method and a device for judging an ocular surface tumor, which improve the accuracy of judging whether the ocular surface tumor is benign or malignant.
A first aspect of an embodiment of the present application provides a method for determining an eye surface tumor, including:
obtaining an eye surface swelling map;
inputting the eye surface swelling map into the eye surface swelling judgment model so as to enable the eye surface swelling judgment model to judge whether the eye surface swelling is benign or malignant and obtain a judgment result; the eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to an eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is good or not after setting the dimensionality of the plurality of second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region.
In a possible implementation manner of the first aspect, after setting the dimensions of the plurality of second feature maps as the first dimensions, the method further includes:
and classifying the ocular surface tumor according to a plurality of second feature maps with the dimensions being the first dimensions, and outputting a classification result.
In a possible implementation manner of the first aspect, after the obtaining the candidate region, the method further includes:
and correcting the position of the candidate region by calculating regression of the candidate region to obtain the corrected candidate region.
In a possible implementation manner of the first aspect, after the obtaining the candidate region, the method further includes:
the candidate region is zoomed according to a first proportion and then is mapped to the eye surface tumor image, and the confidence coefficient of the candidate region is obtained;
and according to the confidence coefficient, combining the judgment result, the classification result and the corrected candidate region to generate a visualization result.
In one possible implementation manner of the first aspect, the obtaining of the eye surface swelling map specifically includes:
an initial input image is obtained, and the initial input image is scaled according to a first proportion to generate an eye surface tumor image and is obtained.
In a possible implementation manner of the first aspect, the generation process of the ocular surface tumor determination model specifically includes:
acquiring a first training sample set; wherein the first training sample set comprises: an ocular surface tumor map containing benign indicia, malignant indicia and ocular surface tumor type indicia;
turning over the first training sample set according to a preset probability and performing HSV color conversion to obtain a second training sample set;
and inputting the second training sample set into the neural network model so as to generate the eye surface tumor judgment model after the neural network model is trained.
In one possible implementation manner of the first aspect, the ocular surface tumor determination model includes: the device comprises a feature extraction layer, an RPN layer, a Roi pooling layer and a classification layer;
the characteristic extraction layer is used for acquiring a first characteristic diagram according to the eye surface swelling map;
the RPN layer is used for acquiring a candidate region according to the first feature map;
the Roi pooling layer is used for obtaining a plurality of second feature maps according to the first feature map and the candidate regions, and setting the dimension of the second feature maps to be a first dimension;
the classification layer is used for judging whether the eye surface tumor is benign or malignant and classifying the eye surface tumor according to a plurality of second feature maps with the dimensions being the first dimension.
In a possible implementation manner of the first aspect, the classification layer is further configured to correct the position of the candidate region by calculating a regression of the candidate region, so as to obtain a corrected candidate region.
In one possible implementation manner of the first aspect, the ocular surface tumor determination model further includes: an output layer;
the output layer is used for outputting the corrected candidate region, the judgment result and the classification result;
the output layer is also used for mapping the candidate region to the eye surface tumor image after scaling according to the first proportion, and outputting after obtaining the confidence coefficient of the candidate region.
A second aspect of the embodiments of the present application provides an apparatus for determining an eye surface swelling, including: the device comprises an acquisition module and a judgment module;
the acquisition module is used for acquiring an eye surface swelling map;
the judging module is used for inputting the eye surface swelling map into the eye surface swelling judging model so as to enable the eye surface swelling judging model to judge whether the eye surface swelling is benign or malignant and obtain a judging result; the eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to an eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is good or not after setting the dimensionality of the plurality of second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region.
Compared with the prior art, the method and the device for judging the eye surface swelling provided by the embodiment of the invention have the beneficial effects that: according to the judging method provided by the embodiment of the invention, an eye surface swelling map is obtained firstly; inputting the eye surface swelling map into the eye surface swelling judgment model so as to judge whether the eye surface swelling is benign or malignant by the eye surface swelling judgment model to obtain a judgment result; the eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to an eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is good or not after setting the dimensionality of the plurality of second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region.
According to the embodiment of the invention, after the eye surface swelling map is input into the eye surface swelling model, the benign and malignant judgment of the eye surface swelling can be automatically carried out, and the judgment result is obtained, so that the interference of manual judgment and the possible misdiagnosis are avoided, and the accuracy of the benign and malignant judgment of the eye surface swelling is improved.
