A comparative study of computer-aided classification systems for focal hepatic lesions from B-mode ultrasound

Abstract

A comparative study of three computer-aided classification (CAC) systems for characterization of focal hepatic lesions (FHLs), such as cyst, hemangioma (HEM), hepatocellular carcinoma (HCC) and metastatic carcinoma (MET), along with normal (NOR) liver tissue is carried out in the present work. In order to develop efficient CAC systems a comprehensive and representative dataset consisting of B-mode ultrasound images with (1) typical and atypical cases of cyst, HEM and MET lesions, (2) small and large HCC lesions and (3) NOR liver cases have been used for designing K-nearest neighbour (KNN), probabilistic neural network (PNN) and a back propagation neural network (BPNN) classifiers. For differential diagnosis between atypical FHLs, expert radiologists often visualize the textural characteristics of regions inside and outside the lesion. Accordingly in the present work, texture features and texture ratio features are computed from regions inside and outside the lesions. A feature set consisting of 208 texture features (i.e. 104 texture features and 104 texture ratio features) is subjected to principal component analysis (PCA) for dimensionality reduction; it is observed that maximum accuracy of 87.7% is obtained for a PCA-BPNN-based CAC system in comparison to 86.1% and 85% as obtained by PCA-PNN and PCA-KNN-based CAC systems. The sensitivity of the proposed PCA-BPNN based CAC system for NOR, Cyst, HEM, HCC and MET cases is 82.5%, 96%, 93.3%, 90% and 82.2%, respectively. The sensitivity values with respect to typical, atypical, small HCC and large HCC cases are 85.9%, 88.1%, 100% and 87%, respectively. Keeping in view the comprehensive and representative dataset used for designing the classifier, the results obtained by the proposed PCA-BPNN-based CAC system are quite encouraging and indicate its usefulness to assist experienced radiologists for interpretation and diagnosis of FHLs.     

Characterization of Primary and Secondary Malignant Liver Lesions from B-Mode Ultrasound

Characterization of hepatocellular carcinomas (HCCs) and metastatic carcinomas (METs) from B-mode ultrasound presents a daunting challenge for radiologists due to their highly overlapping appearances. The differential diagnosis between HCCs and METs is often carried out by observing the texture of regions inside the lesion and the texture of background liver on which the lesion has evolved. The present study investigates the contribution made by texture patterns of regions inside and outside of the lesions for binary classification between HCC and MET lesions. The study is performed on 51 real ultrasound liver images with 54 malignant lesions, i.e., 27 images with 27 solitary HCCs (13 small HCCs and 14 large HCCs) and 24 images with 27 MET lesions (12 typical cases and 15 atypical cases). A total of 120 within-lesion regions of interest and 54 surrounding lesion regions of interest are cropped from 54 lesions. Subsequently, 112 texture features (56 texture features and 56 texture ratio features) are computed by statistical, spectral, and spatial filtering based texture features extraction methods. A two-step methodology is used for feature set optimization, i.e., feature pruning by removal of nondiscriminatory features followed by feature selection by genetic algorithm–support vector machine (SVM) approach. The SVM classifier is designed based on optimum features. The proposed computer-aided diagnostic system achieved the overall classification accuracy of 91.6 % with sensitivity of 90 % and 93.3 % for HCCs and METs, respectively. The promising results obtained by the proposed system indicate its usefulness to assist radiologists in diagnosing liver malignancies.     

Computer-aided diagnosis for the classification of focal liver lesions by use of contrast-enhanced ultrasonography

The authors developed a computer-aided diagnostic (CAD) scheme for classifying focal liver lesions (FLLs) as liver metastasis, hemangioma, and three histologic differentiation types of hepatocellular carcinoma (HCC), by use of microflow imaging (MFI) of contrast-enhanced ultrasonography. One hundred and three FLLs obtained from 97 cases used in this study consisted of 26 metastases (15 hyper- and 11 hypovascularity types), 16 hemangiomas (five hyper- and 11 hypovascularity types) and 61 HCCs: 24 well differentiated (w-HCC), 28 moderately differentiated (m-HCC), and nine poorly differentiated (p-HCC). Pathologies of all cases were determined based on biopsy or surgical specimens. Locations and contours of FLLs on contrast-enhanced images were determined manually by an experienced physician. MFI was obtained with contrast-enhanced low-mechanical-index (MI) pulse subtraction imaging at a fixed plane which included a distinctive cross section of the FLL. In MFI, the inflow high signals in the plane, which were due to the vascular patterns and the contrast agent, were accumulated following flash scanning with a high-MI ultrasound exposure. In the initial step of our computerized scheme, a series of the MFI images was extracted from the original cine clip (AVI format). We applied a smoothing filter and time-sequential running average techniques in order to reduce signal noise on the single MFI image and cyclic noise on the sequential MFI images, respectively. A kidney, vessels, and a liver parenchyma region were segmented automatically by use of the last image of a series of MFI images. The authors estimated time-intensity curves for an FLL by use of a series of the temporally averaged MFI images in order to determine temporal features such as estimated replenishment times at early and delayed phases, flow rates, and peak times. In addition, they extracted morphologic and gray-level image features which were determined based on the physicians' knowledge of the diagnosis of the FLL, such as the size of lesion, vascular patterns, and the presence of hypoechoic regions. They employed a cascade of six independent artificial neural networks (ANNs) by use of extracted temporal and image features for classifying five types of liver diseases. A total of 16 temporal and image features, which were selected from 43 initially extracted features, were used for six different ANNs for making decisions at each decision in the cascade. The ANNs were trained and tested with a leave-one-lesion-out test method. The classification accuracies for the 103 FLLs were 88.5% for metastasis, 93.8% for hemangioma, and 86.9% for all HCCs. In addition, the classification accuracies for histologic differentiation types of HCCs were 79.2% for w-HCC, 50.0% for m-HCC, and 77.8% for p-HCC. The CAD scheme for classifying FLLs by use of the MFI on contrast-enhanced ultrasonography has the potential to improve the diagnostic accuracy in the histologic diagnosis of HCCs and the other liver diseases.     

Discrimination between glioblastoma multiforme and solitary metastasis using morphological features derived from the p:q tensor decomposition of diffusion tensor imaging

The management and treatment of high‐grade glioblastoma multiforme (GBM) and solitary metastasis (MET) are very different and influence the prognosis and subsequent clinical outcomes. In the case of a solitary MET, diagnosis using conventional radiology can be equivocal. Currently, a definitive diagnosis is based on histopathological analysis on a biopsy sample. Here, we present a computerised decision support framework for discrimination between GBM and solitary MET using MRI, which includes: (i) a semi‐automatic segmentation method based on diffusion tensor imaging; (ii) two‐dimensional morphological feature extraction and selection; and (iii) a pattern recognition module for automated tumour classification. Ground truth was provided by histopathological analysis from pre‐treatment stereotactic biopsy or at surgical resection. Our two‐dimensional morphological analysis outperforms previous methods with high cross‐validation accuracy of 97.9% and area under the receiver operating characteristic curve of 0.975 using a neural networks‐based classifier. Copyright © 2014 John Wiley & Sons, Ltd.     

Automated diagnosis of coronary artery diseased patients by heart rate variability analysis using linear and non-linear methods

Abstract

Coronary artery disease (CAD) is a highly considered dangerous disease which may lead to myocardial infarction and even sudden cardiac death. The objective of this work is to evaluate the diagnostic performance features derived from linear and non-linear methods of Heart Rate Variability (HRV) analysis for classification software modules with Normal (NOR) subjects and CAD patients. The proposed methodology follows the recording of electrocardiogram from 60 NOR subjects and 64 CAD patients, RR interval tachogram generation, computing the features from time domain, frequency domain, non-linear methods and its analysis, feature dimension reduction by Principal Component Analysis (PCA) and classification by probabilistic neural network, K nearest neighbour and Support Vector Machine (SVM) classifiers. The results of the study indicate a clear difference in NOR subjects and CAD affected patients by using PCA-SVM classifier with an accuracy of 91.67%, sensitivity of 86.67% and 96.67% for NOR and CAD classes, respectively.     

基于因子分析法的超声肝灌注定量分析初步研究

局灶性肝结节(FLLs)通常可从超声造影图像序列中提取的时间强度曲线进行量化评价.为了克服人工选择的主观性并自动提取曲线,提出一种用于因子分析的基于顶点搜寻的替代-近似(RA)算法.该算法将高维图像序列映射到一维空间,先找到与生理结构对应的两个顶点.然后,根据这两个已知点的信息,在二维空间中可以寻找到第3个顶点.这3个顶点对应着3条不同的目标曲线.实验采用6个在患者自由呼吸运动下采集的肝细胞癌病例对RA算法进行验证.实验结果表明,此方法能准确提取到具有生理意义的因子曲线和相应的因子图,提取的因子曲线与感兴趣区域(ROI)测定曲线的平均相关系数值为0.91±0.03.初步证明了RA算法在FLLs灌注的量化分析具有可行性.     

Comparative Performance Analysis of State-of-the-Art Classification Algorithms Applied to Lung Tissue Categorization

In this paper, we compare five common classifier families in their ability to categorize six lung tissue patterns in high-resolution computed tomography (HRCT) images of patients affected with interstitial lung diseases (ILD) and with healthy tissue. The evaluated classifiers are naive Bayes, k-nearest neighbor, J48 decision trees, multilayer perceptron, and support vector machines (SVM). The dataset used contains 843 regions of interest (ROI) of healthy and five pathologic lung tissue patterns identified by two radiologists at the University Hospitals of Geneva. Correlation of the feature space composed of 39 texture attributes is studied. A grid search for optimal parameters is carried out for each classifier family. Two complementary metrics are used to characterize the performances of classification. These are based on McNemar’s statistical tests and global accuracy. SVM reached best values for each metric and allowed a mean correct prediction rate of 88.3% with high class-specific precision on testing sets of 423 ROIs.     

Usefulness of Texture Analysis for Computerized Classification of Breast Lesions on Mammograms

This work presents the usefulness of texture features in the classification of breast lesions in 5,518 images of regions of interest, which were obtained from the Digital Database for Screening Mammography that included microcalcifications, masses, and normal cases. Sixteen texture features were used, i.e., 13 were based on the spatial gray-level dependence matrix and 3 on the wavelet transform. The nonparametric K-NN classifier was used in the classification stage. The results obtained from receiver operating characteristic analysis indicated that the texture features can be used for separating normal regions and lesions with masses and microcalcifications, yielding the area under the curve (AUC) values of 0.957 and 0.859, respectively. However, the texture features were not very effective for distinguishing between malignant and benign lesions because the AUC was 0.617 for masses and 0.607 for microcalcifications. The study showed that the texture features can be used for the detection of suspicious regions in mammograms.     

Random forests ensemble classifier trained with data resampling strategy to improve cardiac arrhythmia diagnosis

Supervised classification algorithms are commonly used in the designing of computer-aided diagnosis systems. In this study, we present a resampling strategy based Random Forests (RF) ensemble classifier to improve diagnosis of cardiac arrhythmia. Random forests is an ensemble classifier that consists of many decision trees and outputs the class that is the mode of the class's output by individual trees. In this way, an RF ensemble classifier performs better than a single tree from classification performance point of view. In general, multiclass datasets having unbalanced distribution of sample sizes are difficult to analyze in terms of class discrimination. Cardiac arrhythmia is such a dataset that has multiple classes with small sample sizes and it is therefore adequate to test our resampling based training strategy. The dataset contains 452 samples in fourteen types of arrhythmias and eleven of these classes have sample sizes less than 15. Our diagnosis strategy consists of two parts: (i) a correlation based feature selection algorithm is used to select relevant features from cardiac arrhythmia dataset. (ii) RF machine learning algorithm is used to evaluate the performance of selected features with and without simple random sampling to evaluate the efficiency of proposed training strategy. The resultant accuracy of the classifier is found to be 90.0% and this is a quite high diagnosis performance for cardiac arrhythmia. Furthermore, three case studies, i.e., thyroid, cardiotocography and audiology, are used to benchmark the effectiveness of the proposed method. The results of experiments demonstrated the efficiency of random sampling strategy in training RF ensemble classification algorithm.     

SVM-Based CAC System for B-Mode Kidney Ultrasound Images

The present study proposes a computer-aided classification (CAC) system for three kidney classes, viz. normal, medical renal disease (MRD) and cyst using B-mode ultrasound images. Thirty-five B-mode kidney ultrasound images consisting of 11 normal images, 8 MRD images and 16 cyst images have been used. Regions of interest (ROIs) have been marked by the radiologist from the parenchyma region of the kidney in case of normal and MRD cases and from regions inside lesions for cyst cases. To evaluate the contribution of texture features extracted from de-speckled images for the classification task, original images have been pre-processed by eight de-speckling methods. Six categories of texture features are extracted. One-against-one multi-class support vector machine (SVM) classifier has been used for the present work. Based on overall classification accuracy (OCA), features from ROIs of original images are concatenated with the features from ROIs of pre-processed images. On the basis of OCA, few feature sets are considered for feature selection. Differential evolution feature selection (DEFS) has been used to select optimal features for the classification task. DEFS process is repeated 30 times to obtain 30 subsets. Run-length matrix features from ROIs of images pre-processed by Lee’s sigma concatenated with that of enhanced Lee method have resulted in an average accuracy (in %) and standard deviation of 86.3 ± 1.6. The results obtained in the study indicate that the performance of the proposed CAC system is promising, and it can be used by the radiologists in routine clinical practice for the classification of renal diseases.     

Classification of surface EMG signals using optimal wavelet packet method based on Davies-Bouldin criterion

In this paper we present an optimal wavelet packet (OWP) method based on Davies-Bouldin criterion for the classification of surface electromyographic signals. To reduce the feature dimensionality of the outputs of the OWP decomposition, the principle components analysis was employed. Then we chose a neural network classifier to discriminate four types of prosthesis movements. The proposed method achieved a mean classification accuracy of 93.75%, which outperformed the method using the energy of wavelet packet coefficients (with mean classification accuracy 86.25%) and the fuzzy wavelet packet method (87.5%).     

Comparison of Shallow and Deep Learning Methods on Classifying the Regional Pattern of Diffuse Lung Disease

This study aimed to compare shallow and deep learning of classifying the patterns of interstitial lung diseases (ILDs). Using high-resolution computed tomography images, two experienced radiologists marked 1200 regions of interest (ROIs), in which 600 ROIs were each acquired using a GE or Siemens scanner and each group of 600 ROIs consisted of 100 ROIs for subregions that included normal and five regional pulmonary disease patterns (ground-glass opacity, consolidation, reticular opacity, emphysema, and honeycombing). We employed the convolution neural network (CNN) with six learnable layers that consisted of four convolution layers and two fully connected layers. The classification results were compared with the results classified by a shallow learning of a support vector machine (SVM). The CNN classifier showed significantly better performance for accuracy compared with that of the SVM classifier by 6–9%. As the convolution layer increases, the classification accuracy of the CNN showed better performance from 81.27 to 95.12%. Especially in the cases showing pathological ambiguity such as between normal and emphysema cases or between honeycombing and reticular opacity cases, the increment of the convolution layer greatly drops the misclassification rate between each case. Conclusively, the CNN classifier showed significantly greater accuracy than the SVM classifier, and the results implied structural characteristics that are inherent to the specific ILD patterns.     

Differential diagnosis of CT focal liver lesions using texture features, feature selection and ensemble driven classifiers

OBJECTIVES: The aim of the present study is to define an optimally performing computer-aided diagnosis (CAD) architecture for the classification of liver tissue from non-enhanced computed tomography (CT) images into normal liver (C1), hepatic cyst (C2), hemangioma (C3), and hepatocellular carcinoma (C4). To this end, various CAD architectures, based on texture features and ensembles of classifiers (ECs), are comparatively assessed. MATERIALS AND METHODS: Number of regions of interests (ROIs) corresponding to C1-C4 have been defined by experienced radiologists in non-enhanced liver CT images. For each ROI, five distinct sets of texture features were extracted using first order statistics, spatial gray level dependence matrix, gray level difference method, Laws' texture energy measures, and fractal dimension measurements. Two different ECs were constructed and compared. The first one consists of five multilayer perceptron neural networks (NNs), each using as input one of the computed texture feature sets or its reduced version after genetic algorithm-based feature selection. The second EC comprised five different primary classifiers, namely one multilayer perceptron NN, one probabilistic NN, and three k-nearest neighbor classifiers, each fed with the combination of the five texture feature sets or their reduced versions. The final decision of each EC was extracted by using appropriate voting schemes, while bootstrap re-sampling was utilized in order to estimate the generalization ability of the CAD architectures based on the available relatively small-sized data set. RESULTS: The best mean classification accuracy (84.96%) is achieved by the second EC using a fused feature set, and the weighted voting scheme. The fused feature set was obtained after appropriate feature selection applied to specific subsets of the original feature set. CONCLUSIONS: The comparative assessment of the various CAD architectures shows that combining three types of classifiers with a voting scheme, fed with identical feature sets obtained after appropriate feature selection and fusion, may result in an accurate system able to assist differential diagnosis of focal liver lesions from non-enhanced CT images.     

Segmentation, Feature Extraction, and Multiclass Brain Tumor Classification

Multiclass brain tumor classification is performed by using a diversified dataset of 428 post-contrast T1-weighted MR images from 55 patients. These images are of primary brain tumors namely astrocytoma (AS), glioblastoma multiforme (GBM), childhood tumor-medulloblastoma (MED), meningioma (MEN), secondary tumor-metastatic (MET), and normal regions (NR). Eight hundred fifty-six regions of interest (SROIs) are extracted by a content-based active contour model. Two hundred eighteen intensity and texture features are extracted from these SROIs. In this study, principal component analysis (PCA) is used for reduction of dimensionality of the feature space. These six classes are then classified by artificial neural network (ANN). Hence, this approach is named as PCA-ANN approach. Three sets of experiments have been performed. In the first experiment, classification accuracy by ANN approach is performed. In the second experiment, PCA-ANN approach with random sub-sampling has been used in which the SROIs from the same patient may get repeated during testing. It is observed that the classification accuracy has increased from 77 to 91 %. PCA-ANN has delivered high accuracy for each class: AS—90.74 %, GBM—88.46 %, MED—85 %, MEN—90.70 %, MET—96.67 %, and NR—93.78 %. In the third experiment, to remove bias and to test the robustness of the proposed system, data is partitioned in a manner such that the SROIs from the same patient are not common for training and testing sets. In this case also, the proposed system has performed well by delivering an overall accuracy of 85.23 %. The individual class accuracy for each class is: AS—86.15 %, GBM—65.1 %, MED—63.36 %, MEN—91.5 %, MET—65.21 %, and NR—93.3 %. A computer-aided diagnostic system comprising of developed methods for segmentation, feature extraction, and classification of brain tumors can be beneficial to radiologists for precise localization, diagnosis, and interpretation of brain tumors on MR images.     

Hepatocarcinogenesis in liver cirrhosis: imaging diagnosis.

Hepatocellular carcinoma (HCC) frequently occurs in association with liver cirrhosis, as chronic liver disease is one of the most important factors in carcinogenesis. In addition to HCCs, recent reports of pathologic studies of resected specimens from cirrhotic liver describe associated small nodular lesions such as regenerative nodule, dysplastic nodule (adenomatous hyperplasia), and dysplastic nodule with subfocus of HCC (early HCC). In hepatocarcinogenesis of the cirrhotic liver, a regenerative nodule might be the first step in the development of HCC, going through phases of dysplastic nodule, early HCC and early advanced HCC in a multistep fashion. Fortunately, recent advances in various imaging techniques have facilitated the verification of these nodules. In this review, new nomenclature of small hepatocellular nodules, and detection and characterization of hepatic nodules in carcinogenesis with various imaging techniques are described with focus on the premalignant lesions and early stage of HCC. In addition, the efficacy of various imaging techniques for diagnosing them is discussed. Although the terms and definitions of these nodules are still variable and controversial, familiarity with the concept of these borderline lesions is important.     

High efficient system for automatic classification of the electrocardiogram beats

Automatic classification of the electrocardiogram (ECG) signals is an important subject for clinical diagnosis of heart disease. This study investigates the design of a high-efficient system to classify five types of ECG beat namely normal beats and four manifestations of heart arrhythmia, in twofold. First, we propose a system that includes two main modules: a feature extraction module and a classification module. Feature extraction module extracts a suitable combination of the ECG’s morphological characteristics and timing interval features. Discrete wavelet transform is used to extract the morphological features. In the classification module, a multi-class support vector machine (SVM)-based classifier is employed. The parameters of this system are determined based on a trial and error method and its performance is evaluated for the MIT-BIH arrhythmia database. Extensive experiments on the parameters of this system such as classifier kernels and various types of features are conducted. These experiments show that in SVM training, the kernels, kernel parameters, and feature selection have very important roles for SVM classification accuracy. Therefore, most appropriates of these parameters should be used for SVM training. Then at the second fold, a novel hybrid intelligent system (HIS) is proposed that consists of three main modules. In the HIS, further to the two mentioned modules, an optimization module is added. In this module, a genetic algorithm is used for optimization of the relevant parameters of system. These parameters are: wavelet filter type for feature extraction, wavelet decomposition level, and classifier’s parameters. Experimental results show that optimization improves the recognition system, efficiently, and HIS is more superior to the system, which as constant parameters.     

EEG signals classification of epileptic patients via feature selection and voting criteria in intelligent method

Epileptic disease can be diagnosed by using intelligent methods on the Electroencephalograph (EEG) signals. In this paper, wavelet packet transform (WPT) was used in each of the frequency bands and wavelet coefficients were obtained, then the energy and entropy function was done on the wavelet coefficients and used as initial feature vectors. In the next step, eight and 15 features from 30 initial energy and entropy features were selected as the final features because their receiver operating characteristic (ROC) curve areas were higher than others. There were seven classifier inputs. These seven classifiers consisted of four artificial neural networks (ANN) with different structures, support vector machines (SVM), K-nearest neighbours (KNN) and a hybrid network. Each classifier was trained by 0.5, 0.8 and 0.9 EEG signals. After the training process, a fusion network based on a voting criteria was used to make the algorithm robust against the possible changes in each classifier and increase the classification accuracy. Finally, the algorithm was tested by other EEG signals. As a result, normal and epileptic classes were detected with total classification accuracy of 99–100%.     

A method to improve visual similarity of breast masses for an interactive computer-aided diagnosis environment

The purpose of this study was to develop and test a method for selecting "visually similar" regions of interest depicting breast masses from a reference library to be used in an interactive computer-aided diagnosis (CAD) environment. A reference library including 1000 malignant mass regions and 2000 benign and CAD-generated false-positive regions was established. When a suspicious mass region is identified, the scheme segments the region and searches for similar regions from the reference library using a multifeature based k-nearest neighbor (KNN) algorithm. To improve selection of reference images, we added an interactive step. All actual masses in the reference library were subjectively rated on a scale from 1 to 9 as to their "visual margins speculations". When an observer identifies a suspected mass region during a case interpretation he/she first rates the margins and the computerized search is then limited only to regions rated as having similar levels of spiculation (within +/-1 scale difference). In an observer preference study including 85 test regions, two sets of the six "similar" reference regions selected by the KNN with and without the interactive step were displayed side by side with each test region. Four radiologists and five nonclinician observers selected the more appropriate ("similar") reference set in a two alternative forced choice preference experiment. All four radiologists and five nonclinician observers preferred the sets of regions selected by the interactive method with an average frequency of 76.8% and 74.6%, respectively. The overall preference for the interactive method was highly significant (p < 0.001). The study demonstrated that a simple interactive approach that includes subjectively perceived ratings of one feature alone namely, a rating of margin "spiculation," could substantially improve the selection of "visually similar" reference images.