CN110338906B - Intelligent treatment system for photocrosslinking operation and establishment method - Google Patents
Intelligent treatment system for photocrosslinking operation and establishment method Download PDFInfo
- Publication number
- CN110338906B CN110338906B CN201910622122.7A CN201910622122A CN110338906B CN 110338906 B CN110338906 B CN 110338906B CN 201910622122 A CN201910622122 A CN 201910622122A CN 110338906 B CN110338906 B CN 110338906B
- Authority
- CN
- China
- Prior art keywords
- corneal
- treatment
- parameters
- corneal topography
- training
- 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.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B34/00—Computer-aided surgery; Manipulators or robots specially adapted for use in surgery
- A61B34/10—Computer-aided planning, simulation or modelling of surgical operations
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61F—FILTERS IMPLANTABLE INTO BLOOD VESSELS; PROSTHESES; DEVICES PROVIDING PATENCY TO, OR PREVENTING COLLAPSING OF, TUBULAR STRUCTURES OF THE BODY, e.g. STENTS; ORTHOPAEDIC, NURSING OR CONTRACEPTIVE DEVICES; FOMENTATION; TREATMENT OR PROTECTION OF EYES OR EARS; BANDAGES, DRESSINGS OR ABSORBENT PADS; FIRST-AID KITS
- A61F9/00—Methods or devices for treatment of the eyes; Devices for putting-in contact lenses; Devices to correct squinting; Apparatus to guide the blind; Protective devices for the eyes, carried on the body or in the hand
- A61F9/007—Methods or devices for eye surgery
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Surgery (AREA)
- Public Health (AREA)
- Animal Behavior & Ethology (AREA)
- Heart & Thoracic Surgery (AREA)
- Ophthalmology & Optometry (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Veterinary Medicine (AREA)
- Vascular Medicine (AREA)
- Robotics (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Eye Examination Apparatus (AREA)
Abstract
An intelligent treatment system for corneal photocrosslinking surgery, comprising: a corneal topography image acquisition system configured to acquire a user corneal topography image; a photocrosslinking surgery guidance system pre-trained to output parameters of a photocrosslinking surgery by comparing a corneal topography prior to laser crosslinking of the cornea with an expected post-operative corneal topography; a database corresponding to the topographic map of the pre-treated cornea and the parameters of the photo-crosslinking operation is used for pre-training and correcting the photo-crosslinking operation guide system; the image is acquired by a corneal topography image acquisition system, the corneal topography of a patient is acquired, and then the corneal topography is input into a photo-crosslinking surgery guide system which is pre-trained by a database corresponding to the corneal topography and the photo-crosslinking surgery parameters before treatment, so that the parameters of the photo-crosslinking surgery are acquired. A method of establishing an intelligent treatment system is also disclosed. By using the system, efficient, accurate and stable treatment guidance can be provided for corneal photocrosslinking operation.
Description
Technical Field
The invention relates to an ophthalmic medical technology, in particular to an intelligent treatment system for a photo-crosslinking operation and an establishment method of the intelligent treatment system.
Background
The cornea is a transparent tissue at the front end of the eyeball and plays an important role in an eyeball optical system, the refractive power of the cornea accounts for more than three quarters of the power of the eyeball dioptric system, and slight change of the shape of the cornea can cause great change of vision. Myopia, hyperopia and astigmatism are all ophthalmic diseases related to the refractive power of an eyeball system, and the condition of an illness can be relieved by changing the refractive power of a cornea, so that the effect of restoring vision is achieved.
An ultraviolet cross-linking operation method features that the photosensitizer coated on the surface of cornea is cross-linked with the collagen fibres of cornea under the irradiation of ultraviolet ray to increase the mechanical strength of cornea, resulting in the relief of the further deterioration of keratoconus and the decrease of eyeball system's power under myopia, hyperopia and astigmatism. Another method of infrared laser cross-linking surgery is based on a different laser-corneal interaction pattern. In this mode of action, the near infrared laser (parameter 1059.2nm52.06MHz 99fs) induces the formation of a low density plasma that generates reactive oxygen species that react with surrounding proteins, cross-linking with the cornea and spatially induce changes in the mechanical properties of the cornea. Thereby improving the mechanical strength of the cornea.