Secondly, the benign and malignant judgment of the ocular surface tumor can be carried out only by inputting the ocular surface tumor image into the ocular surface tumor model, compared with the method for manually judging the benign and malignant of the ocular surface tumor by an ophthalmologist in the prior art, the method is quicker, the diagnosis and treatment level of the basic hospital and the physical examination center on the ocular surface tumor can be improved in a short time, and the method is suitable for remote medical treatment and timely referral of malignant tumor patients; and is not limited by the shortage of doctor resources, and can be widely applied to common ophthalmology outpatient service and physical examination centers.
Besides judging whether the eye surface tumor is benign or malignant, correcting the position of the candidate region to obtain a corrected candidate region; and obtaining the confidence of the candidate region. And according to a plurality of confidence degrees, combining the judgment result, the classification result and the corrected candidate region, obtaining the candidate region corresponding to the maximum confidence degree in the eye surface swelling map, the judgment result and the classification result thereof, and taking the candidate region as the visualization result of the eye surface swelling map, so that the visualization result has more guiding significance and provides more information for a clinician: the clinician can judge the reliability of the AI report according to the visualization result, thereby selecting to continuously increase the patient detection picture for redetection or combine the clinic to carry out the next diagnosis and treatment.
Moreover, the method for manually judging the quality and the malignancy of the ocular surface tumor by an ophthalmologist needs special training for the ophthalmologist, and the diagnosis of the ocular surface tumor relates to a wide professional field, needs to have cross knowledge of ophthalmology, dermatology and oncology at the same time, and has high cost for the doctors to cultivate, long growth curve and talent shortage. Therefore, the method for judging the ocular surface swelling provided by the embodiment of the invention can be realized only by establishing an ocular surface swelling model, and has better economical efficiency and timeliness.
And finally, the eye surface swelling model is combined with a cloud platform for use, so that the user can judge whether the eye surface swelling is benign or malignant and classify the eye surface swelling at any time and any place, and the convenience for judging whether the eye surface swelling is benign or malignant and classifying the eye surface swelling is further improved.
Drawings
Fig. 1 is a schematic flow chart illustrating a method for determining an eye surface swelling according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a process for generating a model for determining an ocular surface mass according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an apparatus for determining an eye surface swelling object according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the existing method for manually judging whether the ocular surface tumor is benign or malignant by an ophthalmologist, the eyelid and the pigmented nevus conjunctiva are the most common ocular surface tumors on the ocular surface, have similar color and appearance, and are easily confused with benign keratotic lesion or malignant tumor such as basal cell carcinoma and melanoma, so that misdiagnosis or missed diagnosis is caused. Misdiagnosis or missed diagnosis can cause the patient to face a great threat of eyeball loss and life. In addition, even benign tumors, excessive surgical treatment can have serious consequences of disfigurement and impaired vision.
The judgment of the benign and malignant tumor can determine the selection of different tissue damage treatment modes, from conservative follow-up to laser treatment, local excision, destructive enucleation of eyeball and enucleation of orbital contents, and the like, so the judgment of the benign and malignant tumor needs to meet the requirement of high precision.
The preoperative benign and malignant judgment of the tumor is beneficial to selecting a treatment mode which can keep the tissue function as much as possible, and the vision and the appearance are kept as much as possible on the basis of improving the survival rate of patients. At present, pathological examination is still the gold standard for judging tumor properties, but it must perform tumor resection, which generates additional ocular surface reconstruction burden on some benign tumors, and may cause untimely treatment of malignant tumors due to the consideration of biopsy burden.
How to perform noninvasive and accurate judgment on the benign and malignant nature of the ocular surface tumor is a problem to be urgently solved at present.
In order to solve the above technical problems, the present invention provides a method for determining an eye surface swelling. Fig. 1 is a schematic flow chart of a method for determining an eye surface swelling according to an embodiment of the present invention, including S101, S102:
s101: an eye surface swelling map is obtained.
In this embodiment, the eye surface swelling map can be obtained by photographing the eye surface by an eye surface slit lamp microscope camera.
S102: inputting the eye surface swelling map into the eye surface swelling judgment model so that the eye surface swelling judgment model judges whether the eye surface swelling is benign or malignant to obtain a judgment result.
The eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to an eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is benign or malignant after setting the dimensionality of the plurality of second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region. In one embodiment, the model for judging ocular surface masses uses the fast-RCNN algorithm.
In this embodiment, after the setting the dimensions of the plurality of second feature maps as the first dimension, the method further includes: and classifying the ocular surface tumor according to the second feature maps with the first dimension, and outputting a classification result.
