Application of artificial neural networks for the classification of liver lesions by image texture parameters
TL;DR: In this article, a multilayered back-propagation neural network was used for liver lesion classification using B-scan ultrasound images for normal, hemangioma and malignant livers.
Abstract: Ultrasound imaging is a powerful tool for characterizing the state of soft tissues; however, in some cases, where only subtle differences in images are seen as in certain liver lesions such as hemangioma and malignancy, existing B-scan methods are inadequate. More detailed analyses of image texture parameters along with artificial neural networks can be utilized to enhance differentiation. From B-scan ultrasound images, 11 texture parameters comprising of first, second and run length statistics have been obtained for normal, hemangioma and malignant livers. Tissue characterization was then performed using a multilayered backpropagation neural network. The results for 113 cases have been compared with a classification based on discriminant analysis. For linear discriminant analysis, classification accuracy is 79.6% and with neural networks the accuracy is 100%. The present results show that neural networks classify better than discriminant analysis, demonstrating a much potential for clinical application.
Citations
More filters
TL;DR: It was deduced that a situation in which an electron beam is likely to scatter exists in the tumor, which is consistent with the pathological findings reported from previous studies.
Abstract: In recent years, advances in ultrasonographic techniques have allowed the detection of even small hepatocellular carcinomas. However, many cases do not show distinctive ultrasonographic features. A quantitative ultrasonographic procedure for making the diagnosis would therefore be desirable. It is recognized that radio-frequency (RF) signals from hepatic tumors conform to the K distribution. The K distribution is used for the characterization of the statistical properties of backscattering signals from heterogeneous regions. In this study, RF signals were acquired from hemangioma of the liver (hemangioma) and hepatocellular carcinoma (HCC), and parameters of the K distribution, i.e., φ (the mean square of the amplitude of RF signals) and α (the number of effective scatterers), were calculated for the quantitative ultrasonographic evaluation of hepatic tumors. This study included 10 nodules of 10 patients with hemangioma and 27 nodules of 24 patients with HCC. The A-mode RF signals passing through the hepatic tumors were acquired with an Aloka SSD-1000 ultrasound system, and measurement units comprising 279 points were set on the RF signals corresponding to the tumor lesion. The parameters of the K distribution that were most consistent with the histogram were calculated and assessed. With regard to φ, the values were high in hemangioma, and there was a significant (P < 0.05) difference in the φ value between hemangioma and HCC. On the other hand, there was no significant difference in the α value between hemangioma and HCC. The α values of hemangioma were equivalent to those of HCC, while the φ values for hemangioma were higher than those for HCC. It was deduced from these observations that a situation in which an electron beam is likely to scatter exists in the tumor. This is also consistent with the pathological findings reported from previous studies. The use of these parameters is expected to allow quantitative ultrasonographic evaluation of hepatic tumors.
2 citations
01 Jan 2016
TL;DR: This chapter describes specific, texture-based methods for the detection, characterization and recognition of some severe affections and of their evolution phases, using only information from ultrasound images.
Abstract: This chapter describes specific, texture-based methods for the detection, characterization and recognition of some severe affections and of their evolution phases, using only information from ultrasound images. We perform the recognition of the considered affections in supervised manner, and we also discover the disease evolution phases in unsupervised manner. In both cases, the imagistic textural model is defined, consisting of: the relevant features for the characterization of the disease, respectively of its evolution phase; the specific values of the relevant textural features: arithmetic mean, standard deviation, probability distribution. Advanced texture analysis techniques, consisting of textural microstructure co-occurrence matrices based on Laws’ features, are involved in this process. At the end, the imagistic textural model is validated through powerful, supervised classifiers, the resulting accuracy being around 90%.
2 citations
01 Jan 2014
TL;DR: In this study, liver tumors are classified as hepatocellular carcinoma (cancer) and hemangioma (benign) and by using FCM each suspicious tumor region is automatically extracted from liver images and textural features are obtained.
Abstract: The main objective of this study is to provide a Computer-Aided Diagnosis (CAD) system for the diagnosis process of benign and malignant liver tumors from computed tomography (CT). Also it aimed to evaluate the potential role of Fuzzy Clustering Means (FCM) and neural network in the differential diagnosis processes of liver tumors in CT images. In this study, liver tumors are classified as hepatocellular carcinoma (cancer) and hemangioma (benign). By using FCM each suspicious tumor region is automatically extracted from liver images. Consequently, textural features are obtained. These features are used to train the Neural Network (NN) and classify the tumors. The system distinguishes tumors with high accuracy and is therefore clinically useful.
2 citations
TL;DR: In this article , the authors combined the classical approaches with CNN techniques, within B-mode ultrasound images, and the combination was performed at the classifier level, where CNN features obtained at the output of various convolution layers were combined with textural features, then supervised classifiers were employed.
Abstract: Hepatocellular Carcinoma (HCC) is the most frequent malignant liver tumor and the third cause of cancer-related deaths worldwide. For many years, the golden standard for HCC diagnosis has been the needle biopsy, which is invasive and carries risks. Computerized methods are due to achieve a noninvasive, accurate HCC detection process based on medical images. We developed image analysis and recognition methods to perform automatic and computer-aided diagnosis of HCC. Conventional approaches that combined advanced texture analysis, mainly based on Generalized Co-occurrence Matrices (GCM) with traditional classifiers, as well as deep learning approaches based on Convolutional Neural Networks (CNN) and Stacked Denoising Autoencoders (SAE), were involved in our research. The best accuracy of 91% was achieved for B-mode ultrasound images through CNN by our research group. In this work, we combined the classical approaches with CNN techniques, within B-mode ultrasound images. The combination was performed at the classifier level. The CNN features obtained at the output of various convolution layers were combined with powerful textural features, then supervised classifiers were employed. The experiments were conducted on two datasets, acquired with different ultrasound machines. The best performance, above 98%, overpassed our previous results, as well as representative state-of-the-art results.
1 citations
01 Jan 2019
TL;DR: The inside regions ofinterest (IROIs) have been selected from within the FHTs and surrounding regions of interest (SROIs) has been selectedfrom the homogeneous region around the lesion and at the same depth as that of the lesions center.
Abstract: In the current study, an efficient smooth support vector machine (SSVM)-based hierarchical CAC system has been designed for primary benign and malignant focal hepatic tumors. The work has been carried out on a representative and robust image dataset consisting of 76 liver ultrasound images with (a) 16 hemangioma (typical and atypical HEM) images, (b) 28 hepatocellular carcinoma (small HCC and large HCC) images, and (c) 32 metastatic carcinoma (typical and atypical MET) images. The textural characteristics from inside the regions and outside the FHTs are considered equally important for the differential diagnosis. Therefore, in the current study, the inside regions of interest (IROIs) have been selected from within the FHTs and surrounding regions of interest (SROIs) have been selected from the homogeneous region around the lesion and at the same depth as that of the lesion center. Five texture features are computed from the collective dataset consisting of 255 IROIs. In the current study, the SSVM-based multiclass CAC system design has been compared with the SSVM-based hierarchical CAC system design. The SSVM-based hierarchical CAC system consists of two binary classifiers that are arranged in a hierarchical framework. The SSVM classifier-1 classifies the HEM, HCC, and MET images into primary benign (HEM) and malignant (HCC or MET) cases. The malignant cases are further classified by the SSVM classifier-2 into primary malignant (HCC) and secondary malignant (MET) cases. The overall classification accuracy achieved for the multiclass classifier is 82.6% with 20 misclassification cases out of 115 test instances. However, it has been observed that the hierarchical CAC system yields n overall classification accuracy of 89.6% with 12 misclassification cases out of 115 test instances.
1 citations
References
More filters
01 Nov 1973
TL;DR: These results indicate that the easily computable textural features based on gray-tone spatial dependancies probably have a general applicability for a wide variety of image-classification applications.
Abstract: Texture is one of the important characteristics used in identifying objects or regions of interest in an image, whether the image be a photomicrograph, an aerial photograph, or a satellite image. This paper describes some easily computable textural features based on gray-tone spatial dependancies, and illustrates their application in category-identification tasks of three different kinds of image data: photomicrographs of five kinds of sandstones, 1:20 000 panchromatic aerial photographs of eight land-use categories, and Earth Resources Technology Satellite (ERTS) multispecial imagery containing seven land-use categories. We use two kinds of decision rules: one for which the decision regions are convex polyhedra (a piecewise linear decision rule), and one for which the decision regions are rectangular parallelpipeds (a min-max decision rule). In each experiment the data set was divided into two parts, a training set and a test set. Test set identification accuracy is 89 percent for the photomicrographs, 82 percent for the aerial photographic imagery, and 83 percent for the satellite imagery. These results indicate that the easily computable textural features probably have a general applicability for a wide variety of image-classification applications.
20,442 citations
TL;DR: This paper provides an introduction to the field of artificial neural nets by reviewing six important neural net models that can be used for pattern classification and exploring how some existing classification and clustering algorithms can be performed using simple neuron-like components.
Abstract: Artificial neural net models have been studied for many years in the hope of achieving human-like performance in the fields of speech and image recognition. These models are composed of many nonlinear computational elements operating in parallel and arranged in patterns reminiscent of biological neural nets. Computational elements or nodes are connected via weights that are typically adapted during use to improve performance. There has been a recent resurgence in the field of artificial neural nets caused by new net topologies and algorithms, analog VLSI implementation techniques, and the belief that massive parallelism is essential for high performance speech and image recognition. This paper provides an introduction to the field of artificial neural nets by reviewing six important neural net models that can be used for pattern classification. These nets are highly parallel building blocks that illustrate neural net components and design principles and can be used to construct more complex systems. In addition to describing these nets, a major emphasis is placed on exploring how some existing classification and clustering algorithms can be performed using simple neuron-like components. Single-layer nets can implement algorithms required by Gaussian maximum-likelihood classifiers and optimum minimum-error classifiers for binary patterns corrupted by noise. More generally, the decision regions required by any classification algorithm can be generated in a straightforward manner by three-layer feed-forward nets.
7,798 citations
Book•
01 Jan 1974
TL;DR: The objective is to establish an experimental procedure and show direct AFM progression from EMT to EMT using a simple, straightforward, and reproducible procedure.
Abstract: Pathologic basis of disease , Pathologic basis of disease , کتابخانه دیجیتالی دانشگاه علوم پزشکی و خدمات درمانی شهید بهشتی
5,162 citations
01 Apr 1976
TL;DR: In this paper, three standard approaches to automatic texture classification make use of features based on the Fourier power spectrum, on second-order gray level statistics, and on first-order statistics of gray level differences, respectively.
Abstract: Three standard approaches to automatic texture classification make use of features based on the Fourier power spectrum, on second-order gray level statistics, and on first-order statistics of gray level differences, respectively. Feature sets of these types, all designed analogously, were used to classify two sets of terrain samples. It was found that the Fourier features generally performed more poorly, while the other feature sets all performned comparably.
1,379 citations
TL;DR: The gray value distribution of the runs is proposed to be used to define two new features, viz., low gray level run emphasis ( LGRE) and high gray levelrun emphasis ( HGRE).
443 citations