The parameter setting of the existing cornea crosslinking operation is still based on the subjective judgment of doctors for many years of experience to a great extent. In recent years, where the number of ophthalmic disease cases has increased rapidly, the number of doctors with sufficient experience has become increasingly difficult to meet. Moreover, even the subjective judgment of a professional doctor is inevitable to cause careless mistakes, and an effective system for assisting in setting the cross-linking surgical parameters is not provided.
Disclosure of Invention
The invention mainly aims to overcome the defects of the prior art, provides an intelligent treatment system for a photo-crosslinking operation and an establishment method of the intelligent treatment system, solves the problem that the existing corneal crosslinking instrument excessively depends on subjective judgment of a user, can provide clear assistance for the instrument user, and obviously improves the treatment effect and efficiency of the corneal crosslinking operation.
In order to achieve the purpose, the invention adopts the following technical scheme:
in one aspect of the invention, an intelligent treatment system for corneal photocrosslinking surgery includes:
a corneal topography image acquisition system configured to acquire a user corneal topography image;
a photocrosslinking surgery guidance system pre-trained to output parameters of a photocrosslinking surgery by comparing a corneal topography prior to laser crosslinking of the cornea with an expected post-operative corneal topography;
the database corresponding to the corneal topography before treatment and the photocrosslinking operation parameters is used for recording the influence of the photocrosslinking operation parameters on the corneal topography and pre-training and correcting the photocrosslinking operation guide system;
the image is acquired by a corneal topography image acquisition system, the corneal topography of a patient is acquired, and then the corneal topography is input into a photo-crosslinking surgery guide system which is pre-trained by a database corresponding to the corneal topography and the photo-crosslinking surgery parameters before treatment, so that the parameters of the photo-crosslinking surgery are acquired.
Further:
the cornea topography image acquisition system comprises a placido optical system and an image acquisition module, wherein the placido optical system is used for providing red LED light, the image acquisition module is used for acquiring cornea topography images, and preferably, the placido optical system adopts a light emitting diode with the wavelength of 760nm and a light source with the optical power not more than 0.42W.
The corneal topography image acquisition system acquires a corneal topography image again for a patient who finishes a photo-crosslinking operation, stores a preoperative corneal topography image, an postoperative actual corneal topography image and input corneal crosslinking parameters into the database corresponding to the corneal topography image and the photo-crosslinking operation parameters before treatment, and retrains the photo-crosslinking operation guide system along with the updating of the database; preferably, the parameters of the photo-crosslinking operation include irradiation path, irradiation power, irradiation time.
Further comprising: pre-treatment corneal topography map images and corresponding post-treatment corneal topography map databases are used for storing image pairs of the pre-treatment corneal topography map images and the post-treatment corneal topography maps for training and continuously correcting the photocrosslinking operation guide system model.
The light cross-linking operation guiding system adopts a convolution neural network model, two corneal topography maps before cross-linking operation and after expected operation are input, and the convolution neural network model outputs parameters for executing laser operation by executing a regression task; preferably, the activation functions of the hidden layer of the photocrosslinking guiding system are both relus, and the activation function of the output layer is lretlu.
The pre-treatment corneal topography and photo-crosslinking operation parameter corresponding database stores a plurality of corneal topography before and after photo-crosslinking operation and crosslinking parameters based on pre-judgment in advance, preferably, two corneal topography before and after photo-crosslinking operation are added into the pre-treatment corneal topography and photo-crosslinking operation parameter corresponding database, and the photo-crosslinking operation guiding system re-corrects the model parameters after the pre-treatment corneal topography and photo-crosslinking operation parameter corresponding database is updated.
The parameters of the photo-crosslinking operation are parameters of an ultraviolet riboflavin crosslinking operation or parameters of an infrared femtosecond laser noninvasive crosslinking operation.
In another aspect of the present invention, a method for establishing an intelligent treatment system for corneal photocrosslinking surgery includes:
a cornea information acquisition step, configured to acquire cornea information of a patient before and after an operation, including image information and parameter information of the cornea; the method comprises the steps of obtaining a corneal topography, an OCT anterior segment image, corneal topography measuring parameter information and OCT measuring parameter information;
a database establishing step, namely establishing a corresponding database according to the cornea information acquired in the cornea information acquiring step;
a step of training a light cross-linking operation guide system model, which is to output parameters required by implementing the light cross-linking operation by comparing corneal information before corneal laser cross-linking with corneal information after cross-linking through pre-training, wherein preferably, the parameters respectively correspond to two systems of an ultraviolet riboflavin cross-linking operation and an infrared femtosecond laser noninvasive cross-linking operation;
model feedback and correction, namely continuously amplifying a database and correcting a model by collecting corneal information before and after treatment.