In this embodiment, after acquiring the candidate region, the method further includes:
and correcting the position of the candidate region by calculating regression of the candidate region to obtain a corrected candidate region.
Since the candidate region is a suspected lesion region, in order to improve the accuracy of the suspected lesion region, i.e., the localization result of the tumor target, the position of the candidate region needs to be corrected to obtain an accurate localization result of the tumor target.
In this embodiment, after acquiring the candidate region, the method further includes:
the candidate region is scaled according to a first proportion and then mapped to the eye surface tumor image, and the confidence coefficient of the candidate region is obtained;
and generating a visual result by combining the judgment result, the classification result and the corrected candidate region according to the confidence coefficient.
Because a plurality of candidate regions may exist on an ocular surface tumor map, and each candidate region can obtain a corresponding judgment result and classification result after judging whether the ocular surface tumor is benign or malignant. In order to make the judgment result and the classification result have reference significance and improve the judgment accuracy, after obtaining a plurality of confidence degrees of a plurality of candidate regions, according to the plurality of confidence degrees, combining the judgment result and the corrected candidate region, obtaining the candidate region corresponding to the maximum confidence degree and the judgment result thereof, and using the candidate region and the judgment result as the visualization result of the eye surface tumor map, so that the visualization result has guiding significance and provides more information for a clinician. The clinician can judge the reliability of the AI report according to the visualization result, thereby selecting to continuously increase the patient detection picture for redetection or combine the clinic to carry out the next diagnosis and treatment.
In this embodiment, the obtaining the eye surface swelling map specifically includes:
and acquiring an initial input image, and scaling the initial input image according to the first proportion to generate and acquire the eye surface tumor map.
To further explain the generation process of the eye surface swelling judgment model, please refer to fig. 2, fig. 2 is a schematic diagram of the generation process of the eye surface swelling judgment model according to an embodiment of the present invention, including S201-S203:
s201: a first set of training samples is obtained.
Wherein the first training sample set comprises: an ocular surface tumor map containing benign markers, malignant markers, and ocular surface tumor type markers.
In one embodiment, the eye surface tumor type labeling comprises: malignant melanoma of conjunctiva, basal cell carcinoma of eyelid, pigmented nevus of conjunctiva, pigmented nevus of eyelid, and benign keratosis of eyelid.
S202: and turning over the first training sample set according to a preset probability and carrying out HSV color conversion to obtain a second training sample set.
The overturning and HSV color conversion of the first training sample set according to the preset probability aims to increase the number of samples and improve the generalization capability of the model so as to obtain a second training sample set after data enhancement.
S203: and inputting the second training sample set into the neural network model so as to generate the eye surface tumor judgment model after the neural network model is trained.
In this embodiment, the model for judging the ocular surface mass generated after training includes: the device comprises a feature extraction layer, an RPN layer, a Roi pooling layer and a classification layer.
The feature extraction layer is used for acquiring the first feature map according to the eye surface tumor map.
In a specific embodiment, the feature extraction layer extracts features of an image through a plurality of convolution operations, and a network architecture of the feature extraction layer is an image classification network framework currently prevailing, such as VGG16, ResNet50, and the like, and in this embodiment, ResNet50 is used. The output result of the feature extraction layer is the first feature map, which is a feature map with global picture information, and its essence is a multi-dimensional array.
The RPN layer is used for acquiring the candidate region according to the first feature map.
In one embodiment, the RPN layer outputs a plurality of candidate regions of different dimensions in the ocular surface tumor map after acquiring the suspected lesion region according to the first feature map.
The Roi pooling layer is configured to obtain a plurality of second feature maps according to the first feature map and the candidate region, and set a dimension of the plurality of second feature maps as the first dimension.
In a specific embodiment, since the neural network cannot process inputs with different dimensions, the Roi pooling layer obtains a plurality of second feature maps according to the first feature map and the candidate region, and processes the second feature maps into the same dimension, i.e., the first dimension.
In an embodiment, obtaining a plurality of second feature maps according to the first feature map and the candidate region specifically includes: and obtaining a plurality of second feature maps according to the corresponding relation between the first feature map and the candidate region. And the second feature map is a feature map of the candidate region.
The classification layer is used for judging whether the eye surface tumor is benign or malignant and classifying the eye surface tumor according to the second feature maps with the first dimension.
In a specific embodiment, the classification layer is further configured to correct the position of the candidate region by calculating a regression of the candidate region, so as to obtain the corrected candidate region.
