技术领域technical field
本发明涉及医学与计算机技术领域，特别涉及一种基于组合分类器的白内障眼底图像分类方法及分类装置。The invention relates to the fields of medicine and computer technology, in particular to a cataract fundus image classification method and a classification device based on a combined classifier.
背景技术Background technique
相关技术中，首先对白内障眼底图像进行预处理，在预处理后的眼底图像中提取了40个特征，包括1个亮度特征、24个灰度共生矩阵特征和15个灰度-梯度共生矩阵特征。分类器采用的是一个两层的BP神经网络。分类器的输入神经元个数是40个，隐含层神经元个数是10，输出层神经元个数是4个。最后分类的正确率达到了82.5％。In related technologies, the cataract fundus image is first preprocessed, and 40 features are extracted from the preprocessed fundus image, including 1 brightness feature, 24 gray-scale co-occurrence matrix features and 15 gray-gradient co-occurrence matrix features . The classifier uses a two-layer BP neural network. The number of input neurons of the classifier is 40, the number of hidden layer neurons is 10, and the number of output layer neurons is 4. The correct rate of final classification reached 82.5%.
白内障眼底图像的分类主要包括预处理、特征提取和分类器三部分。在图像预处理阶段主要包括提取RGB彩色空间的G通道图像、改进的高低帽算法和三边滤波器处理三部分。经过预处理后的图像变得更清晰，也滤掉了大部分的噪声，更适合后续的特征提取。特征提取主要包含亮度特征和纹理特征提取,其中纹理特征提取包括灰度共生矩阵和灰度-梯度共生矩阵提取两部分。由于视神经盘是眼底图像中最亮的部分，并且它也是将白内障进行分类的一个重要标志，所以以此作为白内障分类的一个标准。纹理特征主要包含灰度共生矩阵特征和灰度-梯度共生矩阵两类。灰度共生矩阵表示的是图像主要的纹理信息，但它不包含图像的边缘信息；而灰度-梯度共生矩阵恰恰弥补了灰度共生矩阵的这个缺陷。分类器采用的是一个两层的BP(Back Propagation，神经网络)神经网络，包含输入层、隐含层和输出层。这里输入层的个数是40，输出层的个数是4，隐含层的个数根据实验结果调整为10，此时的实验效果比较好。The classification of cataract fundus images mainly includes three parts: preprocessing, feature extraction and classifier. In the stage of image preprocessing, it mainly includes three parts: extracting G channel image in RGB color space, improving high-low hat algorithm and three-sided filter processing. The preprocessed image becomes clearer, and most of the noise is filtered out, which is more suitable for subsequent feature extraction. Feature extraction mainly includes brightness feature and texture feature extraction, and texture feature extraction includes gray level co-occurrence matrix and gray level-gradient co-occurrence matrix extraction. Since the optic disc is the brightest part of the fundus image, and it is also an important sign for cataract classification, it is used as a standard for cataract classification. Texture features mainly include gray level co-occurrence matrix feature and gray level-gradient co-occurrence matrix. The gray level co-occurrence matrix represents the main texture information of the image, but it does not contain the edge information of the image; the gray level-gradient co-occurrence matrix just makes up for this defect of the gray level co-occurrence matrix. The classifier adopts a two-layer BP (Back Propagation, neural network) neural network, including an input layer, a hidden layer and an output layer. Here, the number of input layers is 40, the number of output layers is 4, and the number of hidden layers is adjusted to 10 according to the experimental results. The experimental effect at this time is better.
然而，相关技术虽然能对白内障眼底图像进行分类，但分类的正确率还不能达到满意的要求，在特征提取和分类方法方面还待进一步加强。However, although related technologies can classify cataract fundus images, the correct rate of classification cannot meet the satisfactory requirements, and the feature extraction and classification methods need to be further strengthened.
发明内容Contents of the invention
本申请是基于发明人对以下问题的认识和发现作出的：This application is made based on the inventor's recognition and discovery of the following problems:
目前，每一位进行眼科检查的患者都要首先拍摄眼底照片，从而眼科医生可以通过眼底照片中眼底主要部位的清晰程度初步判断患者是否患有白内障及所患白内障疾病的严重程度。其中，当眼底图像中的视神经盘、大小血管均清晰可见的属于正常眼底；视神经盘模糊不清的属于重度白内障；大小血管不可见的属于中度白内障；仅小血管不可见的属于轻度白内障。At present, every patient undergoing an eye examination must first take a fundus photo, so that the ophthalmologist can initially judge whether the patient has cataract and the severity of the cataract disease through the clarity of the main parts of the fundus in the fundus photo. Among them, when the optic disc and large and small blood vessels are clearly visible in the fundus image, it belongs to normal fundus; if the optic disc is blurred, it belongs to severe cataract; if the large and small blood vessels are not visible, it belongs to moderate cataract; .
然而，现在的研究现状是医生凭经验看大量眼底图像并进行分类，而相应的做图像处理的计算机研究人员的工作更多的关注于图像本身的处理，很少有人基于专业医生看眼底图像的经验来做眼底图片的自动分级划分工作，而此项工作对具体应用和医学研究都有很大意义。一方面可以为医生减轻工作量，提高工作效率；另一方面可以对白内障术后恢复情况作很好地估计，患者和医生可以提前预知术后患者视力恢复状况。同时随着人口老龄化和人们对自身健康重视程度的增加，拍摄眼底照片人数也会相应地增加，未来会有更多的眼底图像需要进行分级，这种眼底图像数据库会逐渐地发展成为大数据，这种自动分级系统会大大减少人力资源，也会使工作效率有很大的提高。However, the current research status is that doctors look at a large number of fundus images and classify them based on experience, while the corresponding computer researchers who do image processing focus more on the processing of the images themselves, and few people rely on professional doctors to look at fundus images. Experience to do the automatic grading and division of fundus images, and this work has great significance for specific applications and medical research. On the one hand, it can reduce the workload of doctors and improve work efficiency; on the other hand, it can make a good estimate of the recovery after cataract surgery, and patients and doctors can predict the visual recovery status of patients after surgery in advance. At the same time, as the population ages and people pay more attention to their own health, the number of people taking fundus photos will increase accordingly. In the future, more fundus images will need to be graded. This fundus image database will gradually develop into big data , this automatic grading system will greatly reduce human resources and greatly improve work efficiency.
本发明旨在至少在一定程度上解决上述相关技术中的技术问题之一。The present invention aims at solving one of the technical problems in the related art mentioned above at least to a certain extent.
为此，本发明的一个目的在于提出一种基于组合分类器的白内障眼底图像分类方法，该方法可以提高分类的精确度，更好地满足对于正确率的使用要求。Therefore, an object of the present invention is to propose a cataract fundus image classification method based on a combined classifier, which can improve the accuracy of classification and better meet the use requirements for accuracy.
本发明的另一个目的在于提出一种基于组合分类器的白内障眼底图像分类装置。Another object of the present invention is to propose a cataract fundus image classification device based on a combined classifier.
为达到上述目的，本发明一方面实施例提出了一种基于组合分类器的白内障眼底图像分类方法，包括以下步骤：获取白内障眼底图像；对所述白内障眼底图像进行预处理；分别通过小波变换、轮辅方法和纹理分析从所述处理后的白内障眼底图像中提取N组特征，N为正整数；分别通过支持向量机和BP神经网对所述N组特征进行预测分类，以获取M种预测分类结果，M为大于等于N的正整数；以及通过组合分类器对所述M种预测分类结果进行投票分类，以获取最终分类结果。In order to achieve the above object, an embodiment of the present invention proposes a cataract fundus image classification method based on a combined classifier, comprising the following steps: acquiring a cataract fundus image; preprocessing the cataract fundus image; The wheel-assisted method and texture analysis extract N groups of features from the processed cataract fundus image, and N is a positive integer; respectively predict and classify the N groups of features through support vector machines and BP neural networks to obtain M types of predictions. Classification results, M is a positive integer greater than or equal to N; and classifying the M predicted classification results by combining classifiers to obtain a final classification result.
根据本发明实施例提出的基于组合分类器的白内障眼底图像分类方法，在对白内障眼底图像进行预处理之后，通过小波变换、轮辅方法和纹理分析分别提取多组特征，其次通过支持向量机和BP神经网对多组特征进行预测分类，最后通过组合分类器进行投票分类，以获取最终分类结果，提高分类的精确度，更好地满足对于正确率的使用要求。According to the cataract fundus image classification method based on the combined classifier proposed in the embodiment of the present invention, after preprocessing the cataract fundus image, multiple groups of features are extracted respectively through wavelet transform, wheel-assisted method and texture analysis, and then through support vector machine and The BP neural network predicts and classifies multiple sets of features, and finally votes and classifies by combining classifiers to obtain the final classification result, improve the classification accuracy, and better meet the use requirements for the correct rate.
另外，根据本发明上述实施例的基于组合分类器的白内障眼底图像分类方法还可以具有如下附加的技术特征：In addition, the cataract fundus image classification method based on the combined classifier according to the above-mentioned embodiments of the present invention may also have the following additional technical features:
进一步地，在本发明的一个实施例中，所述N可以为3，所述M可以为6。Further, in an embodiment of the present invention, the N may be 3, and the M may be 6.
进一步地，在本发明的一个实施例中，所述预处理包括：提取RGB彩色空间和G通道图像；对所述G通道图像进行两次直方图均衡化处理；和/或通过高低帽算法和三边滤波器处理所述G通道图像。Further, in one embodiment of the present invention, the preprocessing includes: extracting the RGB color space and the G channel image; performing two histogram equalization processing on the G channel image; and/or using the high-low hat algorithm and A trilateral filter processes the G-channel image.
进一步地，在本发明的一个实施例中，采用所述组合分类器的投票法选取获得票数最多的预测分类结果作为所述最终分类结果。Further, in an embodiment of the present invention, the predicted classification result with the most votes is selected as the final classification result by using the voting method of the combination classifier.
进一步地，在本发明的一个实施例中，所述N组特征中每个特征包括亮度特征和纹理特征，其中，所述纹理特征包括灰度共生矩阵和灰度-梯度共生矩阵。Further, in an embodiment of the present invention, each feature in the N groups of features includes a brightness feature and a texture feature, wherein the texture feature includes a grayscale co-occurrence matrix and a grayscale-gradient co-occurrence matrix.
本发明另一方面实施例提出了一种基于组合分类器的白内障眼底图像分类装置，包括：获取模块，用于获取白内障眼底图像；处理模块，用于对所述白内障眼底图像进行预处理；提取模块，用于分别通过小波变换、轮辅方法和纹理分析从所述处理后的白内障眼底图像中提取N组特征，N为正整数；第一分类模块，用于分别通过小波变换、轮辅方法和纹理分析从所述处理后的白内障眼底图像中提取N组特征，N为正整数；以及第二分类模块，用于通过组合分类器对所述M种预测分类结果进行投票分类，以获取最终分类结果。Another embodiment of the present invention proposes a cataract fundus image classification device based on a combination classifier, including: an acquisition module, used to acquire a cataract fundus image; a processing module, used to preprocess the cataract fundus image; extract A module for extracting N groups of features from the processed cataract fundus image through wavelet transform, wheel-assisted method and texture analysis respectively, where N is a positive integer; the first classification module is used for respectively through wavelet transform and wheel-assisted method and texture analysis to extract N groups of features from the processed cataract fundus image, where N is a positive integer; and a second classification module, which is used to vote and classify the M prediction classification results by combining classifiers to obtain the final classification results.
根据本发明实施例提出的基于组合分类器的白内障眼底图像分类装置，在对白内障眼底图像进行预处理之后，通过小波变换、轮辅方法和纹理分析分别提取多组特征，其次通过支持向量机和BP神经网对多组特征进行预测分类，最后通过组合分类器进行投票分类，以获取最终分类结果，提高分类的精确度，更好地满足对于正确率的使用要求。According to the cataract fundus image classification device based on the combination classifier proposed in the embodiment of the present invention, after preprocessing the cataract fundus image, multiple groups of features are extracted respectively through wavelet transform, wheel-assisted method and texture analysis, and then through support vector machine and The BP neural network predicts and classifies multiple sets of features, and finally votes and classifies by combining classifiers to obtain the final classification result, improve the classification accuracy, and better meet the use requirements for the correct rate.
另外，根据本发明上述实施例的基于组合分类器的白内障眼底图像分类装置还可以具有如下附加的技术特征：In addition, the cataract fundus image classification device based on the combined classifier according to the above-mentioned embodiments of the present invention may also have the following additional technical features:
进一步地，在本发明的一个实施例中，所述N可以为3，所述M可以为6。Further, in an embodiment of the present invention, the N may be 3, and the M may be 6.
进一步地，在本发明的一个实施例中，所述处理模块包括：提取单元，用于提取RGB彩色空间和G通道图像；处理单元，用于对所述G通道图像进行两次直方图均衡化处理和/或通过高低帽算法和三边滤波器处理所述G通道图像。Further, in one embodiment of the present invention, the processing module includes: an extraction unit for extracting an RGB color space and a G channel image; a processing unit for performing two histogram equalization on the G channel image processing and/or processing the G-channel image through a high-low hat algorithm and a trilateral filter.
进一步地，在本发明的一个实施例中，所述第二分类模块具体用于采用所述组合分类器的投票法选取获得票数最多的预测分类结果作为所述最终分类结果。Further, in an embodiment of the present invention, the second classification module is specifically configured to use the voting method of the combination classifier to select the predicted classification result with the most votes as the final classification result.
进一步地，在本发明的一个实施例中，所述N组特征中每个特征包括亮度特征和纹理特征，其中，所述纹理特征包括灰度共生矩阵和灰度-梯度共生矩阵。Further, in an embodiment of the present invention, each feature in the N groups of features includes a brightness feature and a texture feature, wherein the texture feature includes a grayscale co-occurrence matrix and a grayscale-gradient co-occurrence matrix.
本发明附加的方面和优点将在下面的描述中部分给出，部分将从下面的描述中变得明显，或通过本发明的实践了解到。Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
附图说明Description of drawings
本发明的上述和/或附加的方面和优点从结合下面附图对实施例的描述中将变得明显和容易理解，其中：The above and/or additional aspects and advantages of the present invention will become apparent and comprehensible from the description of the embodiments in conjunction with the following drawings, wherein:
图1为根据本发明实施例的基于组合分类器的白内障眼底图像分类方法的流程图；1 is a flowchart of a method for classifying cataract fundus images based on a combined classifier according to an embodiment of the present invention;
图2为根据本发明一个实施例的基于组合分类器的白内障眼底图像分类方法的流程图；Fig. 2 is the flow chart of the cataract fundus image classification method based on combined classifier according to one embodiment of the present invention;
图3为根据本发明一个实施例的白内障眼底图像示意图；Fig. 3 is a schematic diagram of a cataract fundus image according to an embodiment of the present invention;
图4为根据本发明实施例的基于组合分类器的白内障眼底图像分类装置的结构示意图；以及4 is a schematic structural diagram of a cataract fundus image classification device based on a combined classifier according to an embodiment of the present invention; and
图5为根据本发明一个实施例的基于组合分类器的白内障眼底图像分类装置的结构示意图。Fig. 5 is a schematic structural diagram of an apparatus for classifying cataract fundus images based on a combined classifier according to an embodiment of the present invention.
具体实施方式Detailed ways
下面详细描述本发明的实施例，所述实施例的示例在附图中示出，其中自始至终相同或类似的标号表示相同或类似的元件或具有相同或类似功能的元件。下面通过参考附图描述的实施例是示例性的，旨在用于解释本发明，而不能理解为对本发明的限制。Embodiments of the present invention are described in detail below, examples of which are shown in the drawings, wherein the same or similar reference numerals designate the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.
此外，术语“第一”、“第二”仅用于描述目的，而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者更多个该特征。在本发明的描述中，“多个”的含义是两个或两个以上，除非另有明确具体的限定。In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, a feature defined as "first" and "second" may explicitly or implicitly include one or more of these features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.
在本发明中，除非另有明确的规定和限定，术语“安装”、“相连”、“连接”、“固定”等术语应做广义理解，例如，可以是固定连接，也可以是可拆卸连接，或一体地连接；可以是机械连接，也可以是电连接；可以是直接相连，也可以通过中间媒介间接相连，可以是两个元件内部的连通。对于本领域的普通技术人员而言，可以根据具体情况理解上述术语在本发明中的具体含义。In the present invention, unless otherwise clearly specified and limited, terms such as "installation", "connection", "connection" and "fixation" should be understood in a broad sense, for example, it can be a fixed connection or a detachable connection , or integrally connected; it may be mechanically connected or electrically connected; it may be directly connected or indirectly connected through an intermediary, and it may be the internal communication of two components. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention according to specific situations.
在本发明中，除非另有明确的规定和限定，第一特征在第二特征之“上”或之“下”可以包括第一和第二特征直接接触，也可以包括第一和第二特征不是直接接触而是通过它们之间的另外的特征接触。而且，第一特征在第二特征“之上”、“上方”和“上面”包括第一特征在第二特征正上方和斜上方，或仅仅表示第一特征水平高度高于第二特征。第一特征在第二特征“之下”、“下方”和“下面”包括第一特征在第二特征正上方和斜上方，或仅仅表示第一特征水平高度小于第二特征。In the present invention, unless otherwise clearly specified and limited, a first feature being "on" or "under" a second feature may include direct contact between the first and second features, and may also include the first and second features Not in direct contact but through another characteristic contact between them. Moreover, "above", "above" and "above" the first feature on the second feature include that the first feature is directly above and obliquely above the second feature, or simply means that the first feature is horizontally higher than the second feature. "Below", "below" and "under" the first feature to the second feature include that the first feature is directly above and obliquely above the second feature, or simply means that the first feature is less horizontal than the second feature.
下面参照附图描述根据本发明实施例提出的基于组合分类器的白内障眼底图像分类方法及分类装置，首先将参照附图描述根据本发明实施例提出的基于组合分类器的白内障眼底图像分类方法。参照图1所示，该分类方法包括以下步骤：The cataract fundus image classification method and classification device based on the combined classifier according to the embodiment of the present invention will be described below with reference to the accompanying drawings. First, the cataract fundus image classification method based on the combined classifier according to the embodiment of the present invention will be described with reference to the accompanying drawings. Shown in Fig. 1 with reference to, this classification method comprises the following steps:
S101，获取白内障眼底图像。S101, acquiring a cataract fundus image.
S102，对白内障眼底图像进行预处理。S102. Preprocessing the cataract fundus image.
进一步地，在本发明的一个实施例中，预处理包括：提取RGB彩色空间和G通道图像；对G通道图像进行两次直方图均衡化处理；和/或通过高低帽算法和三边滤波器处理G通道图像。Further, in one embodiment of the present invention, the preprocessing includes: extracting the RGB color space and the G channel image; performing two histogram equalization processing on the G channel image; and/or passing the high-low hat algorithm and the trilateral filter Process G channel images.
S103，分别通过小波变换、轮辅方法和纹理分析从处理后的白内障眼底图像中提取N组特征，N为正整数。S103, extracting N groups of features from the processed cataract fundus image respectively through wavelet transform, wheel-assisted method and texture analysis, where N is a positive integer.
进一步地，在本发明的一个实施例中，N组特征中每个特征包括亮度特征和纹理特征，其中，纹理特征包括灰度共生矩阵和灰度-梯度共生矩阵。Further, in an embodiment of the present invention, each feature in the N groups of features includes a brightness feature and a texture feature, wherein the texture feature includes a grayscale co-occurrence matrix and a grayscale-gradient co-occurrence matrix.
进一步地，在本发明的一个实施例中，N可以为3，M可以为6。Further, in an embodiment of the present invention, N may be 3, and M may be 6.
具体地，参照图2所示，在本发明的一个实施例中，根据对预处理后的图像进行分析，使用三种不同的方法可以提取三组特征：Specifically, as shown in FIG. 2, in one embodiment of the present invention, according to the analysis of the preprocessed image, three groups of features can be extracted using three different methods:
a.使用小波变换提取特征(小波特征)a. Use wavelet transform to extract features (wavelet features)
在图像预处理阶段提取RGB彩色空间的G通道图像，对G通道图像进行两次直方图均衡化操作。对预处理后的图像使用三级Haar小波分解，不同类别的眼底图像经过Haar小波变换后第三级的系数分布不同。因此可以用来作为特征用于分类眼底图像。In the image preprocessing stage, the G channel image of the RGB color space is extracted, and two histogram equalization operations are performed on the G channel image. The three-level Haar wavelet decomposition is used for the preprocessed image, and the coefficient distribution of the third level is different for different types of fundus images after Haar wavelet transformation. Therefore, it can be used as features for classifying fundus images.
b.使用轮辅方法进行特征提取(轮辅特征)b. Feature extraction using wheel-assisted method (wheel-assisted feature)
眼底图像的预处理和a中小波变换的一样。参照图3所示，在预处理的眼底图像上取18条直径(每条之间等角度)和5个不同半径的经向圆。当图像中的血管穿过直径或者经向圆，将会产生一个局部峰值，眼底图像的血管信息越多，波动越大。The preprocessing of the fundus image is the same as the wavelet transform in a. As shown in FIG. 3 , 18 meridional circles with diameters (equal angles between each bar) and 5 different radii are taken on the preprocessed fundus image. When the blood vessels in the image pass through the diameter or meridian circle, a local peak will be generated, and the more blood vessel information in the fundus image, the greater the fluctuation.
c.使用上面介绍的论文中的方法进行特征提取(纹理特征)c. Use the method in the paper introduced above for feature extraction (texture features)
在图像预处理阶段主要包括提取RGB彩色空间的G通道图像、改进的高低帽算法和三边滤波器处理三部分。在预处理后的眼底图像中提取了40个特征，包括1个亮度特征、24个灰度共生矩阵特征和15个灰度-梯度共生矩阵特征。In the stage of image preprocessing, it mainly includes three parts: extracting G channel image in RGB color space, improving high-low hat algorithm and three-sided filter processing. 40 features were extracted from the preprocessed fundus images, including 1 brightness feature, 24 gray-scale co-occurrence matrix features and 15 gray-gradient co-occurrence matrix features.
S104，分别通过支持向量机和BP神经网对N组特征进行预测分类，以获取M种预测分类结果，M为大于等于N的正整数。S104. Predict and classify the N groups of features through the support vector machine and the BP neural network respectively, so as to obtain M kinds of prediction and classification results, where M is a positive integer greater than or equal to N.
进一步地，对于提取的三组特征分别使用支持向量机和BP神经网络进行预测分类，将产生六种预测的结果。Further, using support vector machine and BP neural network to predict and classify the extracted three groups of features, six prediction results will be produced.
其中，在本发明的一个实施例中，使用支持向量机和BP神经网络可以将白内障眼底图像分为正常、轻度、中度和重度四个类别。Wherein, in one embodiment of the present invention, cataract fundus images can be classified into four categories: normal, mild, moderate and severe by using support vector machine and BP neural network.
S105，通过组合分类器对M种预测分类结果进行投票分类，以获取最终分类结果。S105. Vote and classify M types of predicted classification results by combining classifiers to obtain a final classification result.
进一步地，在本发明的一个实施例中，采用组合分类器的投票法选取获得票数最多的预测分类结果作为最终分类结果。Further, in one embodiment of the present invention, the predicted classification result with the most votes is selected as the final classification result by using the voting method of combined classifiers.
其中，对六种预测结果使用组合分类器中常用的投票法，选择六中预测结果中获得投票数最多的类别作为最终的类别。Among them, the voting method commonly used in combined classifiers is used for the six prediction results, and the category with the most votes among the six prediction results is selected as the final category.
在本发明的实施例中，本发明实施例对白内障眼底图像做图像预处理操作，对预处理后的眼底照片进行分析，使用三种不同的方法提取了三组特征，其次对于提取到的三组特征，分别使用支持向量机和BP神经网络将白内障眼底图像分为正常、轻度、中度和重度四个类别，最后采用组合分类器中的投票法获得白内障眼底图像的最终类别，最后的分类正确率能达到85.6％，相比于相关技术中论文的分类方法提高了3.1％的正确率，对具体应用和医学研究都有很大意义。In the embodiment of the present invention, the embodiment of the present invention performs an image preprocessing operation on the cataract fundus image, analyzes the preprocessed fundus photo, uses three different methods to extract three groups of features, and then for the extracted three Group features, use support vector machine and BP neural network to divide cataract fundus images into four categories: normal, mild, moderate and severe, and finally use the voting method in the combined classifier to obtain the final category of cataract fundus images, and finally The correct rate of classification can reach 85.6%, which is 3.1% higher than the classification method of papers in related technologies, which has great significance for specific applications and medical research.
根据本发明实施例提出的基于组合分类器的白内障眼底图像分类方法，在对白内障眼底图像进行预处理之后，通过小波变换、轮辅方法和纹理分析分别提取多组特征，其次通过支持向量机和BP神经网对多组特征进行预测分类，最后通过组合分类器进行投票分类，以获取最终分类结果，提高分类的精确度，更好地满足对于正确率的使用要求。According to the cataract fundus image classification method based on the combined classifier proposed in the embodiment of the present invention, after preprocessing the cataract fundus image, multiple groups of features are extracted respectively through wavelet transform, wheel-assisted method and texture analysis, and then through support vector machine and The BP neural network predicts and classifies multiple sets of features, and finally votes and classifies by combining classifiers to obtain the final classification result, improve the classification accuracy, and better meet the use requirements for the correct rate.
其次参照附图描述根据本发明实施例提出的基于组合分类器的白内障眼底图像分类装置。参照图4所示。该分类装置10包括：获取模块100、处理模块200、提取模块300、第一分类模块400和第二分类模块500。Next, a cataract fundus image classification device based on a combined classifier according to an embodiment of the present invention will be described with reference to the accompanying drawings. Refer to Figure 4. The classification device 10 includes: an acquisition module 100 , a processing module 200 , an extraction module 300 , a first classification module 400 and a second classification module 500 .
其中，获取模块100用于获取白内障眼底图像。处理模块200用于对白内障眼底图像进行预处理。提取模块300用于分别通过小波变换、轮辅方法和纹理分析从处理后的白内障眼底图像中提取N组特征，N为正整数。第一分类模块400用于分别通过小波变换、轮辅方法和纹理分析从处理后的白内障眼底图像中提取N组特征，N为正整数。第二分类模块500用于通过组合分类器对M种预测分类结果进行投票分类，以获取最终分类结果。本发明实施例的分类装置10通过小波变换、轮辅方法和纹理分析提取特征，其次通过支持向量机和BP神经网进行预测分类，从而获取分类结果，提高分类的精确度。Wherein, the acquiring module 100 is used for acquiring a cataract fundus image. The processing module 200 is used for preprocessing the cataract fundus image. The extraction module 300 is used to extract N groups of features from the processed cataract fundus image through wavelet transform, wheel-assisted method and texture analysis respectively, where N is a positive integer. The first classification module 400 is used to extract N groups of features from the processed cataract fundus image through wavelet transform, wheel-assisted method and texture analysis respectively, where N is a positive integer. The second classification module 500 is configured to vote and classify M types of predicted classification results by combining classifiers to obtain a final classification result. The classification device 10 of the embodiment of the present invention extracts features through wavelet transform, wheel-assisted method and texture analysis, and then predicts and classifies through support vector machine and BP neural network, so as to obtain classification results and improve classification accuracy.
进一步地，在本发明的一个实施例中，N可以为3，M可以为6。Further, in an embodiment of the present invention, N may be 3, and M may be 6.
进一步地，在本发明的一个实施例中，N组特征中每个特征包括亮度特征和纹理特征，其中，纹理特征包括灰度共生矩阵和灰度-梯度共生矩阵。Further, in an embodiment of the present invention, each feature in the N groups of features includes a brightness feature and a texture feature, wherein the texture feature includes a grayscale co-occurrence matrix and a grayscale-gradient co-occurrence matrix.
进一步地，在本发明的一个实施例中，参照图5所示，处理模块200包括：提取单元201和处理单元202。Further, in an embodiment of the present invention, as shown in FIG. 5 , the processing module 200 includes: an extraction unit 201 and a processing unit 202 .
其中，提取单元201用于提取RGB彩色空间和G通道图像。处理单元202用于对G通道图像进行两次直方图均衡化处理和/或通过高低帽算法和三边滤波器处理G通道图像。Wherein, the extracting unit 201 is used for extracting RGB color space and G channel image. The processing unit 202 is configured to perform two histogram equalization processes on the G-channel image and/or process the G-channel image through a high-low hat algorithm and a trilateral filter.
具体地，参照图2所示，在本发明的一个实施例中，根据对预处理后的图像进行分析，使用三种不同的方法可以提取三组特征：Specifically, as shown in FIG. 2, in one embodiment of the present invention, according to the analysis of the preprocessed image, three groups of features can be extracted using three different methods:
a.使用小波变换提取特征(小波特征)a. Use wavelet transform to extract features (wavelet features)
在图像预处理阶段提取RGB彩色空间的G通道图像，对G通道图像进行两次直方图均衡化操作。对预处理后的图像使用三级Haar小波分解，不同类别的眼底图像经过Haar小波变换后第三级的系数分布不同。因此可以用来作为特征用于分类眼底图像。In the image preprocessing stage, the G channel image of the RGB color space is extracted, and two histogram equalization operations are performed on the G channel image. The three-level Haar wavelet decomposition is used for the preprocessed image, and the coefficient distribution of the third level is different for different types of fundus images after Haar wavelet transformation. Therefore, it can be used as features for classifying fundus images.
b.使用轮辅方法进行特征提取(轮辅特征)b. Feature extraction using wheel-assisted method (wheel-assisted feature)
眼底图像的预处理和a中小波变换的一样。参照图3所示，在预处理的眼底图像上取18条直径(每条之间等角度)和5个不同半径的经向圆。当图像中的血管穿过直径或者经向圆，将会产生一个局部峰值，眼底图像的血管信息越多，波动越大。The preprocessing of the fundus image is the same as the wavelet transform in a. As shown in FIG. 3 , 18 meridional circles with diameters (equal angles between each bar) and 5 different radii are taken on the preprocessed fundus image. When the blood vessels in the image pass through the diameter or meridian circle, a local peak will be generated, and the more blood vessel information in the fundus image, the greater the fluctuation.
c.使用上面介绍的论文中的方法进行特征提取(纹理特征)c. Use the method in the paper introduced above for feature extraction (texture features)
在图像预处理阶段主要包括提取RGB彩色空间的G通道图像、改进的高低帽算法和三边滤波器处理三部分。在预处理后的眼底图像中提取了40个特征，包括1个亮度特征、24个灰度共生矩阵特征和15个灰度-梯度共生矩阵特征。In the stage of image preprocessing, it mainly includes three parts: extracting G channel image in RGB color space, improving high-low hat algorithm and three-sided filter processing. 40 features were extracted from the preprocessed fundus images, including 1 brightness feature, 24 gray-scale co-occurrence matrix features and 15 gray-gradient co-occurrence matrix features.
进一步地，使用支持向量机和BP神经网络可以将白内障眼底图像分为正常、轻度、中度和重度四个类别，其次对六种预测结果使用组合分类器中常用的投票法，选择六中预测结果中获得投票数最多的类别作为最终的类别。Further, using the support vector machine and BP neural network, the fundus images of cataract can be divided into four categories: normal, mild, moderate and severe. Secondly, the voting method commonly used in the combination classifier is used for the six prediction results, and the six prediction results are selected. The category with the most votes in the prediction results is the final category.
进一步地，在本发明的一个实施例中，第二分类模块500具体用于采用组合分类器的投票法选取获得票数最多的预测分类结果作为最终分类结果。Further, in an embodiment of the present invention, the second classification module 500 is specifically configured to select the predicted classification result with the most votes as the final classification result by using a voting method of combined classifiers.
在本发明的实施例中，本发明实施例对白内障眼底图像做图像预处理操作，对预处理后的眼底照片进行分析，使用三种不同的方法提取了三组特征，其次对于提取到的三组特征，分别使用支持向量机和BP神经网络将白内障眼底图像分为正常、轻度、中度和重度四个类别，最后采用组合分类器中的投票法获得白内障眼底图像的最终类别，最后的分类正确率能达到85.6％，相比于相关技术中论文的分类方法提高了3.1％的正确率，对具体应用和医学研究都有很大意义。In the embodiment of the present invention, the embodiment of the present invention performs an image preprocessing operation on the cataract fundus image, analyzes the preprocessed fundus photo, uses three different methods to extract three groups of features, and then for the extracted three Group features, use support vector machine and BP neural network to divide cataract fundus images into four categories: normal, mild, moderate and severe, and finally use the voting method in the combined classifier to obtain the final category of cataract fundus images, and finally The correct rate of classification can reach 85.6%, which is 3.1% higher than the classification method of papers in related technologies, which has great significance for specific applications and medical research.
根据本发明实施例提出的基于组合分类器的白内障眼底图像分类装置，在对白内障眼底图像进行预处理之后，通过小波变换、轮辅方法和纹理分析分别提取多组特征，其次通过支持向量机和BP神经网对多组特征进行预测分类，最后通过组合分类器进行投票分类，以获取最终分类结果，提高分类的精确度，更好地满足对于正确率的使用要求。According to the cataract fundus image classification device based on the combination classifier proposed in the embodiment of the present invention, after preprocessing the cataract fundus image, multiple groups of features are extracted respectively through wavelet transform, wheel-assisted method and texture analysis, and then through support vector machine and The BP neural network predicts and classifies multiple sets of features, and finally votes and classifies by combining classifiers to obtain the final classification result, improve the classification accuracy, and better meet the use requirements for the correct rate.
流程图中或在此以其他方式描述的任何过程或方法描述可以被理解为，表示包括一个或更多个用于实现特定逻辑功能或过程的步骤的可执行指令的代码的模块、片段或部分，并且本发明的优选实施方式的范围包括另外的实现，其中可以不按所示出或讨论的顺序，包括根据所涉及的功能按基本同时的方式或按相反的顺序，来执行功能，这应被本发明的实施例所属技术领域的技术人员所理解。Any process or method descriptions in flowcharts or otherwise described herein may be understood to represent modules, segments or portions of code comprising one or more executable instructions for implementing specific logical functions or steps of the process , and the scope of preferred embodiments of the invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including substantially concurrently or in reverse order depending on the functions involved, which shall It is understood by those skilled in the art to which the embodiments of the present invention pertain.
在流程图中表示或在此以其他方式描述的逻辑和/或步骤，例如，可以被认为是用于实现逻辑功能的可执行指令的定序列表，可以具体实现在任何计算机可读介质中，以供指令执行系统、装置或设备(如基于计算机的系统、包括处理器的系统或其他可以从指令执行系统、装置或设备取指令并执行指令的系统)使用，或结合这些指令执行系统、装置或设备而使用。就本说明书而言，"计算机可读介质"可以是任何可以包含、存储、通信、传播或传输程序以供指令执行系统、装置或设备或结合这些指令执行系统、装置或设备而使用的装置。计算机可读介质的更具体的示例(非穷尽性列表)包括以下：具有一个或多个布线的电连接部(电子装置)，便携式计算机盘盒(磁装置)，随机存取存储器(RAM)，只读存储器(ROM)，可擦除可编辑只读存储器(EPROM或闪速存储器)，光纤装置，以及便携式光盘只读存储器(CDROM)。另外，计算机可读介质甚至可以是可在其上打印所述程序的纸或其他合适的介质，因为可以例如通过对纸或其他介质进行光学扫描，接着进行编辑、解译或必要时以其他合适方式进行处理来以电子方式获得所述程序，然后将其存储在计算机存储器中。The logic and/or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced listing of executable instructions for implementing logical functions, can be embodied in any computer-readable medium, For use with instruction execution systems, devices, or devices (such as computer-based systems, systems including processors, or other systems that can fetch instructions from instruction execution systems, devices, or devices and execute instructions), or in conjunction with these instruction execution systems, devices or equipment used. For the purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use in or in conjunction with an instruction execution system, device or device. More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connection with one or more wires (electronic device), portable computer disk case (magnetic device), random access memory (RAM), Read Only Memory (ROM), Erasable and Editable Read Only Memory (EPROM or Flash Memory), Fiber Optic Devices, and Portable Compact Disc Read Only Memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program can be printed, since the program can be read, for example, by optically scanning the paper or other medium, followed by editing, interpretation or other suitable processing if necessary. The program is processed electronically and stored in computer memory.
应当理解，本发明的各部分可以用硬件、软件、固件或它们的组合来实现。在上述实施方式中，多个步骤或方法可以用存储在存储器中且由合适的指令执行系统执行的软件或固件来实现。例如，如果用硬件来实现，和在另一实施方式中一样，可用本领域公知的下列技术中的任一项或他们的组合来实现：具有用于对数据信号实现逻辑功能的逻辑门电路的离散逻辑电路，具有合适的组合逻辑门电路的专用集成电路，可编程门阵列(PGA)，现场可编程门阵列(FPGA)等。It should be understood that various parts of the present invention can be realized by hardware, software, firmware or their combination. In the embodiments described above, various steps or methods may be implemented by software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one or combination of the following techniques known in the art: Discrete logic circuits, ASICs with suitable combinational logic gates, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.
本技术领域的普通技术人员可以理解实现上述实施例方法携带的全部或部分步骤是可以通过程序来指令相关的硬件完成，所述的程序可以存储于一种计算机可读存储介质中，该程序在执行时，包括方法实施例的步骤之一或其组合。Those of ordinary skill in the art can understand that all or part of the steps carried by the methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. During execution, one or a combination of the steps of the method embodiments is included.
此外，在本发明各个实施例中的各功能单元可以集成在一个处理模块中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个模块中。上述集成的模块既可以采用硬件的形式实现，也可以采用软件功能模块的形式实现。所述集成的模块如果以软件功能模块的形式实现并作为独立的产品销售或使用时，也可以存储在一个计算机可读取存储介质中。In addition, each functional unit in each embodiment of the present invention may be integrated into one processing module, each unit may exist separately physically, or two or more units may be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software function modules. If the integrated modules are realized in the form of software function modules and sold or used as independent products, they can also be stored in a computer-readable storage medium.
上述提到的存储介质可以是只读存储器，磁盘或光盘等。The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, and the like.
在本说明书的描述中，参考术语“一个实施例”、“一些实施例”、“示例”、“具体示例”、或“一些示例”等的描述意指结合该实施例或示例描述的具体特征、结构、材料或者特点包含于本发明的至少一个实施例或示例中。在本说明书中，对上述术语的示意性表述不一定指的是相同的实施例或示例。而且，描述的具体特征、结构、材料或者特点可以在任何的一个或多个实施例或示例中以合适的方式结合。In the description of this specification, descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or characteristic is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
尽管上面已经示出和描述了本发明的实施例，可以理解的是，上述实施例是示例性的，不能理解为对本发明的限制，本领域的普通技术人员在不脱离本发明的原理和宗旨的情况下在本发明的范围内可以对上述实施例进行变化、修改、替换和变型。Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be construed as limitations to the present invention. Variations, modifications, substitutions, and modifications to the above-described embodiments are possible within the scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510275513.8ACN104881683B (en) | 2015-05-26 | 2015-05-26 | Cataract eye fundus image sorting technique based on assembled classifier and sorter |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510275513.8ACN104881683B (en) | 2015-05-26 | 2015-05-26 | Cataract eye fundus image sorting technique based on assembled classifier and sorter |
| Publication Number | Publication Date |
|---|---|
| CN104881683A CN104881683A (en) | 2015-09-02 |
| CN104881683Btrue CN104881683B (en) | 2018-08-28 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510275513.8AExpired - Fee RelatedCN104881683B (en) | 2015-05-26 | 2015-05-26 | Cataract eye fundus image sorting technique based on assembled classifier and sorter |
| Country | Link |
|---|---|
| CN (1) | CN104881683B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106530295A (en)* | 2016-11-07 | 2017-03-22 | 首都医科大学 | Fundus image classification method and device of retinopathy |
| CN108095683A (en)* | 2016-11-11 | 2018-06-01 | 北京羽医甘蓝信息技术有限公司 | The method and apparatus of processing eye fundus image based on deep learning |
| CN106846293B (en)* | 2016-12-14 | 2020-08-07 | 海纳医信（北京）软件科技有限责任公司 | Image processing method and device |
| CN106934798B (en)* | 2017-02-20 | 2020-08-21 | 苏州体素信息科技有限公司 | Diabetic retinopathy classification and classification method based on deep learning |
| CN108172291B (en)* | 2017-05-04 | 2020-01-07 | 深圳硅基智能科技有限公司 | Diabetic retinopathy recognition system based on fundus images |
| CN109390053B (en)* | 2017-08-02 | 2021-01-08 | 上海市第六人民医院 | Fundus image processing method, fundus image processing apparatus, computer device, and storage medium |
| CN110019900B (en)* | 2017-08-29 | 2021-04-20 | 中国移动通信有限公司研究院 | Method and device for structure labeling of fundus images |
| CN109102494A (en)* | 2018-07-04 | 2018-12-28 | 中山大学中山眼科中心 | A kind of After Cataract image analysis method and device |
| CN109036550A (en)* | 2018-07-04 | 2018-12-18 | 中山大学中山眼科中心 | A kind of cataract analysis method and device based on artificial intelligence |
| CN109102885B (en)* | 2018-08-20 | 2021-03-05 | 北京邮电大学 | Automatic cataract grading method based on combination of convolutional neural network and random forest |
| CN109875499B (en)* | 2019-02-22 | 2021-03-09 | 中山大学中山眼科中心 | Method and device for screening risk degree of congenital cataract |
| CN109872322A (en)* | 2019-02-27 | 2019-06-11 | 电子科技大学 | A method for localization of nuclear cataract lesions based on cascade detection model |
| CN110327013B (en)* | 2019-05-21 | 2022-02-15 | 北京至真互联网技术有限公司 | Fundus image detection method, device and equipment and storage medium |
| CN112220445A (en)* | 2020-09-29 | 2021-01-15 | 温州医科大学附属眼视光医院 | Mobile terminal-based cataract and after cataract analysis system and device |
| CN113379702B (en)* | 2021-06-08 | 2024-05-28 | 广州医软智能科技有限公司 | Blood vessel path extraction method and device for microcirculation image |
| CN116188810B (en)* | 2023-04-25 | 2023-07-21 | 浙江一山智慧医疗研究有限公司 | Method, device and application for extracting characteristics of optic disc |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102202557A (en)* | 2008-05-20 | 2011-09-28 | 科学、技术与研究机构 | An automatic opacity detection system for cortical cataract diagnosis |
| CN104114079A (en)* | 2011-10-24 | 2014-10-22 | Iriss医疗科技有限公司 | System and method for identifying eye conditions |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102202557A (en)* | 2008-05-20 | 2011-09-28 | 科学、技术与研究机构 | An automatic opacity detection system for cortical cataract diagnosis |
| CN104114079A (en)* | 2011-10-24 | 2014-10-22 | Iriss医疗科技有限公司 | System and method for identifying eye conditions |
| Title |
|---|
| A computer-aided healthcare system for cataract classification and grading based on fundus image analysis;Liye Guo et al;《Computers in Industry》;20141111;第72-80页* |
| 基于分类器技术的白内障组织图像去噪方法;李爱丽 等;《计算机工程与设计》;20130831;第34卷(第8期);第2834-2838页* |
| 多分类器组合及其遥感分类研究;张少佳;《中国优秀硕士学位论文全文数据库基础科学辑》;20110215(第02期);第A008-43页* |
| 白内障影像评估的新方式;李建军;《眼科》;20101231;第19卷(第2期);第76-78页* |
| Publication number | Publication date |
|---|---|
| CN104881683A (en) | 2015-09-02 |
| Publication | Publication Date | Title |
|---|---|---|
| CN104881683B (en) | Cataract eye fundus image sorting technique based on assembled classifier and sorter | |
| Joshi et al. | Glaucoma detection using image processing and supervised learning for classification | |
| CN109886273B (en) | A CMR Image Segmentation and Classification System | |
| Pan et al. | Classification of Malaria-Infected Cells Using Deep | |
| Prasad et al. | Early detection of diabetic retinopathy from digital retinal fundus images | |
| CN106530295A (en) | Fundus image classification method and device of retinopathy | |
| JP2022551683A (en) | Methods and systems for non-invasive genetic testing using artificial intelligence (AI) models | |
| CN110991254B (en) | Ultrasound image video classification prediction method and system | |
| Reddy et al. | An efficient multilevel thresholding scheme for heart image segmentation using a hybrid generalized adversarial network | |
| Fan et al. | Principal component analysis based cataract grading and classification | |
| CN111798440A (en) | Medical image artifact automatic identification method, system and storage medium | |
| Nalbalwar et al. | Detection of brain tumor by using ANN | |
| Bagheri et al. | Deep neural network based polyp segmentation in colonoscopy images using a combination of color spaces | |
| Sadek et al. | Automatic classification of bright retinal lesions via deep network features | |
| WO2023198224A1 (en) | Method for constructing magnetic resonance image preliminary screening model for mental disorders | |
| CN110163839A (en) | Leopard pattern fundus image recognition method, model training method and equipment | |
| Francis et al. | Ensemble learning approach for multi-class classification of Alzheimer’s stages using magnetic resonance imaging | |
| Asaduzzaman et al. | ALZENET: deep learning-based early prediction of Alzheimer's disease through magnetic resonance imaging analysis | |
| Jain et al. | Brain tumor detection using mlops and hybrid multi-cloud | |
| Ponni Bala et al. | Early detection and classification of microaneurysms in retinal fundus images using sequential learning methods | |
| Sujithra et al. | Identification of glaucoma in fundus images utilizing gray wolf optimization with deep convolutional neural network-based resnet50 model | |
| Joshi et al. | Intelligent Fusion Approach for MRI and CT Imaging using CNN with Wavelet Transform Approach | |
| Hai et al. | Real time burning image classification using support vector machine | |
| Valério et al. | DeepMammo: Deep transfer learning for lesion classification of mammographic images | |
| Schaefer et al. | Scleroderma capillary pattern identification using texture descriptors and ensemble classification |
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee | Granted publication date:20180828 |