Focus detection method and system based on image recognition

Abstract

The invention discloses a focus detection method and system based on image recognition, and the method comprises the steps: obtaining a to-be-detected medical image, inputting the medical image into a pre-trained focus region positioning model, and determining a preliminary focus region containing a focus in the to-be-detected medical image; obtaining a preliminary lesion area image of the preliminary lesion area, and inputting the preliminary lesion area image into a pre-trained lesion area extraction model to obtain a target lesion area image of the lesion; and inputting the target lesion area image into a pre-trained lesion recognition model to obtain a lesion type in the to-be-detected medical image. The problem of low accuracy during focus detection in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a lesion detection method and system based on image recognition. Background Art

[0002] Nowadays, medical imaging diagnosis mainly relies on traditional methods such as subjective visual analysis and manual annotation by doctors. Doctors use their experience and clinical knowledge combined with image features to identify and diagnose lesions. However, manual analysis and annotation of images is not only time-consuming, but also doctors are easily fatigued when faced with a large number of images, which may lead to missed diagnoses or misdiagnoses.

[0003] Due to these limitations, medical image recognition has gradually been introduced to improve the accuracy and efficiency of lesion detection with the help of deep learning and computer vision technology. By adopting image recognition technology, the burden on medical staff can be significantly reduced and the objectivity and consistency of diagnosis can be improved.

[0004] Currently, when it comes to lesion detection, medical images are directly tested to identify lesion information. However, when the lesion in the image is small, there are many interference factors, and the area and specific type of the lesion need to be identified at one time, which may lead to inaccurate lesion detection. Summary of the invention

[0005] In view of this, an object of the present invention is to provide a lesion detection method and system based on image recognition, aiming to solve the problem of low accuracy in lesion detection in the prior art.

[0006] An object of the present invention is to provide a lesion detection method based on image recognition, which is used to detect lesions in medical images. The method comprises: Acquire a medical image to be detected, and input the medical image into a pre-trained lesion region localization model to determine a preliminary lesion region containing a lesion in the medical image to be detected; Acquire a preliminary lesion region image of a preliminary lesion region, and input the preliminary lesion region image into a pre-trained lesion region extraction model to obtain a target lesion region image of the lesion; The target lesion area image is input into a pre-trained lesion recognition model to obtain the lesion type in the medical image to be detected.

[0007] Furthermore, in the above-mentioned lesion detection method based on image recognition, the training process of the lesion area positioning model includes: Collecting historical medical images containing lesions to obtain a training set and a validation set of the lesion area localization model; Constructing a first detection algorithm based on preset parameters, and using the first detection algorithm for training according to the training set and the validation set of the lesion area positioning model; Until the performance of the lesion area localization model meets the preset standard, the lesion area localization model is trained to obtain the lesion area localization model.

[0008] Furthermore, in the above-mentioned lesion detection method based on image recognition, the training process of the lesion area extraction model includes: Collecting historical preliminary lesion area images to obtain a training set and a validation set for the lesion area extraction model; Constructing a second detection algorithm based on preset parameters, and using the second detection algorithm for training according to the training set and the validation set of the lesion area extraction model; Until the performance of the lesion region extraction model meets a preset standard, the lesion region extraction model is trained to obtain the lesion region extraction model.

[0009] Furthermore, in the above-mentioned lesion detection method based on image recognition, the training process of the lesion recognition model includes: Construct a preset convolutional neural network, and collect a preset number of historical target lesion area images and corresponding lesion types as training samples; The target lesion area image containing the lesion and the corresponding lesion type are respectively used as the input and output of the convolutional neural network, and the convolutional neural network is deep-learned and trained until the recognition result output by the convolutional neural network meets the set accuracy, thereby obtaining the lesion recognition model.

[0010] Furthermore, in the above-mentioned lesion detection method based on image recognition, before the step of collecting historical medical images containing lesions to obtain a training set and a verification set of the lesion area localization model, the step further includes: Acquire historical medical images, and respectively extract lesion area images and non-lesion area images from the historical medical images; The lesion area image is randomly rotated, non-uniformly scaled, and distorted to obtain a first lesion area image; A target historical medical image is determined according to the first lesion area image and the non-lesion area image.

[0011] Furthermore, in the above-mentioned lesion detection method based on image recognition, after the step of acquiring historical medical images and respectively extracting lesion area images and non-lesion area images from the historical medical images, the step further includes: Acquiring contour information of the current lesion region image, and determining a diffusion law of the lesion region according to the lesion type of the lesion in the lesion region image; Performing contour expansion on the lesion region image according to the contour information and the diffusion law to obtain an expanded lesion region image; The center point of the lesion area image is obtained, a coordinate system is established with the center point as the coordinate origin, the features in the lesion area image are filled into the extended lesion area image in the same quadrant to obtain a target extended lesion area image, and then the target historical medical image is determined together with the non-lesion area image.

[0012] Furthermore, in the above-mentioned lesion detection method based on image recognition, the step of filling the features in the lesion area image into the extended lesion area image in the same quadrant to obtain the target extended lesion area image comprises: Acquire a rectangular area enclosing the extended lesion area image, and perform grid division on the rectangular area; Acquire a central grid of the lesion region image in the same quadrant, and determine a target central grid corresponding to the central grid in the extended lesion region image; Based on the positional relationship between the central grid and other grids, and the target central grid, the features of the other grids are filled into the grids without features in the extended lesion region image to obtain the target extended lesion region image.

[0013] Another object of the present invention is to provide a lesion detection system based on image recognition, for detecting lesions in medical images, the system comprising: An acquisition module, used for acquiring a medical image to be detected, and inputting the medical image into a pre-trained lesion region localization model to determine a preliminary lesion region containing a lesion in the medical image to be detected; An extraction module, used for acquiring a preliminary lesion region image of a preliminary lesion region, and inputting the preliminary lesion region image into a pre-trained lesion region extraction model to obtain a target lesion region image of the lesion; The detection module is used to input the target lesion area image into a pre-trained lesion recognition model to obtain the lesion type in the medical image to be detected.

[0014] Another object of the present invention is to provide a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of any one of the methods described above.

[0015] Another object of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.

[0016] The present invention uses a trained lesion region positioning model to preliminarily locate the lesion region, preliminarily identifies the area where the lesion may exist, and then uses a lesion region extraction model to lock the specific location of the lesion with a smaller lesion range when encountering a lesion with a smaller lesion range, and finally uses a lesion recognition model to identify the type of the lesion, thereby realizing the detection of the lesion region and type, and due to the multi-level recognition method adopted, when identifying the lesion type, the interference of other factors is avoided, and the accuracy of lesion recognition is improved. The problem of low accuracy in lesion detection in the prior art is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flow chart of a lesion detection method based on image recognition in the first embodiment of the present invention; Figure 2 This is a structural block diagram of a lesion detection system based on image recognition in the third embodiment of the present invention.

[0018] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0019] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0020] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0022] The following will describe in detail how to accurately detect small lesions in medical images in conjunction with specific embodiments and drawings.

[0023] Embodiment 1 See also Figure 1, shown is a lesion detection method based on image recognition in the first embodiment of the present invention, and the method includes steps S10 to S12.

[0024] Step S10, obtaining a medical image to be detected, and inputting the medical image into a pre-trained lesion region localization model to determine a preliminary lesion region containing a lesion in the medical image to be detected.

[0025] Among them, the medical image is an image containing lesions collected by medical equipment, such as a CT image. The lesion area localization model is used to roughly estimate the area of ​​the lesion in the medical image and preliminarily identify the area where the lesion may exist.

[0026] Specifically, a relatively lightweight target detection algorithm (such as YOLOv5, SSD, etc.) can be selected to quickly identify areas where lesions may exist. For example, when training the lesion area localization model, historical medical images containing lesions are collected to obtain a training set and a validation set of the lesion area localization model; a first detection algorithm (such as YOLOv5, SSD, etc.) is constructed based on preset parameters, and the first detection algorithm is used for training according to the training set and the validation set of the lesion area localization model, wherein a suitable loss function can be selected for training (such as a combination of cross entropy loss and bounding box regression loss), and optimizers such as Adam and SGD are used to adjust hyperparameters such as the learning rate; multiple cycles of iterative training are performed on the training set, and the performance of the model is monitored using the validation set to prevent overfitting, until the performance of the lesion area localization model meets the preset standard, so as to train the lesion area localization model.

[0027] Step S11, obtaining a preliminary lesion region image of the preliminary lesion region, and inputting the preliminary lesion region image into a pre-trained lesion region extraction model to obtain a target lesion region image of the lesion.

[0028] Specifically, after determining the preliminary lesion area, a preliminary lesion area image is captured, and a lesion area extraction model is used to obtain a more accurate target lesion area image of the lesion. Specifically, historical preliminary lesion area images are collected to obtain a training set and a validation set of the lesion area extraction model; a second detection algorithm is constructed based on preset parameters, and the second detection algorithm is used for training according to the training set and the validation set of the lesion area extraction model; until the performance of the lesion area extraction model meets the preset standard, the lesion area extraction model is trained to obtain the lesion area extraction model, wherein the lesion area extraction model aims to accurately identify the specific contour of the lesion, and network models such as U-Net and Mask R-CNN can be used to collect a large number of training data sets, and the lesion contour is accurately annotated, usually using polygons or other forms to ensure that the annotation accurately reflects the boundary of the lesion, and finally training is performed to obtain the final lesion area extraction model.

[0029] In addition, in some optional embodiments of the present invention, since the extracted preliminary lesion area images are of different sizes, in order to ensure that the image size of the data set used for training is consistent while maintaining the original features, the obtained preliminary lesion area images can be normalized, such as uniformly scaled.

[0030] Step S12, inputting the target lesion area image into a pre-trained lesion recognition model to obtain the lesion type in the medical image to be detected.

[0031] Among them, the lesion recognition model grasps the internal logic of lesion type recognition. By inputting the target lesion area image into the pre-trained lesion recognition model, the lesion type in the medical image to be detected can be accurately identified.

[0032] Specifically, a preset convolutional neural network is constructed, and a preset number of historical target lesion area images and corresponding lesion types are collected as training samples; the target lesion area images containing lesions and the corresponding lesion types are used as input and output of the convolutional neural network, respectively, and the convolutional neural network is deep learned and trained until the recognition result output by the convolutional neural network meets the set accuracy, thereby obtaining a lesion recognition model, wherein the more detailed process of selecting and training the neural network can be understood by those skilled in the art and will not be elaborated here.

[0033] In summary, a lesion detection method based on image recognition in the above-mentioned embodiment of the present invention first uses the trained lesion area positioning model to preliminarily locate the lesion area, preliminarily identifies the area where the lesion may exist, and then uses the lesion area extraction model to lock the specific location of the lesion with a smaller lesion range when encountering a lesion with a smaller lesion range, and finally uses the lesion recognition model to identify the type of lesion, thereby realizing the detection of the lesion area and type, and due to the multi-level recognition method adopted, when identifying the lesion type, the interference of other factors is avoided, and the accuracy of lesion recognition is improved. The problem of low accuracy in lesion detection in the prior art is solved.

[0034] Embodiment 2 This embodiment also proposes a lesion detection method based on image recognition. The difference between the lesion detection method based on image recognition in this embodiment and the lesion detection method based on image recognition in Embodiment 1 is that: Before the step of acquiring historical medical images containing lesions to obtain the training set and the verification set of the lesion area localization model, the following steps are also included: Acquire historical medical images, and respectively extract lesion area images and non-lesion area images from the historical medical images; The lesion area image is randomly rotated, non-uniformly scaled, and distorted to obtain a first lesion area image; A target historical medical image is determined according to the first lesion area image and the non-lesion area image.

[0035] In practice, image samples with lesions are difficult to obtain. In order to ensure the accuracy of model training, it is necessary to ensure sufficient diverse sample data. Therefore, in order to greatly enhance the diversity of the data set, existing lesion area images in existing historical medical images are extracted, the lesions are randomly rotated, different orientations are simulated, the lesions are non-uniformly scaled, lesions of different sizes are adapted, and affine transformation or perspective transformation is used to change the shape of the lesions, simulate the variation of biological tissues, and change the lesion area by geometric transformation. Finally, the lesion area image is synthesized with the extracted non-lesion area image to obtain the corresponding expanded target historical medical image.

[0036] In addition, in practice, the above-mentioned method of improving data diversity has a certain degree of randomness. In order to improve the diversity of sample data in a targeted manner, after extracting the lesion area, the diffusion law of the lesion area can be determined according to the current lesion type, so that the current lesion area can be diffused to expand the diversity of samples. Specifically, the contour information of the current lesion area image is obtained, and the diffusion law of the lesion area is determined according to the lesion type of the lesion in the lesion area image. The contour of the lesion area image is extended according to the contour information and the diffusion law to obtain an extended lesion area image. At this time, the contour features of the lesion area are extended. In order to further and completely expand the features of the lesion area Fill, obtain the center point of the lesion area image, establish a coordinate system with the center point as the coordinate origin, fill the features in the lesion area image into the extended lesion area image in the same quadrant to obtain a target extended lesion area image, and then determine the target historical medical image with the non-lesion area image, wherein the center point can be the geometric center of the lesion area image, such as the centroid, and establish a coordinate system based on the center point, wherein the extended lesion area image and various parts of the lesion area image are respectively distributed in different quadrants of the coordinate system, and the features of the lesion area image in the same quadrant are randomly filled into the part of the extended lesion area image that does not contain the features, thereby obtaining the final target extended lesion area image.

[0037] In addition, in some optional embodiments of the present invention, the step of filling the features in the lesion area image into the extended lesion area image in the same quadrant to obtain a target extended lesion area image includes: Acquire a rectangular area enclosing the extended lesion area image, and perform grid division on the rectangular area; Acquire a central grid of the lesion region image in the same quadrant, and determine a target central grid corresponding to the central grid in the extended lesion region image; Based on the positional relationship between the central grid and other grids, and the target central grid, the features of the other grids are filled into the grids without features in the extended lesion region image to obtain the target extended lesion region image.

[0038] Among them, when performing feature migration and filling, the lesion area is divided into multiple small grids by grid division, and the migration is performed according to the pixel features in each grid. This method can not only retain the continuity of local features, but also use spatial positions in the extended lesion area for more flexible feature migration. In order to effectively migrate features, the key is how to correspond the grids according to a specific mapping strategy. In the embodiment of the present invention, a center point mapping strategy is adopted. The center mapping is based on the relationship between the center point of the lesion area and the center point of the extended area. The corresponding target center grid or center point position is found in the extended area, and each grid of the lesion area is mapped to the extended lesion area according to the relative position of the grid.

[0039] Exemplarily, the features of the previous lesion area may be retained or not retained during filling, that is, the features may be used to fill the entire extended lesion area, or the features may be filled after removing the overlapping parts with the lesion area. In some optional embodiments of the present invention, a rectangular area may be enclosed in the lesion area, which can correspond to the rectangular area of ​​the extended lesion area, thereby facilitating the filling between grid features.

[0040] In summary, a lesion detection method based on image recognition in the above-mentioned embodiment of the present invention first uses the trained lesion area positioning model to preliminarily locate the lesion area, preliminarily identifies the area where the lesion may exist, and then uses the lesion area extraction model to lock the specific location of the lesion with a smaller lesion range when encountering a lesion with a smaller lesion range, and finally uses the lesion recognition model to identify the type of lesion, thereby realizing the detection of the lesion area and type, and due to the multi-level recognition method adopted, when identifying the lesion type, the interference of other factors is avoided, and the accuracy of lesion recognition is improved. The problem of low accuracy in lesion detection in the prior art is solved.

[0041] Embodiment 3 See also Figure 2 , shown is a lesion detection system based on image recognition proposed in the third embodiment of the present invention, which is used to detect lesions in medical images. The system includes: An acquisition module 100 is used to acquire a medical image to be detected, and input the medical image into a pre-trained lesion region localization model to determine a preliminary lesion region containing a lesion in the medical image to be detected; An extraction module 200 is used to obtain a preliminary lesion region image of a preliminary lesion region, and input the preliminary lesion region image into a pre-trained lesion region extraction model to obtain a target lesion region image of the lesion; The detection module 300 is used to input the target lesion area image into a pre-trained lesion recognition model to obtain the lesion type in the medical image to be detected.

[0042] Furthermore, in the above-mentioned lesion detection system based on image recognition, the training process of the lesion area positioning model includes: Collecting historical medical images containing lesions to obtain a training set and a validation set of the lesion area localization model; Constructing a first detection algorithm based on preset parameters, and using the first detection algorithm for training according to the training set and the validation set of the lesion area positioning model; Until the performance of the lesion area localization model meets the preset standard, the lesion area localization model is trained to obtain the lesion area localization model.

[0043] Furthermore, in the above-mentioned lesion detection system based on image recognition, the training process of the lesion region extraction model includes: Collecting historical preliminary lesion area images to obtain a training set and a validation set for the lesion area extraction model; Constructing a second detection algorithm based on preset parameters, and using the second detection algorithm for training according to the training set and the validation set of the lesion area extraction model; Until the performance of the lesion region extraction model meets a preset standard, the lesion region extraction model is trained to obtain the lesion region extraction model.

[0044] Furthermore, in the above-mentioned lesion detection system based on image recognition, the training process of the lesion recognition model includes: Construct a preset convolutional neural network, and collect a preset number of historical target lesion area images and corresponding lesion types as training samples; The target lesion area image containing the lesion and the corresponding lesion type are respectively used as the input and output of the convolutional neural network, and the convolutional neural network is deep-learned and trained until the recognition result output by the convolutional neural network meets the set accuracy, thereby obtaining the lesion recognition model.

[0045] Furthermore, in the above-mentioned lesion detection system based on image recognition, before the step of collecting historical medical images containing lesions to obtain a training set and a verification set of the lesion area localization model, the step further includes: Acquire historical medical images, and respectively extract lesion area images and non-lesion area images from the historical medical images; The lesion area image is randomly rotated, non-uniformly scaled, and distorted to obtain a first lesion area image; A target historical medical image is determined according to the first lesion area image and the non-lesion area image.

[0046] Furthermore, in the above-mentioned lesion detection system based on image recognition, after the step of acquiring historical medical images and respectively extracting lesion area images and non-lesion area images from the historical medical images, the step further includes: Acquiring contour information of the current lesion region image, and determining a diffusion law of the lesion region according to the lesion type of the lesion in the lesion region image; Performing contour expansion on the lesion region image according to the contour information and the diffusion law to obtain an expanded lesion region image; The center point of the lesion area image is obtained, a coordinate system is established with the center point as the coordinate origin, the features in the lesion area image are filled into the extended lesion area image in the same quadrant to obtain a target extended lesion area image, and then the target historical medical image is determined together with the non-lesion area image.

[0047] Furthermore, in the above-mentioned lesion detection system based on image recognition, the step of filling the features in the lesion area image into the extended lesion area image in the same quadrant to obtain the target extended lesion area image comprises: Acquire a rectangular area enclosing the extended lesion area image, and perform grid division on the rectangular area; Acquire a central grid of the lesion region image in the same quadrant, and determine a target central grid corresponding to the central grid in the extended lesion region image; Based on the positional relationship between the central grid and other grids, and the target central grid, the features of the other grids are filled into the grids without features in the extended lesion region image to obtain the target extended lesion region image.

[0048] The functions or operation steps implemented when the above modules are executed are substantially the same as those in the above method embodiments, and will not be repeated here.

[0049] Embodiment 4 Another aspect of the present invention further provides a readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the steps of the method described in any one of the above-mentioned embodiments 1 to 2 are implemented.

[0050] Embodiment 5 Another aspect of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in any one of the above-mentioned embodiments 1 to 2 when executing the program.

[0051] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0052] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be specifically implemented in any computer-readable storage medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, a "computer-readable storage medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0053] More specific examples (a non-exhaustive list) of computer-readable storage media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable storage medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0054] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0055] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0056] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A lesion detection method based on image recognition, characterized in that: For detecting lesions in a medical image, the method comprises: Acquire a medical image to be detected, and input the medical image into a pre-trained lesion region localization model to determine a preliminary lesion region containing a lesion in the medical image to be detected; Acquire a preliminary lesion region image of a preliminary lesion region, and input the preliminary lesion region image into a pre-trained lesion region extraction model to obtain a target lesion region image of the lesion; The target lesion area image is input into a pre-trained lesion recognition model to obtain the lesion type in the medical image to be detected.

2. The method for detecting lesions based on image recognition according to claim 1, characterized in that: The training process of the lesion area localization model includes: Collecting historical medical images containing lesions to obtain a training set and a validation set of the lesion area localization model; Constructing a first detection algorithm based on preset parameters, and using the first detection algorithm for training according to the training set and the validation set of the lesion area positioning model; Until the performance of the lesion area localization model meets the preset standard, the lesion area localization model is trained to obtain the lesion area localization model.

3. The lesion detection method based on image recognition according to claim 2, characterized in that: The training process of the lesion region extraction model includes: Collecting historical preliminary lesion area images to obtain a training set and a validation set for the lesion area extraction model; Constructing a second detection algorithm based on preset parameters, and using the second detection algorithm for training according to the training set and the validation set of the lesion area extraction model; Until the performance of the lesion region extraction model meets a preset standard, the lesion region extraction model is trained to obtain the lesion region extraction model.

4. The method for detecting lesions based on image recognition according to claim 1, characterized in that: The training process of the lesion recognition model includes: Construct a preset convolutional neural network, and collect a preset number of historical target lesion area images and corresponding lesion types as training samples; The target lesion area image containing the lesion and the corresponding lesion type are respectively used as the input and output of the convolutional neural network, and the convolutional neural network is deep-learned and trained until the recognition result output by the convolutional neural network meets the set accuracy, thereby obtaining the lesion recognition model.

5. The method for lesion detection based on image recognition according to claim 1, characterized in that: Before the step of collecting historical medical images containing lesions to obtain the training set and the verification set of the lesion area localization model, the following steps are also included: Acquire historical medical images, and respectively extract lesion area images and non-lesion area images from the historical medical images; The lesion area image is randomly rotated, non-uniformly scaled, and distorted to obtain a first lesion area image; A target historical medical image is determined according to the first lesion area image and the non-lesion area image.

6. The method for detecting lesions based on image recognition according to claim 5, characterized in that: After the step of acquiring the historical medical image and respectively extracting the lesion area image and the non-lesion area image from the historical medical image, the method further comprises: Acquiring contour information of the current lesion region image, and determining a diffusion law of the lesion region according to the lesion type of the lesion in the lesion region image; Performing contour expansion on the lesion region image according to the contour information and the diffusion law to obtain an expanded lesion region image; The center point of the lesion area image is obtained, a coordinate system is established with the center point as the coordinate origin, the features in the lesion area image are filled into the extended lesion area image in the same quadrant to obtain a target extended lesion area image, and then the target historical medical image is determined together with the non-lesion area image.

7. The method for detecting lesions based on image recognition according to claim 6, characterized in that: The step of filling the features in the lesion area image into the extended lesion area image in the same quadrant to obtain a target extended lesion area image comprises: Acquire a rectangular area enclosing the extended lesion area image, and perform grid division on the rectangular area; Acquire a central grid of the lesion region image in the same quadrant, and determine a target central grid corresponding to the central grid in the extended lesion region image; Based on the positional relationship between the central grid and other grids, and the target central grid, the features of the other grids are filled into the grids without features in the extended lesion region image to obtain the target extended lesion region image.

8. A lesion detection system based on image recognition, characterized in that: For detecting lesions in medical images, the system comprises: An acquisition module, used for acquiring a medical image to be detected, and inputting the medical image into a pre-trained lesion region localization model to determine a preliminary lesion region containing a lesion in the medical image to be detected; An extraction module, used for acquiring a preliminary lesion region image of a preliminary lesion region, and inputting the preliminary lesion region image into a pre-trained lesion region extraction model to obtain a target lesion region image of the lesion; The detection module is used to input the target lesion area image into a pre-trained lesion recognition model to obtain the lesion type in the medical image to be detected.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the program.

Images