In this embodiment, the model for judging ocular surface swelling further includes: an output layer;
the output layer is used for outputting the corrected candidate region, the judgment result and the classification result;
the output layer is further used for mapping the candidate region to the ocular surface tumor map after scaling according to the first proportion, obtaining the confidence coefficient of the candidate region and outputting the confidence coefficient.
From the above, the output result of the output layer includes: the corrected candidate region, the judgment result, the classification result and the confidence coefficient of the candidate region; and according to the confidence coefficient, combining the judgment result, the classification result and the corrected candidate region to generate a visualization result.
Wherein the classification result comprises: the four-classification result and the five-classification result can be designed according to actual requirements.
To further explain the device for determining an ocular surface swelling object, please refer to fig. 3, fig. 3 is a schematic structural diagram of a device for determining an ocular surface swelling object according to an embodiment of the present invention, including: an acquisition module 301 and a judgment module 302.
The acquisition module 301 is configured to acquire an eye surface swelling map;
the judging module 302 is configured to input the eye surface swelling map into an eye surface swelling judging model, so that the eye surface swelling judging model performs benign and malignant judgment on the eye surface swelling to obtain a judging result; the eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to an eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is good or not after setting the dimensionality of the second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region.
Specifically, after the candidate region is obtained, the method further includes:
and correcting the position of the candidate region by calculating regression of the candidate region to obtain a corrected candidate region.
The embodiment of the invention is firstly used for acquiring an eye surface swelling map through an acquisition module 301; the judging module 302 is used for inputting the eye surface swelling map into the eye surface swelling judging model so as to enable the eye surface swelling judging model to judge whether the eye surface swelling is benign or malignant to obtain a judging result; the eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to an eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is good or not after setting the dimensionality of the plurality of second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region.
According to the embodiment of the invention, after the eye surface swelling map is input into the eye surface swelling model, the benign and malignant judgment of the eye surface swelling can be automatically carried out, and the judgment result is obtained, so that the interference of manual judgment and the possible misdiagnosis are avoided, and the accuracy of the benign and malignant judgment of the eye surface swelling is improved.
Secondly, the benign and malignant judgment of the ocular surface tumor can be carried out only by inputting the ocular surface tumor image into the ocular surface tumor model, compared with the method for manually judging the benign and malignant of the ocular surface tumor by an ophthalmologist in the prior art, the method is quicker, the diagnosis and treatment level of the basic hospital and the physical examination center on the ocular surface tumor can be improved in a short time, and the method is suitable for remote medical treatment and timely referral of malignant tumor patients; and is not limited by the shortage of doctor resources, and can be widely applied to common ophthalmology outpatient service and physical examination centers.
Besides judging whether the eye surface tumor is benign or malignant, correcting the position of the candidate region to obtain a corrected candidate region; and after the confidence degrees of the candidate regions are obtained, the candidate region corresponding to the maximum confidence degree in the eye surface swelling map, the judgment result and the classification result are obtained according to a plurality of confidence degrees and by combining the judgment result, the classification result and the corrected candidate regions, and the candidate region, the judgment result and the classification result are used as the visualization result of the eye surface swelling map, so that the visualization result has guiding significance, and more information is provided for a clinician: the clinician can judge the reliability of the AI report according to the visualization result, thereby selecting to continuously increase the patient detection picture for redetection or combine the clinic to carry out the next diagnosis and treatment.
Moreover, the method for manually judging the quality and the malignancy of the ocular surface tumor by an ophthalmologist needs special training for the ophthalmologist, and the diagnosis of the ocular surface tumor relates to a wide professional field, needs to have cross knowledge of ophthalmology, dermatology and oncology at the same time, and has high cost for the doctors to cultivate, long growth curve and talent shortage. Therefore, the method for judging the ocular surface swelling provided by the embodiment of the invention can be realized only by establishing an ocular surface swelling model, and has better economical efficiency and timeliness.
And finally, the eye surface swelling model is combined with a cloud platform for use, so that the user can judge whether the eye surface swelling is benign or malignant and classify the eye surface swelling at any time and any place, and the convenience for judging whether the eye surface swelling is benign or malignant and classifying the eye surface swelling is further improved.
While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention.
Claims (10)
1. A method for judging an ocular surface tumor, comprising:
obtaining an eye surface swelling map;
inputting the eye surface swelling map into an eye surface swelling judgment model so as to enable the eye surface swelling judgment model to judge whether the eye surface swelling is benign or malignant and obtain a judgment result; the eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to the eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is good or not after setting the dimensionality of the plurality of second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region.
2. The method for determining an eye surface tumor according to claim 1, further comprising, after the setting the dimensions of the plurality of second feature maps as the first dimension:
and classifying the ocular surface tumor according to the second feature maps with the first dimension, and outputting a classification result.
3. The method of claim 2, further comprising, after obtaining the candidate region:
and correcting the position of the candidate region by calculating regression of the candidate region to obtain a corrected candidate region.
4. The method of claim 3, further comprising, after obtaining the candidate region:
the candidate region is scaled according to a first proportion and then mapped to the eye surface tumor image, and the confidence coefficient of the candidate region is obtained;
and generating a visual result by combining the judgment result, the classification result and the corrected candidate region according to the confidence coefficient.
5. The method for determining ocular surface swelling according to claim 4, wherein the obtaining of the ocular surface swelling map specifically comprises:
and acquiring an initial input image, and scaling the initial input image according to the first proportion to generate and acquire the eye surface tumor map.
6. The method for judging an ocular surface swelling material as claimed in claim 5, wherein the process of generating the ocular surface swelling material judgment model comprises:
acquiring a first training sample set; wherein the first set of training samples comprises: an ocular surface tumor map containing benign indicia, malignant indicia and ocular surface tumor type indicia;
turning the first training sample set according to a preset probability and performing HSV color conversion to obtain a second training sample set;
inputting the second training sample set into a neural network model so as to generate the eye surface tumor judgment model after the neural network model is trained.
7. The method of claim 6, wherein the model for determining ocular surface swelling comprises: the device comprises a feature extraction layer, an RPN layer, a Roi pooling layer and a classification layer;
the feature extraction layer is used for acquiring the first feature map according to the eye surface tumor map;
the RPN layer is used for acquiring the candidate region according to the first feature map;
the Roi pooling layer is configured to obtain a plurality of second feature maps according to the first feature map and the candidate region, and set a dimension of the plurality of second feature maps to the first dimension;
the classification layer is used for judging whether the eye surface tumor is benign or malignant and classifying the eye surface tumor according to the second feature maps with the first dimension.
8. The method as claimed in claim 7, wherein the classification layer is further configured to calculate a regression of the candidate region, and correct the position of the candidate region to obtain the corrected candidate region.
9. The method of claim 8, wherein the model further comprises: an output layer;
the output layer is used for outputting the corrected candidate region, the judgment result and the classification result;
the output layer is further used for mapping the candidate region to the ocular surface tumor map after scaling according to the first proportion, obtaining the confidence coefficient of the candidate region and outputting the confidence coefficient.
10. An apparatus for judging an eye surface swelling, comprising: the device comprises an acquisition module and a judgment module;
the acquisition module is used for acquiring an eye surface swelling map;
the judging module is used for inputting the eye surface swelling map into an eye surface swelling judging model so as to enable the eye surface swelling judging model to judge whether the eye surface swelling is benign or malignant to obtain a judging result; the eye surface swelling judgment model is used for acquiring a first feature map and a candidate region according to the eye surface swelling map, then acquiring a plurality of second feature maps according to the first feature map and the candidate region, and judging whether the eye surface swelling is good or not after setting the dimensionality of the plurality of second feature maps to be a first dimensionality; wherein the candidate region is a suspected lesion region.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111044291.0A CN113822861A (en) | 2021-09-07 | 2021-09-07 | Method and device for judging eye surface swelling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111044291.0A CN113822861A (en) | 2021-09-07 | 2021-09-07 | Method and device for judging eye surface swelling |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113822861A true CN113822861A (en) | 2021-12-21 |
Family
ID=78922128
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111044291.0A Pending CN113822861A (en) | 2021-09-07 | 2021-09-07 | Method and device for judging eye surface swelling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113822861A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838114A (en) * | 2019-11-11 | 2020-02-25 | 苏州锐一仪器科技有限公司 | Pulmonary nodule detection method, device and computer storage medium |
| CN111144238A (en) * | 2019-12-11 | 2020-05-12 | 重庆邮电大学 | Article detection method and system based on Faster R-CNN |
| CN111652850A (en) * | 2020-05-08 | 2020-09-11 | 怀光智能科技（武汉）有限公司 | Screening system based on mobile device |
| CN112102256A (en) * | 2020-08-22 | 2020-12-18 | 复旦大学 | Narrow-band endoscopic image-oriented cancer focus detection and diagnosis system for early esophageal squamous carcinoma |
| CN112699868A (en) * | 2020-12-17 | 2021-04-23 | 深圳视见医疗科技有限公司 | Image identification method and device based on deep convolutional neural network |
| CN112786189A (en) * | 2021-01-05 | 2021-05-11 | 重庆邮电大学 | Intelligent diagnosis system for new coronary pneumonia based on deep learning |
-
2021
- 2021-09-07 CN CN202111044291.0A patent/CN113822861A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838114A (en) * | 2019-11-11 | 2020-02-25 | 苏州锐一仪器科技有限公司 | Pulmonary nodule detection method, device and computer storage medium |
| CN111144238A (en) * | 2019-12-11 | 2020-05-12 | 重庆邮电大学 | Article detection method and system based on Faster R-CNN |
| CN111652850A (en) * | 2020-05-08 | 2020-09-11 | 怀光智能科技（武汉）有限公司 | Screening system based on mobile device |
| CN112102256A (en) * | 2020-08-22 | 2020-12-18 | 复旦大学 | Narrow-band endoscopic image-oriented cancer focus detection and diagnosis system for early esophageal squamous carcinoma |
| CN112699868A (en) * | 2020-12-17 | 2021-04-23 | 深圳视见医疗科技有限公司 | Image identification method and device based on deep convolutional neural network |
| CN112786189A (en) * | 2021-01-05 | 2021-05-11 | 重庆邮电大学 | Intelligent diagnosis system for new coronary pneumonia based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| WANGTING LI ET AL.: "Dense anatomical annotation of slit-lamp images improves the performance of deep learning for the diagnosis of ophthalmic disorders", 《NATURE BIOMEDICAL ENGINEERING》, pages 1 - 11 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wannous et al. | Enhanced assessment of the wound-healing process by accurate multiview tissue classification | |
| Wortsman | Sonography of facial cutaneous basal cell carcinoma | |
| Flores et al. | Intraoperative imaging during Mohs surgery with reflectance confocal microscopy: initial clinical experience | |
| CN107595390B (en) | Real-time matching fusion method of ultrasonic image and CT image | |
| CN108618749B (en) | Retina blood vessel three-dimensional reconstruction method based on portable digital fundus camera | |
| CN110403576A (en) | Application of the three-dimensional photoacoustic imaging in tumor of breast points-scoring system | |
| CN111598949B (en) | Automatic positioning method for head and neck acupoints of human body | |
| Aubreville et al. | Transferability of deep learning algorithms for malignancy detection in confocal laser endomicroscopy images from different anatomical locations of the upper gastrointestinal tract | |
| Wu et al. | Diabetic macular edema grading based on improved Faster R-CNN and MD-ResNet | |
| CN110960219A (en) | Intelligent auxiliary diagnosis system for skin ulcer wound surfaces | |
| WO2011122936A2 (en) | A methodology and apparatus for objective assessment and rating of psoriasis lesion thickness using digital imaging | |
| CN111466952B (en) | Real-time conversion method and system for ultrasonic endoscope and CT three-dimensional image | |
| Lange et al. | Computer-aided-diagnosis (CAD) for colposcopy | |
| CN113822861A (en) | Method and device for judging eye surface swelling | |
| Gill et al. | Artifacts and landmarks: pearls and pitfalls for in vivo reflectance confocal microscopy of the skin using the tissue-coupled device | |
| Mantena et al. | Low-cost, smartphone-based specular imaging and automated analysis of the corneal endothelium | |
| CN115910300A (en) | Medical virtual platform system based on artificial intelligence and information processing method | |
| WO2022120212A1 (en) | Apparatus and method for detecting cervical cancer | |
| CN114343555A (en) | Colposcope artificial intelligence auxiliary quality control management method and device | |
| Chuh et al. | Outcomes of dermoscope-guided surgical procedures in primary care: case-control study | |
| CN114140399A (en) | OCT image processing device for qualitative and boundary positioning in brain tumor operation | |
| Richarz et al. | Challenges for new adopters in pre-surgical margin assessment by handheld reflectance confocal microscope of basal cell carcinoma; a prospective single-center study | |
| CN111603199B (en) | Three-dimensional reconstruction ultrasonic diagnosis system based on body surface positioning measuring instrument | |
| CA3112555C (en) | Spectrally encoded optical polarization imaging for detecting skin cancer margins | |
| US20230380682A1 (en) | Devices, systems, and methods to measure corneal topography |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |