Cataracts are the leading cause of visual impairment and blindness globally. Over the years, researchers have achieved significant progress in developing state-of-the-art machine learning techniques for automatic cataract classification and grading, aiming to prevent cataracts early and improve clinicians’ diagnosis efficiency. This survey provides a comprehensive survey of recent advances in machine learning techniques for cataract classification/grading based on ophthalmic images. We summarize existing literature from two research directions: conventional machine learning methods and deep learning methods. This survey also provides insights into existing works of both merits and limitations. In addition, we discuss several challenges of automatic cataract classification/grading based on machine learning techniques and present possible solutions to these challenges for future research.
相似文献
Recently, deep learning, especially convolutional neural networks, has achieved the remarkable results in natural image classification and segmentation. At the same time, in the field of medical image segmentation, researchers use deep learning techniques for tasks such as tumor segmentation, cell segmentation, and organ segmentation. Automatic tumor segmentation plays an important role in radiotherapy and clinical practice and is the basis for the implementation of follow-up treatment programs. This paper reviews the tumor segmentation methods based on deep learning in recent years. We first introduce the common medical image types and the evaluation criteria of segmentation results in tumor segmentation. Then, we review the tumor segmentation methods based on deep learning from technique view and tumor view, respectively. The technique view reviews the researches from the architecture of the deep learning and the tumor view reviews from the type of tumors.
相似文献Cardiovascular diseases (CVDs) in India and globally are the major cause of mortality, as revealed by the World Health Organization (WHO). The irregularities in the pace of heartbeats, called cardiac arrhythmias or heart arrhythmias, are one of the commonly diagnosed CVDs caused by ischemic heart disease, hypertension, alcohol intake, and stressful lifestyle. Other than the listed CVDs, the abnormality in the cardiac rhythm caused by the long term mental stress (stimulated by Autonomic Nervous System (ANS)) is a challenging issue for researchers. Early detection of cardiac arrhythmias through automatic electronic techniques is an important research field since the invention of electrocardiogram (ECG or EKG) and advanced machine learning algorithms. ECG (EKG) provides the record of variations in electrical activity associated with the cardiac cycle, used by cardiologists and researchers as a gold standard to study the heart function. The present work is aimed to provide an extensive survey of work done by researchers in the area of automated ECG analysis and classification of regular & irregular classes of heartbeats by conventional and modern artificial intelligence (AI) methods. The artificial intelligence (AI) based methods have emerged popularly during the last decade for the automatic and early diagnosis of clinical symptoms of arrhythmias. In this work, the literature is explored for the last two decades to review the performance of AI and other computer-based techniques to analyze the ECG signals for the prediction of cardiac (heart rhythm) disorders. The existing ECG feature extraction techniques and machine learning (ML) methods used for ECG signal analysis and classification are compared using the performance metrics like specificity, sensitivity, accuracy, positive predictivity value, etc. Some popular AI methods, which include, artificial neural networks (ANN), Fuzzy logic systems, and other machine learning algorithms (support vector machines (SVM), k-nearest neighbor (KNN), etc.) are considered in this review work for the applications of cardiac arrhythmia classification. The popular ECG databases available publicly to evaluate the classification accuracy of the classifier are also mentioned. The aim is to provide the reader, the prerequisites, the methods used in the last two decades, and the systematic approach, all at one place to further purse a research work in the area of cardiovascular abnormalities detection using the ECG signal. As a contribution to the current work, future challenges for real-time remote ECG acquisition and analysis using the emerging technologies like wireless body sensor network (WBSN) and the internet of things (IoT) are identified.
相似文献Stuttering speech recognition is a well-studied concept in speech signal processing. Classification of speech disorder is the main focus of this study. Classification of stuttered speech is becoming more important with the enhancement of machine learning and deep learning. In this study, some of the recent and most influencing stuttering speech recognition methods are reviewed with a discussion on different categories of stuttering. The stuttering speech recognition process is divided mainly into four segments-input speech pre-emphasis, segmentation, feature extraction, and stutter classification. All these segments are briefly elaborated and related researches are discussed. It is observed that different traditional machine learning and deep learning classification approaches are employed to recognize stuttered speech in last few decades. A comprehensive analysis is presented on different feature extraction and classification method with their efficiency.
相似文献In machine learning, image classification accuracy generally depends on image segmentation and feature extraction methods with the extracted features and its qualities. The main focus of this paper is to determine the defected area of mangoes using image segmentation algorithm for improving the classification accuracy. The Enhanced Fuzzy based K-means clustering algorithm is designed for increasing the efficiency of segmentation. Proposed segmentation method is compared with K-means and Fuzzy C-means clustering methods. The geometric, texture and colour based features are used in the feature extraction. Process of feature selection is done by Maximally Correlated Principal Component Analysis (MCPCA). Finally, in the classification step, severe portions of the affected area are analyzed by Backpropagation Based Discriminant Classifier (BBDC). Proposed classifier is compared with BPNN and Naive Bayes classifiers. The images are classified into three classes in final output like Class A –good quality mango, Class B-average quality mango, and Class C-poor quality mango. Finally, the evaluated results of the proposed model examine various defected and healthy mango images and prove that the proposed method has the highest accuracy when compared with existing methods.
相似文献The mechanism of detecting the neurodegenerative disorder from Magnetic Resonance Images (MRIs) is one of the demanding and critical process in recent days. For this purpose, the existing works introduced some of the segmentation and classification techniques, which were used to detect the abnormal region from the brain images. However, it limits the problems of over segmentation, inefficient classification, and more complexity. The early predictions and the diagnosis process of neurodegenerative-disorders were accomplished by the use of segmentation and classification approaches of various methods. The proposed methodology focused on developing an integrated segmentation and classification techniques for an accurate brain disease classification. Here, the most extensively used segmentation techniques such Particle Swarm Optimization (PSO) and Self-Organizing Map (SOM) techniques are integrated for enabling an efficient image segmentation. In addition, it segments the Grey Matter (GM), White Matter (WM) and Cerebrospinal Fluid (CSF) regions. Consequently, the most suitable features are extracted from the segmented image by using the Neighbor Intensity Pattern (NIP) extraction technique. Based on these features, the normal and abnormal regions are classified by the use of an integrated Neural Network and K-Nearest Neighbor (KNN) classification techniques. The hybridization of the work is, that it integrates the benefits of various segmentation and classification techniques, which leads to increased detection efficiency and classification accuracy. The performance of these techniques are evaluated by using two different datasets such as ADNI and PPMI, which contains more number of brain MRIs. Also, various performance parameters have been utilized to test the results of the proposed system. Moreover, the traditional classification techniques are considered to compare the results of the proposed classification technique. During experimental evaluation, the performance of the techniques are validated by using different measures, and the results are compared with other existing techniques for analyzing the efficiency of proposed mechanism. At last, the results stated that the NN-KNN outperforms the other techniques by exactly locating the affected regions. The proposed framework exhibits the higher performance of accuracy level with 98.6%, sensitivity rate of 95%, exposed 96% of specificity rate and acquires the efficient precision rate of 99.21%. In future, this work can be expanded by using some advanced techniques for classifying other brain diseases.
相似文献Brain tumor classification is a significant issue in Computer-Aided Diagnosis (CAD) for clinical applications. The classification process is crucial and plays a major role to diagnosis the brain tumors. The existing works focus on recognizing brain tumors through diverse classification approaches. Though, the conventional classification approaches are suffered from high false alarm rates. To improve the early-stage brain tumor diagnosis via classification, the main intention of this paper is to introduce a novel brain tumor segmentation and classification model. The dataset gathered from the two benchmark sources is subjected to pre-processing for enhancing the quality of images, and skull stripping for extracting the region of interest from the skull. Further, a new segmentation approach termed Adaptive Fuzzy Active Contour Fusion Model (AFACFM) with a new fitness function is developed. Here, the enhancement of the segmentation is performed by the hybrid Jaya-Tunicate Swarm Algorithm (J-TSA). Next, the combination of Convolutional Neural Network (CNN) and Fuzzy classifier is performed in the final classification phase. The deep features are extracted from the pooling layer of CNN, which are subjected to the Fuzzy classifier for classifying the images into normal, benign, and malignant. As a modification, the parameters of the CNN and Fuzzy classifier are tuned by the proposed J-TSA. The comparative analysis is finally done, and this work demonstrates the potential of using deep learning in MRI images to provide a non-invasive tool for simultaneous and automated tumor segmentation and classification. Through the performance analysis, the accuracy of the designed CNN-Fuzzy using J-TSA was 77%, 29%, 19%, 8.7%, 6.8%, and 1.6% enhanced than SVM, NN, DBN, CNN, Fuzzy, and CNN-Fuzzy, respectively.
相似文献