Further:
the training comprises the following steps:
training between the data set and the data set in the mode of training corneal topography parameters or OCT anterior segment image parameters and treatment parameters before and after BP neural network training treatment;
training between the image and the data in a mode of comparing a corneal topography or an OCT anterior segment image before and after CNN neural network training with treatment parameters;
training between images in a mode of utilizing an antagonistic neural network to train a preoperative and postoperative corneal topography or an OCT anterior segment image and a treatment parameter for imaging comparison;
the BP neural network adopts a back propagation algorithm of a feedforward network with two layers of sigmoid units, the input layer of the BP neural network is an m-dimensional vector and comprises a plurality of groups of data before and after treatment, and the output layer of the BP neural network comprises n-dimensional vectors of treatment parameters;
the CNN neural network inputs images before and after an operation to execute a regression task and outputs parameters for executing laser operation, and an activation function of a hidden layer and an activation function of an output layer are Lrel;
the antagonistic neural network depth adopts convolution to generate an antagonistic network DCGAN, and a convolution with a step length (Stride) is used for replacing a pooling layer in a discriminator D; batch Normalization (Batch Normalization) was used at G, D to help model convergence; in G, the activation function uses the ReLU function, and the last layer uses the tanh function, which is returned to the pixel value in use; in D, the activation functions all use leakage ReLU as the activation function.
Imaging the irradiation path, irradiation time and irradiation power in the treatment parameters, which specifically comprises: the laser path parameter image is two-dimensionalized into a closed curve or a non-closed curve, the time and power parameters are converted into energy density, the value size of the energy density is represented by the color shade, the treatment parameter image is a closed curve or a non-closed curve, and the laser crosslinking energy size is represented by the color density.
The training method is used for an antagonistic neural network with obvious laser treatment path change characteristics and a BP neural network or a CNN network with dominant laser time and power change characteristics.
The invention has the following beneficial effects:
the invention provides an intelligent treatment system for a corneal photo-crosslinking operation, which can provide efficient, accurate and stable treatment guidance for the corneal photo-crosslinking operation. The intelligent treatment system for the corneal photocrosslinking operation can convert subjective experience into more accurate and efficient treatment. In the system, the light cross-linking operation guide system can configure different laser parameters according to topographic maps after different expected cross-linking operations, and meanwhile, the light cross-linking operation guide system can be trained further after the corresponding database is updated, so that the model accuracy is improved. Due to the advantages, the system can be used for effectively and accurately treating myopia, hyperopia, astigmatism and other ophthalmic diseases related to the refractive power of the optical system of the eyeball.
Drawings
FIG. 1 is a block diagram of an intelligent treatment system for corneal photocrosslinking surgery according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a corneal topography image capture system in accordance with an embodiment of the present invention;
FIG. 3 is a block diagram of a training method according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a training method BP neural network according to an embodiment of the present invention:
FIG. 5 is a block diagram of the structure of the training procedure of the DCGAN model according to an embodiment of the present invention:
FIG. 6 is a block diagram of a training method and database structure according to an embodiment of the present invention;
fig. 7 is a block diagram of a database structure utilized by the photocrosslinking surgical guidance system according to an embodiment of the present invention.
Detailed Description
The embodiments of the present invention will be described in detail below. It should be emphasized that the following description is merely exemplary in nature and is not intended to limit the scope of the invention or its application.
Referring to fig. 1 to 7, in an embodiment, an intelligent treatment system for corneal photocrosslinking surgery includes a corneal topography image collecting system, a photocrosslinking surgery guidance system, and a database of a pre-treatment corneal topography map and a photocrosslinking surgery parameter correspondence, wherein the corneal topography image collecting system is configured to collect a corneal topography image of a user, the photocrosslinking surgery guidance system is pre-trained to output parameters of the photocrosslinking surgery by comparing the corneal laser crosslinking corneal topography map with an expected post-surgery corneal topography map, the pre-treatment corneal topography map and the photocrosslinking surgery parameter correspondence database are used to record the influence of the photocrosslinking surgery parameters on the corneal topography map, and are used to pre-train and correct the photocrosslinking surgery guidance system; the image is acquired by a corneal topography image acquisition system, the corneal topography of a patient is acquired, and then the corneal topography is input into a photo-crosslinking surgery guide system which is pre-trained by a database corresponding to the corneal topography and the photo-crosslinking surgery parameters before treatment, so that the parameters of the photo-crosslinking surgery are acquired.
In a preferred embodiment, as shown in fig. 2, the corneal topography image capturing system comprises a placido optical system for providing red LED light and an image capturing module for capturing a corneal topography image, and more preferably, the placido optical system employs a light emitting diode with a wavelength of 760nm and a light source with a light power of not more than 0.42W. The cornea topographic map image acquisition system can also comprise a three-dimensional motion platform, a jaw support bracket, a workstation and other matched equipment.
In a preferred embodiment, the corneal topography image acquisition system acquires a corneal topography image again for a patient who has completed a photo-crosslinking operation, stores a pre-operation corneal topography image, an post-operation actual corneal topography image and input corneal crosslinking parameters into the database corresponding to the pre-treatment corneal topography image and the photo-crosslinking operation parameters, and retrains the photo-crosslinking operation guidance system with the updating of the database; more preferably, the parameters of the photo-crosslinking operation include irradiation path, irradiation power, irradiation time.
Referring to fig. 1, in a preferred embodiment, the intelligent treatment system for corneal photocrosslinking surgery further comprises: pre-treatment corneal topography map images and corresponding post-treatment corneal topography map databases are used for storing image pairs of the pre-treatment corneal topography map images and the post-treatment corneal topography maps for training and continuously correcting the photocrosslinking operation guide system model.
Referring to fig. 6, in a preferred embodiment, the photocrosslinking guidance system inputs two corneal topography maps before crosslinking operation and after expected operation by using a convolutional neural network model, and the convolutional neural network model outputs a parameter for performing laser operation by performing a regression task; more preferably, the activation functions of the hidden layers of the photocrosslinking guiding system are both relus, and the activation function of the output layer is lreol.
In a preferred embodiment, some corneal topography maps before and after the photo-crosslinking operation and a pre-determined crosslinking parameter are pre-stored in the database corresponding to the pre-treatment corneal topography map and the photo-crosslinking operation parameter, more preferably, two corneal topography maps before and after the photo-crosslinking operation are added to the database corresponding to the pre-treatment corneal topography map and the photo-crosslinking operation parameter, and the photo-crosslinking operation guidance system re-corrects the model parameter after the database corresponding to the pre-treatment corneal topography map and the photo-crosslinking operation parameter is updated.
In different embodiments, the parameter of the photo-crosslinking operation may be a parameter for performing an ultraviolet riboflavin crosslinking operation, or may be a parameter for performing an infrared femtosecond laser non-invasive crosslinking operation.
In another embodiment, a method for establishing an intelligent treatment system for corneal photocrosslinking surgery comprises the following steps:
a cornea information acquisition step, configured to acquire cornea information of a patient before and after an operation, including image information and parameter information of the cornea; the method specifically comprises a corneal topography, an OCT anterior segment image, measurement parameter information of a corneal topographer and OCT measurement parameter information; corneal topography parameter information includes, but is not limited to, the meridian direction and value of the Ks maximum refractive power, Kf (the meridian direction and value thereof forming a 90-degree angle with Ks), SAI (surface asymmetry index), etc.; OCT measurement parameter information includes, but is not limited to, corneal thickness (thinnest at the center of the cornea), anterior chamber thickness, pupil width, etc.;
a database establishing step of establishing a corresponding database according to the cornea information acquired in the cornea information acquiring step, as shown in fig. 3, 4, 6 and 7;
the method comprises a step of training a light cross-linking operation guide system model, wherein pre-training is carried out to output parameters required for implementing the light cross-linking operation by comparing corneal information before corneal laser cross-linking with corneal information after cross-linking, and the parameters can respectively correspond to two systems of an ultraviolet riboflavin cross-linking operation and an infrared femtosecond laser noninvasive cross-linking operation; see the training architecture as shown in fig. 3-6;
model feedback and correction, namely continuously amplifying a database and correcting a model by collecting corneal information before and after treatment.
In a preferred embodiment, the training comprises:
training between the data sets in a mode of training corneal topography parameters before and after treatment or OCT (optical coherence tomography) anterior segment image parameters and treatment parameters by a BP neural network (see figure 4);
training between images and data in a mode that a corneal topography map or an OCT anterior segment image before and after CNN neural network training is compared with treatment parameters (see figure 3);
training between images by using an antagonistic neural network to train preoperative and postoperative corneal topography or OCT anterior segment images to graphically compare treatment parameters (see fig. 5);
the BP neural network adopts a back propagation algorithm of a feedforward network with two layers of sigmoid units, an input layer is an m-dimensional vector and comprises a plurality of groups of data before and after treatment, and an output layer is an n-dimensional vector and comprises treatment parameters;
the CNN neural network inputs images before and after an operation to execute a regression task and outputs parameters for executing laser operation, and an activation function of a hidden layer and an activation function of an output layer are Lrel;
wherein, the antagonistic neural network depth adopts convolution to generate an antagonistic network DCGAN, and a convolution with step length (Stride) is used for replacing a pooling layer in a discriminator D; batch Normalization (Batch Normalization) was used at G, D to help model convergence; in G, the activation function uses the ReLU function, and the last layer uses the tanh function, which is returned to the pixel value in use; in D, the activation functions all use leakage ReLU as the activation function.
In a preferred embodiment, the method further comprises imaging the irradiation path, irradiation time and irradiation power in the treatment parameters, which specifically comprises: the laser path parameter image is two-dimensionalized into a closed curve or a non-closed curve, the time and power parameters are converted into energy density, the value size of the energy density is represented by the color shade, the treatment parameter image is a closed curve or a non-closed curve, and the laser crosslinking energy size is represented by the color density.
Referring to fig. 7, in a preferred embodiment, the corneal parameter database includes a vectorized database (illumination path, illumination time, illumination power) and an imaged database.
In a preferred embodiment, in the training method, a countermeasure neural network is adopted for the laser treatment path change characteristic which is obvious, and a BP neural network or a CNN network is adopted for the laser time and power change characteristic which is dominant. Relevant judgment thresholds can be preset to judge whether the laser treatment path change characteristics are obvious or not and whether the laser treatment path change characteristics are dominant or not.
Exemplary embodiments and application examples of the present invention are further described below with reference to the accompanying drawings.
Referring to fig. 1 to 4, in an embodiment, an intelligent treatment system for a photo-crosslinking operation includes a corneal topography image acquisition system, a photo-crosslinking operation guidance system, and a database of a corneal topography and a corneal photo-crosslinking operation parameter correspondence before treatment.
As shown in fig. 2, the corneal topography image capturing system includes a placido optical system for providing red LED light and an image capturing module for capturing a corneal topography image.
The photocrosslinking operation guide system is pre-trained to output parameters required for implementing photocrosslinking operation by comparing a corneal topography before operation with an expected corneal topography after operation, and the parameters respectively correspond to two systems of ultraviolet riboflavin crosslinking operation and infrared femtosecond laser noninvasive crosslinking operation.
The database corresponding to the corneal topographic map and the corneal photocrosslinking surgical parameters is used for recording the influence of the corneal crosslinking parameters on the corneal topographic map so as to pre-train and correct the photocrosslinking surgical guidance system.
The image is acquired by a corneal topography image acquisition system, the corneal topography of a patient is acquired, and then the corneal topography is input into a photo-crosslinking surgery guide system which is pre-trained by a database corresponding to the pre-treatment corneal topography and corneal photo-crosslinking surgery parameters, so that the parameters of the photo-crosslinking surgery are acquired.
In a preferred embodiment, the corneal topography image collecting system collects a corneal topography image again for a patient who has completed a photo-crosslinking operation, stores a pre-operation corneal topography image, an post-operation actual corneal topography image and input corneal crosslinking parameters into the database corresponding to the corneal photo-crosslinking operation parameters, and retrains the photo-crosslinking operation guiding system with the database.
In a preferred embodiment, as shown in fig. 4, the photocrosslinking surgical guidance system takes a CNN convolutional neural network model, inputs two corneal topography maps before the crosslinking operation and after the expected operation, and the convolutional neural network model outputs an irradiation path, an irradiation time and an irradiation power by performing a regression task.
In a more preferred embodiment, the activation functions of the hidden layer and the output layer of the photocrosslinking guiding system are both relus, and the activation function of the output layer is lreul, so as to ensure that negative value information is not lost.
In a preferred embodiment, the pre-treatment corneal topography and corneal photocrosslinking operation parameter correspondence database stores a plurality of corneal topography before and after photocrosslinking operation and laser crosslinking parameters based on pre-judgment in advance.
In a more preferred embodiment, the database corresponding to the corneal topography before and after the photocrosslinking operation is added to the corneal topography before and after the treatment and the corneal photocrosslinking operation parameter, and the photocrosslinking operation guidance system revises the model parameter after the database corresponding to the corneal topography before and after the treatment and the corneal photocrosslinking operation parameter is updated.
As shown in fig. 2, in a preferred embodiment, the image acquisition module mainly includes a corneal topography instrument host, a computer host, and a power indicator data acquisition card, the corneal topography instrument host includes a placido optical system, a three-dimensional motion platform, and a manual remote sensing, and the placido optical system is a red LED visible light module, a placido ring, a focusing CCD, an imaging CCD, or the like. The red LED adopts a light emitting diode with the wavelength of 760nm, adopts a light source with the light power not more than 0.42W,
according to the embodiment of the invention, in the artificial intelligence guide system for corneal crosslinking, the corneal topography image acquisition system is configured to clearly acquire the corneal topography image of a user and fully reflect corneal information; the light cross-linking operation guide system can output parameters required by the cross-linking operation by comparing a corneal topography before cross-linking with an expected post-operation corneal topography after pre-training, and the parameters respectively correspond to two systems of an ultraviolet riboflavin cross-linking operation and an infrared femtosecond laser noninvasive cross-linking operation; the database corresponding to the corneal photocrosslinking operation parameters and the corneal topographic map before treatment can record the influence of the corneal crosslinking parameters on the corneal topographic map, and play a role in revising the crosslinking guide system. The system can greatly improve the efficiency, accuracy and stability of the photo-crosslinking operation process, and reduce errors caused by subjective judgment of doctors.
In some application examples, the examinee should sit naturally on a measuring chair in front of the topographic map, the chin of the examinee should be placed on a mandible support, the examinee should not blink as much as possible, the obtained topographic map area is enlarged, the high-precision corneal topographic map of the examinee is obtained, then the corneal topographic map is input into a photo-crosslinking operation guiding system which is pre-trained by a database corresponding to the corneal topographic map and the corneal photo-crosslinking operation parameters before treatment, and photo-crosslinking operation parameter vectors including but not limited to an irradiation path, an irradiation time and an irradiation power are obtained. Then the operator can select corresponding parameters on the cross-linking instrument to perform the surgery according to the output vector of the photo-cross-linking surgery guidance system.
After the operation is finished, the cornea topographic map image of the patient is collected again, the cornea topographic map before the operation, the actual cornea topographic map after the operation and the input crosslinking parameters are stored into a database corresponding to the cornea topographic map before treatment and the cornea optical crosslinking operation parameters, and the optical crosslinking operation guide system is retrained and the parameters such as convolution kernel and the like are updated along with the updating of the database.
The foregoing is a more detailed description of the invention in connection with specific/preferred embodiments and is not intended to limit the practice of the invention to those descriptions. It will be apparent to those skilled in the art that various substitutions and modifications can be made to the described embodiments without departing from the spirit of the invention, and these substitutions and modifications should be considered to fall within the scope of the invention.
Claims (3)
1. A method of establishing an intelligent treatment system for corneal photocrosslinking surgery, comprising:
a cornea information acquisition step, configured to acquire cornea information of a patient before and after an operation, including image information and parameter information of the cornea; the method specifically comprises the following steps: corneal topography, OCT anterior segment image; measuring parameter information of a corneal topographer and measuring parameter information of OCT (optical coherence tomography);
a database establishing step, namely establishing a corresponding database according to the cornea information acquired in the cornea information acquiring step;
the method comprises a step of training a light cross-linking operation guide system model, wherein pre-training is carried out to output parameters of the light cross-linking operation by comparing corneal information before corneal laser cross-linking with corneal information after cross-linking, and the parameters respectively correspond to two systems of an ultraviolet riboflavin cross-linking operation and an infrared femtosecond laser noninvasive cross-linking operation;
model feedback and correction, namely continuously amplifying a database and correcting a model by collecting corneal information before and after treatment;
in the training, a confrontation neural network is adopted obviously for the laser treatment path change characteristics, and a BP neural network or a CNN neural network is adopted for the laser time and power change characteristic dominance; and presetting related judgment threshold values to judge whether the laser treatment path change characteristics are obvious or not and whether the laser treatment path change characteristics are dominant or not.
2. The method of claim 1, wherein the training comprises:
training between the data set and the data set in the mode of training corneal topography parameters or OCT anterior segment image parameters and treatment parameters before and after BP neural network training treatment;
training images and data, wherein the mode is that a corneal topography or an OCT image is compared with treatment parameters before and after CNN neural network training;
training between images in a mode of utilizing an antagonistic neural network to train a preoperative and postoperative corneal topography or an OCT anterior segment image and a treatment parameter for imaging comparison;
the BP neural network adopts a back propagation algorithm of a feedforward network with two layers of sigmoid units, the input layer of the BP neural network is an m-dimensional vector and comprises a plurality of groups of data before and after treatment, and the output layer of the BP neural network comprises n-dimensional vectors of treatment parameters;
the CNN neural network inputs images before and after an operation to execute a regression task and outputs parameters for executing laser operation, and an activation function of a hidden layer and an activation function of an output layer are Lrel;
the depth of the antagonistic neural network adopts convolution to generate an antagonistic network DCGAN, and a convolution layer with step length is used for replacing a pooling layer in a discriminator D; batch Normalization (Batch Normalization) was used at G, D to help model convergence; in G, the activation function uses the ReLU function, and the last layer uses the tanh function, which is returned to the pixel value in use; in D, the activation functions all use leakage ReLU as the activation function.
3. The method according to claim 1 or 2, wherein the imaging of the irradiation path, irradiation time, irradiation power among the treatment parameters comprises: the laser path parameter image is two-dimensionalized into a closed curve or a non-closed curve, the time and power parameters are converted into energy density, the value size of the energy density is represented by the color shade, the treatment parameter image is a closed curve or a non-closed curve, and the laser crosslinking energy size is represented by the color density.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910622122.7A CN110338906B (en) | 2019-07-10 | 2019-07-10 | Intelligent treatment system for photocrosslinking operation and establishment method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910622122.7A CN110338906B (en) | 2019-07-10 | 2019-07-10 | Intelligent treatment system for photocrosslinking operation and establishment method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110338906A CN110338906A (en) | 2019-10-18 |
| CN110338906B true CN110338906B (en) | 2020-10-30 |
Family
ID=68174853
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910622122.7A Active CN110338906B (en) | 2019-07-10 | 2019-07-10 | Intelligent treatment system for photocrosslinking operation and establishment method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110338906B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110989199A (en) * | 2019-11-21 | 2020-04-10 | 明灏科技（北京）有限公司 | Calculating device for cornea moulding mirror power |
| CN112656507B (en) * | 2020-12-31 | 2022-08-26 | 江苏省人民医院（南京医科大学第一附属医院） | Method for constructing BP neural network model suitable for SMILE surgical cutting thickness prediction |
| CN113515561B (en) * | 2021-07-20 | 2023-12-12 | 苏州景昱医疗器械有限公司 | Parameter comparison method, device, electronic equipment and computer readable storage medium |
| CN114618090A (en) * | 2022-03-14 | 2022-06-14 | 重庆大学 | Energy control method of strong pulse laser xerophthalmia therapeutic instrument |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104382689A (en) * | 2014-12-02 | 2015-03-04 | 中国科学院光电研究院 | Femtosecond laser system using for imaging and operation at the same time |
| CN108577966A (en) * | 2018-03-20 | 2018-09-28 | 河南省洛阳正骨医院（河南省骨科医院） | Intelligent bone surgery system |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101259009B (en) * | 2008-03-26 | 2010-12-08 | 中国科学院光电技术研究所 | Corneal topography measuring instrument |
| CN102908118A (en) * | 2012-11-09 | 2013-02-06 | 苏州四海通仪器有限公司 | Corneal topography instrument |
| ITFI20120262A1 (en) * | 2012-11-28 | 2014-05-29 | Strumenti Oftalmici C S O S R L Costruzioni | PLACID DISC PROJECTOR WITH LED LIGHTING FOR A CORNEAL TOPOGRAPHY SYSTEM |
| CN103278771B (en) * | 2013-05-22 | 2015-06-17 | 广西电网公司电力科学研究院 | Identification method for abnormal state of three-phase asynchronous motor at industrial site |
| US10327953B2 (en) * | 2013-10-31 | 2019-06-25 | Optimedica Corporation | Laser eye surgery lens fragmentation |
| WO2016161481A1 (en) * | 2015-04-08 | 2016-10-13 | Wavefront Biometric Technologies Pty Limited | Multi-biometric authentication |
| EP3462412B1 (en) * | 2017-09-28 | 2019-10-30 | Siemens Healthcare GmbH | Determining a two-dimensional mammography data set |
| CN110163230A (en) * | 2018-06-15 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of image labeling method and device |
| CN109620223A (en) * | 2018-12-07 | 2019-04-16 | 北京工业大学 | A kind of rehabilitation of stroke patients system brain-computer interface key technology method |
| CN109977841A (en) * | 2019-03-20 | 2019-07-05 | 中南大学 | A kind of face identification method based on confrontation deep learning network |
| CN109994195B (en) * | 2019-03-22 | 2020-12-29 | 清华大学深圳研究生院 | Artificial intelligence guide system for corneal crosslinking |
| CN109966662B (en) * | 2019-04-30 | 2020-11-24 | 四川省肿瘤医院 | System for verifying radiation therapy dosage |
-
2019
- 2019-07-10 CN CN201910622122.7A patent/CN110338906B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104382689A (en) * | 2014-12-02 | 2015-03-04 | 中国科学院光电研究院 | Femtosecond laser system using for imaging and operation at the same time |
| CN108577966A (en) * | 2018-03-20 | 2018-09-28 | 河南省洛阳正骨医院（河南省骨科医院） | Intelligent bone surgery system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110338906A (en) | 2019-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110338906B (en) | Intelligent treatment system for photocrosslinking operation and establishment method | |
| US10485416B2 (en) | Ocular modeling methods and apparatus | |
| US11487136B2 (en) | Customized wavefront-guided methods, systems, and devices to correct higher-order aberration | |
| CN109994195B (en) | Artificial intelligence guide system for corneal crosslinking | |
| CN105451638B (en) | For the integrated OCT dioptrics meter systems of eye biometrics | |
| US10123687B2 (en) | Method for optimizing the selection of the IOL to be implanted in an eye | |
| US6585375B2 (en) | Method for producing an artificial ocular lense | |
| CN110179581B (en) | Correction method for preventing and controlling myopia and ametropia based on tension regulation mechanism | |
| JP2005525194A (en) | Manufacturing method of individual specification intraocular lens | |
| EP3781013B1 (en) | Customized ablation to correct visual ametropia | |
| CN113940812A (en) | Cornea center positioning method for excimer laser cornea refractive surgery | |
| RU2643417C1 (en) | Method for positioning the toric intraocular lens during implantation | |
| US6610048B1 (en) | Prolate shaped corneal reshaping | |
| CN115462955A (en) | Accurate corneal cross-linking astigmatism correction system guided by corneal topography imaging | |
| WO2001024728A2 (en) | Prolate shaped corneal reshaping | |
| US20020115988A1 (en) | Keratometric to apical radius conversion | |
| CN115836838A (en) | Diopter accurate evaluation method and application | |
| CN106420110A (en) | Pediatric intraocular lens degree prediction method | |
| CN113240711B (en) | Real-time individual crosslinking system | |
| CN110960349B (en) | Prediction method for diopter adjustment value of SMILE refractive surgery | |
| KR102460451B1 (en) | Method for measuring anomalies of refraction using a reflection image of pupil in visible light | |
| WO2002056791A2 (en) | Keratometric to apical radius conversion | |
| CN117138244A (en) | Self-adaptive near amblyopia therapeutic instrument | |
| Amer et al. | Aberrometry from astronomy to vision science | |
| CN113096074A (en) | Personalized cornea crosslinking system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |