Diagnosis of retinal health in digital fundus images using continuous wavelet transform (CWT) and entropies.

Vision is paramount to humans to lead an active personal and professional life. The prevalence of ocular diseases is rising, and diseases such as glaucoma, Diabetic Retinopathy (DR) and Age-related Macular Degeneration (AMD) are the leading causes of blindness in developed countries. Identifying these diseases in mass screening programmes is time-consuming, labor-intensive and the diagnosis can be subjective. The use of an automated computer aided diagnosis system will reduce the time taken for analysis and will also reduce the inter-observer subjective variabilities in image interpretation. In this work, we propose one such system for the automatic classification of normal from abnormal (DR, AMD, glaucoma) images. We had a total of 404 normal and 1082 abnormal fundus images in our database. As the first step, 2D-Continuous Wavelet Transform (CWT) decomposition on the fundus images of two classes was performed. Subsequently, energy features and various entropies namely Yager, Renyi, Kapoor, Shannon, and Fuzzy were extracted from the decomposed images. Then, adaptive synthetic sampling approach was applied to balance the normal and abnormal datasets. Next, the extracted features were ranked according to the significances using Particle Swarm Optimization (PSO). Thereupon, the ranked and selected features were used to train the random forest classifier using stratified 10-fold cross validation. Overall, the proposed system presented a performance rate of 92.48%, and a sensitivity and specificity of 89.37% and 95.58% respectively using 15 features. This novel system shows promise in detecting abnormal fundus images, and hence, could be a valuable adjunct eye health screening tool that could be employed in polyclinics, and thereby reduce the workload of specialists at hospitals.

Ovarian tissue characterization in ultrasound: a review.

Ovarian cancer is the most common cause of death among gynecological malignancies. We discuss different types of clinical and nonclinical features that are used to study and analyze the differences between benign and malignant ovarian tumors. Computer-aided diagnostic (CAD) systems of high accuracy are being developed as an initial test for ovarian tumor classification instead of biopsy, which is the current gold standard diagnostic test. We also discuss different aspects of developing a reliable CAD system for the automated classification of ovarian cancer into benign and malignant types. A brief description of the commonly used classifiers in ultrasound-based CAD systems is also given.

An automated technique for carotid far wall classification using grayscale features and wall thickness variability.

PURPOSE: To test a computer-aided diagnostic method for differentiating symptomatic from asymptomatic carotid B-mode ultrasonographic images. METHODS: Our system (called Atheromatic) automatically computed the intima-media thickness (IMT) of the carotid far wall using AtheroEdge, calculated nonlinear features based on higher order spectra, and used these features and IMT and IMT variability (IMTVpoly ) to associate each image to a feature vector that was then labeled as symptomatic or asymptomatic (Sym/Asym) by a multiclassifiers system. We tested this method on a database of 118 carotid artery images from 37 symptomatic and 22 asymptomatic patients RESULTS: The highest accuracy (99.1%) was obtained by the support vector machine classifier using seven features. These features, relevant to discriminate Sym/Asym, included IMT and IMTVpoly , along with the bispectral entropies of the distal wall image at 77°, 78°, and 79° angles. CONCLUSIONS: Classification in Sym/Asym of the far carotid wall is feasible and accurate and could be useful for the early detection of atherosclerosis and to identify patients with higher cardiovascular risk.

Computer-aided diagnostic system for detection of Hashimoto thyroiditis on ultrasound images from a Polish population.

OBJECTIVES: Computer-aided diagnostic (CAD) techniques aid physicians in better diagnosis of diseases by extracting objective and accurate diagnostic information from medical data. Hashimoto thyroiditis is the most common type of inflammation of the thyroid gland. The inflammation changes the structure of the thyroid tissue, and these changes are reflected as echogenic changes on ultrasound images. In this work, we propose a novel CAD system (a class of systems called ThyroScan) that extracts textural features from a thyroid sonogram and uses them to aid in the detection of Hashimoto thyroiditis. METHODS: In this paradigm, we extracted grayscale features based on stationary wavelet transform from 232 normal and 294 Hashimoto thyroiditis-affected thyroid ultrasound images obtained from a Polish population. Significant features were selected using a Student t test. The resulting feature vectors were used to build and evaluate the following 4 classifiers using a 10-fold stratified cross-validation technique: support vector machine, decision tree, fuzzy classifier, and K-nearest neighbor. RESULTS: Using 7 significant features that characterized the textural changes in the images, the fuzzy classifier had the highest classification accuracy of 84.6%, sensitivity of 82.8%, specificity of 87.0%, and a positive predictive value of 88.9%. CONCLUSIONS: The proposed ThyroScan CAD system uses novel features to noninvasively detect the presence of Hashimoto thyroiditis on ultrasound images. Compared to manual interpretations of ultrasound images, the CAD system offers a more objective interpretation of the nature of the thyroid. The preliminary results presented in this work indicate the possibility of using such a CAD system in a clinical setting after evaluating it with larger databases in multicenter clinical trials.

GyneScan: an improved online paradigm for screening of ovarian cancer via tissue characterization.

Ovarian cancer is the fifth highest cause of cancer in women and the leading cause of death from gynecological cancers. Accurate diagnosis of ovarian cancer from acquired images is dependent on the expertise and experience of ultrasonographers or physicians, and is therefore, associated with inter observer variabilities. Computer Aided Diagnostic (CAD) techniques use a number of different data mining techniques to automatically predict the presence or absence of cancer, and therefore, are more reliable and accurate. A review of published literature in the field of CAD based ovarian cancer detection indicates that many studies use ultrasound images as the base for analysis. The key objective of this work is to propose an effective adjunct CAD technique called GyneScan for ovarian tumor detection in ultrasound images. In our proposed data mining framework, we extract several texture features based on first order statistics, Gray Level Co-occurrence Matrix and run length matrix. The significant features selected using t-test are then used to train and test several supervised learning based classifiers such as Probabilistic Neural Networks (PNN), Support Vector Machine (SVM), Decision Tree (DT), k-Nearest Neighbor (KNN), and Naive Bayes (NB). We evaluated the developed framework using 1300 benign and 1300 malignant images. Using 11 significant features in KNN/PNN classifiers, we were able to achieve 100% classification accuracy, sensitivity, specificity, and positive predictive value in detecting ovarian tumor. Even though more validation using larger databases would better establish the robustness of our technique, the preliminary results are promising. This technique could be used as a reliable adjunct method to existing imaging modalities to provide a more confident second opinion on the presence/absence of ovarian tumor.

Semiautomated analysis of carotid artery wall thickness in MRI.

PURPOSE: To develop a semiautomatic method based on level set method (LSM) for carotid arterial wall thickness (CAWT) measurement. MATERIALS AND METHODS: Magnetic resonance imaging (MRI) of diseased carotid arteries was acquired from 10 patients. Ground truth (GT) data for arterial wall segmentation was collected from three experienced vascular clinicians. The semiautomatic variational LSM was employed to segment lumen and arterial wall outer boundaries on 102 MR images. Two computer-based measurements, arterial wall thickness (WT) and arterial wall area (AWA), were computed and compared with GT. Linear regression, Bland-Altman, and bias correlation analysis on WT and AWA were applied for evaluating the performance of the semiautomatic method. RESULTS: Arterial wall thickness measured by radial distance measure (RDM) and polyline distance measure (PDM) correlated well between GT and variational LSM (r = 0.83 for RDM and r = 0.64 for PDM, P < 0.05). The absolute arterial wall area bias between LSM and three observers is less than 10%, suggesting LSM can segment arterial wall well compared with manual tracings. The Jaccard Similarity (Js ) analysis showed a good agreement for the segmentation results between proposed method and GT (Js 0.71 ± 0.08), the mean curve distance for lumen boundary is 0.34 ± 0.2 mm between the proposed method and GT, and 0.47 ± 0.2 mm for outer wall boundary. CONCLUSION: The proposed LSM can generate reasonably accurate lumen and outer wall boundaries compared to manual segmentation, and can work similar to a human reader. However, it tends to overestimate CAWT and AWA compared to the manual segmentation for arterial wall with small area.

Automated classification of patients with coronary artery disease using grayscale features from left ventricle echocardiographic images.

Coronary Artery Disease (CAD), caused by the buildup of plaque on the inside of the coronary arteries, has a high mortality rate. To efficiently detect this condition from echocardiography images, with lesser inter-observer variability and visual interpretation errors, computer based data mining techniques may be exploited. We have developed and presented one such technique in this paper for the classification of normal and CAD affected cases. A multitude of grayscale features (fractal dimension, entropies based on the higher order spectra, features based on image texture and local binary patterns, and wavelet based features) were extracted from echocardiography images belonging to a huge database of 400 normal cases and 400 CAD patients. Only the features that had good discriminating capability were selected using t-test. Several combinations of the resultant significant features were used to evaluate many supervised classifiers to find the combination that presents a good accuracy. We observed that the Gaussian Mixture Model (GMM) classifier trained with a feature subset made up of nine significant features presented the highest accuracy, sensitivity, specificity, and positive predictive value of 100%. We have also developed a novel, highly discriminative HeartIndex, which is a single number that is calculated from the combination of the features, in order to objectively classify the images from either of the two classes. Such an index allows for an easier implementation of the technique for automated CAD detection in the computers in hospitals and clinics.

Prostate tissue characterization/classification in 144 patient population using wavelet and higher order spectra features from transrectal ultrasound images.

In this work, we have proposed an on-line computer-aided diagnostic system called "UroImage" that classifies a Transrectal Ultrasound (TRUS) image into cancerous or non-cancerous with the help of non-linear Higher Order Spectra (HOS) features and Discrete Wavelet Transform (DWT) coefficients. The UroImage system consists of an on-line system where five significant features (one DWT-based feature and four HOS-based features) are extracted from the test image. These on-line features are transformed by the classifier parameters obtained using the training dataset to determine the class. We trained and tested six classifiers. The dataset used for evaluation had 144 TRUS images which were split into training and testing sets. Three-fold and ten-fold cross-validation protocols were adopted for training and estimating the accuracy of the classifiers. The ground truth used for training was obtained using the biopsy results. Among the six classifiers, using 10-fold cross-validation technique, Support Vector Machine and Fuzzy Sugeno classifiers presented the best classification accuracy of 97.9% with equally high values for sensitivity, specificity and positive predictive value. Our proposed automated system, which achieved more than 95% values for all the performance measures, can be an adjunct tool to provide an initial diagnosis for the identification of patients with prostate cancer. The technique, however, is limited by the limitations of 2D ultrasound guided biopsy, and we intend to improve our technique by using 3D TRUS images in the future.

Computed tomography carotid wall plaque characterization using a combination of discrete wavelet transform and texture features: A pilot study.

In 30% of stroke victims, the cause of stroke has been found to be the stenosis caused by plaques in the carotid artery. Early detection of plaque and subsequent classification of the same into symptomatic and asymptomatic can help the clinicians to choose only those patients who are at a higher risk of stroke for risky surgeries and stenosis treatments. Therefore, in this work, we have proposed a non-invasive computer-aided diagnostic technique to classify the detected plaque into the two classes. Computed tomography (CT) images of the carotid artery images were used to extract Local Binary Pattern (LBP) features and wavelet energy features. Significant features were then used to train and test several supervised learning algorithm based classifiers. The Support Vector Machine (SVM) classifier with various kernel configurations was evaluated using LBP and wavelet features. The SVM classifier presented the highest accuracy of 88%, sensitivity of 90.2%, and specificity of 86.5% for radial basis function (RBF) kernel function. The CT images of the carotid artery provide unique 3D images of the artery and plaque that could be used for calculating percentage of stenosis. Our proposed technique enables automatic classification of plaque into asymptomatic and symptomatic with high accuracy, and hence, it can be used for deciding the course of treatment. We have also proposed a single-valued integrated index (Atheromatic Index) using the significant features which can provide a more objective and faster prediction of the class.

Effect of complex wavelet transform filter on thyroid tumor classification in three-dimensional ultrasound.

Ultrasonography has great potential in differentiating malignant thyroid nodules from the benign ones. However, visual interpretation is limited by interobserver variability, and further, the speckle distribution poses a challenge during the classification process. This article thus presents an automated system for tumor classification in three-dimensional contrast-enhanced ultrasonography data sets. The system first processes the contrast-enhanced ultrasonography images using complex wavelet transform-based filter to mitigate the effect of speckle noise. The higher order spectra features are then extracted and used as input for training and testing a fuzzy classifier. In the off-line training system, higher order spectra features are extracted from a set of images known as the training images. These higher order spectra features along with the clinically assigned ground truth are used to train the classifier and obtain an estimate of the classifier or training parameters. The ground truth tells the class label of the image (i.e. whether the image belongs to a benign or malignant nodule). During the online testing phase, the estimated classifier parameters are applied on the higher order spectra features that are extracted from the testing images to predict their class labels. The predicted class labels are compared with their corresponding original ground truth to evaluate the performance of the classifier. Without utilizing the complex wavelet transform filter, the fuzzy classifier demonstrated an accuracy of 91.6%, while utilizing the complex wavelet transform filter, the accuracy significantly boosted to 99.1%.

Ovarian tumor characterization and classification using ultrasound-a new online paradigm.

Among gynecological malignancies, ovarian cancer is the most frequent cause of death. Image mining algorithms have been predominantly used to give the physicians a more objective, fast, and accurate second opinion on the initial diagnosis made from medical images. The objective of this work is to develop an adjunct computer-aided diagnostic technique that uses 3D ultrasound images of the ovary to accurately characterize and classify benign and malignant ovarian tumors. In this algorithm, we first extract features based on the textural changes and higher-order spectra information. The significant features are then selected and used to train and evaluate the decision tree (DT) classifier. The proposed technique was validated using 1,000 benign and 1,000 malignant images, obtained from ten patients with benign and ten with malignant disease, respectively. On evaluating the classifier with tenfold stratified cross validation, the DT classifier presented a high accuracy of 97 %, sensitivity of 94.3 %, and specificity of 99.7 %. This high accuracy was achieved because of the use of the novel combination of the four features which adequately quantify the subtle changes and the nonlinearities in the pixel intensity variations. The rules output by the DT classifier are comprehensible to the end-user and, hence, allow the physicians to more confidently accept the results. The preliminary results show that the features are discriminative enough to yield good accuracy. Moreover, the proposed technique is completely automated, accurate, and can be easily written as a software application for use in any computer.

Comparison of walking parameters obtained from the young, elderly and adults with support.

Data mining techniques are highly useful in the study of various medical signals and images in order to obtain useful information to better predict the diagnosis or prognosis or treatment options for the patient. Study of the human walking pattern helps us understand the variability of motion during activities such as high performance walking and normal walking. A comparison of the parameters quantifying this variability in motion in normal young and elderly subjects and the subjects who need support will aid in better understanding of the relationship among walking patterns, age and disabilities. In this study, we measured the tri-axial acceleration along three directions: anteroposterior, lateral and vertical. We also measured gyrational pitch, roll and yaw. These parameters were obtained using sensors attached to the back, left thigh and right thigh of the three classes of subjects (normal, elderly and adults with support) during the three types of exercises: 10-m normal walk, 10-m high performance walk and stepping. These recorded signals were then subjected to wavelet packet decomposition, and three entropies, namely approximate entropy and two bispectral entropies, were obtained from the resultant wavelet coefficients. On analysing these entropies, we could observe the following: (1) the entropy steadily decreases with the increase in age and with the presence of impairments, and (2) the entropy decreases among all the three types of exercises, namely normal walking and high performance walking. We feel that the results of this work can help in the design of supporting devices for elderly subjects.

An integrated diabetic index using heart rate variability signal features for diagnosis of diabetes.

Electrocardiogram (ECG) signals are difficult to interpret, and clinicians must undertake a long training process to learn to diagnose diabetes from subtle abnormalities in these signals. To facilitate these diagnoses, we have developed a technique based on the heart rate variability signal obtained from ECG signals. This technique uses digital signal processing methods and, therefore, automates the detection of diabetes from ECG signals. In this paper, we describe the signal processing techniques that extract features from heart rate (HR) signals and present an analysis procedure that uses these features to diagnose diabetes. Through statistical analysis, we have identified the correlation dimension, Poincaré geometry properties (SD2), and recurrence plot properties (REC, DET, L(mean)) as useful features. These features differentiate the HR data of diabetic patients from those of patients who do not have the illness, and have been validated by using the AdaBoost classifier with the perceptron weak learner (yielding a classification accuracy of 86%). We then developed a novel diabetic integrated index (DII) that is a combination of these nonlinear features. The DII indicates whether a particular HR signal was taken from a person with diabetes. This index aids the automatic detection of diabetes, thereby allowing a more objective assessment and freeing medical professionals for other tasks.

Ovarian tumor characterization using 3D ultrasound.

Among gynecological malignancies, ovarian cancer is the most frequent cause of death. Preoperative determination of whether a tumor is benign or malignant has often been found to be difficult. Because of such inconclusive findings from ultrasound images and other tests, many patients with benign conditions have been offered unnecessary surgeries thereby increasing patient anxiety and healthcare cost. The key objective of our work is to develop an adjunct Computer Aided Diagnostic (CAD) technique that uses ultrasound images of the ovary and image mining algorithms to accurately classify benign and malignant ovarian tumor images. In this algorithm, we extract texture features based on Local Binary Patterns (LBP) and Laws Texture Energy (LTE) and use them to build and train a Support Vector Machine (SVM) classifier. Our technique was validated using 1000 benign and 1000 malignant images, and we obtained a high accuracy of 99.9% using a SVM classifier with a Radial Basis Function (RBF) kernel. The high accuracy can be attributed to the determination of the novel combination of the 16 texture based features that quantify the subtle changes in the images belonging to both classes. The proposed algorithm has the following characteristics: cost-effectiveness, complete automation, easy deployment, and good end-user comprehensibility. We have also developed a novel integrated index, Ovarian Cancer Index (OCI), which is a combination of the texture features, to present the physicians with a more transparent adjunct technique for ovarian tumor classification.

Data mining framework for fatty liver disease classification in ultrasound: a hybrid feature extraction paradigm.

PURPOSE: Fatty liver disease (FLD) is an increasing prevalent disease that can be reversed if detected early. Ultrasound is the safest and ubiquitous method for identifying FLD. Since expert sonographers are required to accurately interpret the liver ultrasound images, lack of the same will result in interobserver variability. For more objective interpretation, high accuracy, and quick second opinions, computer aided diagnostic (CAD) techniques may be exploited. The purpose of this work is to develop one such CAD technique for accurate classification of normal livers and abnormal livers affected by FLD. METHODS: In this paper, the authors present a CAD technique (called Symtosis) that uses a novel combination of significant features based on the texture, wavelet transform, and higher order spectra of the liver ultrasound images in various supervised learning-based classifiers in order to determine parameters that classify normal and FLD-affected abnormal livers. RESULTS: On evaluating the proposed technique on a database of 58 abnormal and 42 normal liver ultrasound images, the authors were able to achieve a high classification accuracy of 93.3% using the decision tree classifier. CONCLUSIONS: This high accuracy added to the completely automated classification procedure makes the authors' proposed technique highly suitable for clinical deployment and usage.

Atherosclerotic risk stratification strategy for carotid arteries using texture-based features.

Plaques in the carotid artery result in stenosis, which is one of the main causes for stroke. Patients have to be carefully selected for stenosis treatments as they carry some risk. Since patients with symptomatic plaques have greater risk for strokes, an objective classification technique that classifies the plaques into symptomatic and asymptomatic classes is needed. We present a computer aided diagnostic (CAD) based ultrasound characterization methodology (a class of Atheromatic systems) that classifies the patient into symptomatic and asymptomatic classes using two kinds of datasets: (1) plaque regions in ultrasound carotids segmented semi-automatically and (2) far wall gray-scale intima-media thickness (IMT) regions along the common carotid artery segmented automatically. For both kinds of datasets, the protocol consists of estimating texture-based features in frameworks of local binary patterns (LBP) and Law's texture energy (LTE) and applying these features for obtaining the training parameters, which are then used for classification. Our database consists of 150 asymptomatic and 196 symptomatic plaque regions and 342 IMT wall regions. When using the Atheromatic-based system on semiautomatically determined plaque regions, support vector machine (SVM) classifier was adapted with highest accuracy of 83%. The accuracy registered was 89.5% on the far wall gray-scale IMT regions when using SVM, K-nearest neighbor (KNN) or radial basis probabilistic neural network (RBPNN) classifiers. LBP/LTE-based techniques on both kinds of carotid datasets are noninvasive, fast, objective and cost-effective for plaque characterization and, hence, will add more value to the existing carotid plaque diagnostics protocol. We have also proposed an index for each type of datasets: AtheromaticPi, for carotid plaque region, and AtheromaticWi, for IMT carotid wall region, based on the combination of the respective significant features. These indices show a separation between symptomatic and asymptomatic by 4.53 units and 4.42 units, respectively, thereby supporting the texture hypothesis classification.

An integrated index for the identification of diabetic retinopathy stages using texture parameters.

Diabetes is a condition of increase in the blood sugar level higher than the normal range. Prolonged diabetes damages the small blood vessels in the retina resulting in diabetic retinopathy (DR). DR progresses with time without any noticeable symptoms until the damage has occurred. Hence, it is very beneficial to have the regular cost effective eye screening for the diabetes subjects. This paper documents a system that can be used for automatic mass screenings of diabetic retinopathy. Four classes are identified: normal retina, non-proliferative diabetic retinopathy (NPDR), proliferative diabetic retinopathy (PDR), and macular edema (ME). We used 238 retinal fundus images in our analysis. Five different texture features such as homogeneity, correlation, short run emphasis, long run emphasis, and run percentage were extracted from the digital fundus images. These features were fed into a support vector machine classifier (SVM) for automatic classification. SVM classifier of different kernel functions (linear, radial basis function, polynomial of order 1, 2, and 3) was studied. Receiver operation characteristics (ROC) curves were plotted to select the best classifier. Our proposed system is able to identify the unknown class with an accuracy of 85.2%, and sensitivity, specificity, and area under curve (AUC) of 98.9%, 89.5%, and 0.972 respectively using SVM classifier with polynomial kernel of order 3. We have also proposed a new integrated DR index (IDRI) using different features, which is able to identify the different classes with 100% accuracy.

Thermography based breast cancer detection using texture features and Support Vector Machine.

Breast cancer is a leading cause of death nowadays in women throughout the world. In developed countries, it is the most common type of cancer in women, and it is the second or third most common malignancy in developing countries. The cancer incidence is gradually increasing and remains a significant public health concern. The limitations of mammography as a screening and diagnostic modality, especially in young women with dense breasts, necessitated the development of novel and more effective strategies with high sensitivity and specificity. Thermal imaging (thermography) is a noninvasive imaging procedure used to record the thermal patterns using Infrared (IR) camera. The aim of this study is to evaluate the feasibility of using thermal imaging as a potential tool for detecting breast cancer. In this work, we have used 50 IR breast images (25 normal and 25 cancerous) collected from Singapore General Hospital, Singapore. Texture features were extracted from co-occurrence matrix and run length matrix. Subsequently, these features were fed to the Support Vector Machine (SVM) classifier for automatic classification of normal and malignant breast conditions. Our proposed system gave an accuracy of 88.10%, sensitivity and specificity of 85.71% and 90.48% respectively.

Wavelet-based energy features for glaucomatous image classification.

Texture features within images are actively pursued for accurate and efficient glaucoma classification. Energy distribution over wavelet subbands is applied to find these important texture features. In this paper, we investigate the discriminatory potential of wavelet features obtained from the daubechies (db3), symlets (sym3), and biorthogonal (bio3.3, bio3.5, and bio3.7) wavelet filters. We propose a novel technique to extract energy signatures obtained using 2-D discrete wavelet transform, and subject these signatures to different feature ranking and feature selection strategies. We have gauged the effectiveness of the resultant ranked and selected subsets of features using a support vector machine, sequential minimal optimization, random forest, and naïve Bayes classification strategies. We observed an accuracy of around 93% using tenfold cross validations to demonstrate the effectiveness of these methods.