Filippo Molinari

Automated diagnosis of epileptic EEG using entropies

Abstract Epilepsy is a neurological disorder characterized by the presence of recurring seizures. Like many other neurological disorders, epilepsy can be assessed by the electroencephalogram (EEG). The EEG signal is highly non-linear and non-stationary, and hence, it is difficult to characterize and interpret it. However, it is a well-established clinical technique with low associated costs. In this work, we propose a methodology for the automatic detection of normal , pre-ictal , and ictal conditions from recorded EEG signals. Four entropy features namely Approximate Entropy ( ApEn ), Sample Entropy ( SampEn ), Phase Entropy 1 ( S1 ), and Phase Entropy 2 ( S2 ) were extracted from the collected EEG signals. These features were fed to seven different classifiers: Fuzzy Sugeno Classifier (FSC), Support Vector Machine (SVM), K-Nearest Neighbour (KNN), Probabilistic Neural Network (PNN), Decision Tree (DT), Gaussian Mixture Model (GMM), and Naive Bayes Classifier (NBC). Our results show that the Fuzzy classifier was able to differentiate the three classes with a high accuracy of 98.1%. Overall, compared to previous techniques, our proposed strategy is more suitable for diagnosis of epilepsy with higher accuracy.

Review: A state of the art review on intima-media thickness (IMT) measurement and wall segmentation techniques for carotid ultrasound

Last 10 years have witnessed the growth of many computer applications for the segmentation of the vessel wall in ultrasound imaging. Epidemiological studies showed that the thickness of the major arteries is an early and effective marker of onset of cardiovascular diseases. Ultrasound imaging, being real-time, economic, reliable, safe, and now seems to become a standard in vascular assessment methodology. This review is an attempt to discuss the most performing methodologies that have been developed so far to perform computer-based segmentation and intima-media thickness (IMT) measurement of the carotid arteries in ultrasound images. First we will present the rationale and the clinical relevance of computer-based measurements in clinical practice, followed by the challenges that one has to face when approaching the segmentation of ultrasound vascular images. The core of the paper is the presentation, discussion, benchmarking and evaluation of different segmentation techniques, including: edge-detection, active contours, dynamic programming, local statistics, Hough transform, statistical modeling, and integration of these approaches. Also, we will discuss and compare the different performance metrics that have been proposed and used to perform the validation. Best performing user-dependent techniques show an average IMT measurement error of about 1μm when compared to human tracings [57], whereas completely automated techniques show errors of about 10μm. The review ends with a discussion about the current standards in carotid wall segmentation and in an overview of the future perspectives, which may include the adoption of advanced and intelligent strategies to let the computer technique measure the IMT in the image portion where measurement is more reliable.

Characterization of a Completely User-Independent Algorithm for Carotid Artery Segmentation in 2-D Ultrasound Images

The analysis of the carotid artery wall is crucial for the diagnosis of serious cardiovascular pathologies or for the assessment of a subjects cardiovascular risk. Several algorithms have been proposed for the segmentation of ultrasound carotid artery images, but almost all require a certain degree of user interaction. We recently developed a completely user-independent algorithm for the segmentation of the common-carotid-artery wall; specifically, the algorithm traces the contour of the interfaces between the lumen and the intima layer and between the media and adventitia layers. In this paper, we show the characterization of the algorithm in terms of segmentation error. Moreover, we compare the output of the algorithm with the segmentations manually traced by four experts, using the percent statistics test and testing the automatically generated segmentation against the average human segmentations. We show that our algorithms segmentation is not statistically different from that of a trained operator and that the segmentation error is lower than 1 pixel for both the lumen-intima interface and for the media-adventitia interface.

US-Guided Percutaneous Radiofrequency Thermal Ablation for the Treatment of Solid Benign Hyperfunctioning or Compressive Thyroid Nodules

The aim of the study was to define the effectiveness and safety of ultrasound-guided percutaneous radiofrequency (RF) thermal ablation in the treatment of compressive solid benign thyroid nodules. Thirty-one patients not eligible for surgery or radioiodine (131I) treatment underwent RF ablation for benign nodules; a total of 33 nodules were treated (2 patients had 2 nodules treated in the same session): 10 cold nodules and 23 hyperfunctioning. Fourteen patients complained of compressive symptoms. Nodule volume, thyroid function and compressive symptoms were evaluated before treatment and at 1, 3 and 6 mo. Ultrasound-guided RF ablation was performed using a Starbust RITA needle, with nine expandable prongs; total exposure time was 6 to 10 min at 95 degrees C in one area or more of the nodule. Baseline volume (measured at the time of RF ablation) was 27.7 +/- 21.5 mL (mean +/- SD), but significantly decreased during follow-up: 19.2 +/- 16.2 at 1 mo (-32.7%; p < 0.001), 15.9 +/- 14.1 mL at 3 mo (-46.4 %; p < 0.001) and 14.6 +/- 12.6 mL at 6 mo (-50.7%; p < 0.001). After treatment, all patients with cold nodules remained euthyroid: five patients with hot nodules normalized thyroid function, and the remaining sixteen showed a partial remission of hyperthyroidism. Besides a sensation of heat and mild swelling of the neck, no major complications were observed. Improvement in compressive symptoms was reported by 13 patients, with a reduction on severity scale from 6.1 +/- 1.4 to 2.2 +/- 1.9 (p < 0.0001). Radiofrequency was effective and safe in reducing volume by about 50% and compressive symptoms in large benign nodules. Hyperfunction was fully controlled in 24% of patients and partially reduced in the others.

An Integrated Approach to Computer-Based Automated Tracing and Its Validation for 200 Common Carotid Arterial Wall Ultrasound Images A New Technique

Objective. Most of the algorithms for the segmentation of the common carotid artery (CCA) wall require human interaction to locate the vessel in the ultrasound image. The aim of this article is to show an accurate algorithm for the computer‐based automated tracing of the CCA in longitudinal B‐mode ultrasound images. Methods. Two hundred images (100 normal CCAs, 50 CCAs with an increased intima‐media thickness, 30 with fibrous plaques, and 20 with anechoic plaques) were processed to delineate the region of interest containing the CCA. The strategy is an integrated approach (carotid artery layer extraction using an integrated approach [CALEXia]) consisting of geometric feature extraction, line fitting, and classification. The output of the algorithm is the tracings of the proximal and distal adventitia layers. Performance of the algorithm was validated against human tracings considered the ground truth. Results. The mean distance errors ± SD using this integrated approach were 1.05 ± 1.04 pixels (0.07 ± 0.07 mm) for proximal or near adventitia and 2.68 ± 3.94 pixels (0.17 ± 0.24 mm) for distal or far adventitia. Sixteen of 200 images were not perfectly traced because of the presence of both plaques and blood backscattering. The computational cost ensures the possibility for near real‐time detection. Conclusions. Although the CALEXia algorithm automatically detects the CCA, it is also robust and validated over a large database. This can constitute a general basis for a completely automated segmentation procedure widely applicable to other anatomies.

Intima-media thickness: setting a standard for a completely automated method of ultrasound measurement

The intima-media thickness (IMT) of the common carotid artery is a widely used clinical marker of severe cardiovascular diseases. IMT is usually manually measured on longitudinal B-mode ultrasound images. Many computer-based techniques for IMT measurement have been proposed to overcome the limits of manual segmentation. Most of these, however, require a certain degree of user interaction. In this paper we describe a new, completely automated layer extraction technique (named CALEXia) for the segmentation and IMT measurement of the carotid wall in ultrasound images. CALEXia is based on an integrated approach consisting of feature extraction, line fitting, and classification that enables the automated tracing of the carotid adventitial walls. IMT is then measured by relying on a fuzzy K-means classifier. We tested CALEXia on a database of 200 images. We compared CALEXia?s performance with those of a previously developed methodology that was based on signal analysis (CULEXsa). Three trained operators manually segmented the images and the average profiles were considered as the ground truth. The average error from CALEXia for lumen-intima (LI) and media- adventitia (MA) interface tracings were 1.46 ? 1.51 pixel (0.091 ? 0.093 mm) and 0.40 ? 0.87 pixel (0.025 ? 0.055 mm), respectively. The corresponding errors for CULEXsa were 0.55 ? 0.51 pixels (0.035 ? 0.032 mm) and 0.59 ? 0.46 pixels (0.037 ? 0.029 mm). The IMT measurement error was equal to 0.87 ? 0.56 pixel (0.054 ? 0.035 mm) for CALEXia and 0.12 ? 0.14 pixel (0.01 ? 0.01 mm) for CULEXsa. Thus, CALEXia showed limited performance in segmenting the LI interface, but outperformed CULEXsa in the MA interface and in the number of images correctly processed (190 for CALEXia and 184 for CULEXsa). Based upon two complementary strategies, we anticipate fusing them for further IMT improvements.

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 Laws 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.

Symptomatic vs. Asymptomatic Plaque Classification in Carotid Ultrasound

Quantitative characterization of carotid atherosclerosis and classification into symptomatic or asymptomatic type is crucial in both diagnosis and treatment planning. This paper describes a computer-aided diagnosis (CAD) system which analyzes ultrasound images and classifies them into symptomatic and asymptomatic based on the textural features. The proposed CAD system consists of three modules. The first module is preprocessing, which conditions the images for the subsequent feature extraction. The feature extraction stage uses image texture analysis to calculate Standard deviation, Entropy, Symmetry, and Run Percentage. Finally, classification is performed using AdaBoost and Support Vector Machine for automated decision making. For Adaboost, we compared the performance of five distinct configurations (Least Squares, Maximum- Likelihood, Normal Density Discriminant Function, Pocket, and Stumps) of this algorithm. For Support Vector Machine, we compared the performance using five different configurations (linear kernel, polynomial kernel configurations of different orders and radial basis function kernels). SVM with radial basis function kernel for support vector machine presented the best classification result: classification accuracy of 82.4%, sensitivity of 82.9%, and specificity of 82.1%. We feel that texture features coupled with the Support Vector Machine classifier can be used to identify the plaque tissue type. An Integrated Index, called symptomatic asymptomatic carotid index (SACI), is proposed using texture features to discriminate symptomatic and asymptomatic carotid ultrasound images using just one index or number. We hope this SACI can be used as an adjunct tool by the vascular surgeons for daily screening.

Cost-Effective and Non-Invasive Automated Benign & Malignant Thyroid Lesion Classification in 3D Contrast-Enhanced Ultrasound Using Combination of Wavelets and Textures: A Class of ThyroScan™ Algorithms

Ultrasound has great potential to aid in the differential diagnosis of malignant and benign thyroid lesions, but interpretative pitfalls exist and the accuracy is still poor. To overcome these difficulties, we developed and analyzed a range of knowledge representation techniques, which are a class of ThyroScan™ algorithms from Global Biomedical Technologies Inc., California, USA, for automatic classification of benign and malignant thyroid lesions. The analysis is based on data obtained from twenty nodules (ten benign and ten malignant) taken from 3D contrast-enhanced ultrasound images. Fine needle aspiration biopsy and histology confirmed malignancy. Discrete Wavelet Transform (DWT) and texture algorithms are used to extract relevant features from the thyroid images. The resulting feature vectors are fed to three different classifiers: K-Nearest Neighbor (K-NN), Probabilistic Neural Network (PNN), and Decision Tree (DeTr). The performance of these classifiers is compared using Receiver Operating Characteristic (ROC) curves. Our results show that combination of DWT and texture features coupled with K-NN resulted in good performance measures with the area of under the ROC curve of 0.987, a classification accuracy of 98.9%, a sensitivity of 98%, and a specificity of 99.8%. Finally, we have proposed a novel integrated index called Thyroid Malignancy Index (TMI), which is made up of texture features, to diagnose benign or malignant nodules using just one index. We hope that this TMI will help clinicians in a more objective detection of benign and malignant thyroid lesions.

AUTOMATIC COMPUTER-BASED TRACINGS (ACT) IN LONGITUDINAL 2-D ULTRASOUND IMAGES USING DIFFERENT SCANNERS

Objective. The aim of this paper is to show an algorithm for the automatic computer-based tracing (ACT) of common carotid artery (CCA) in longitudinal B-mode ultrasound images characterized by four main features: (i) user-independence; (ii) suitability to normal and pathological images; (iii) robustness to noise; and (iv) independent of ultrasound OEM scanner. Methods. Three hundred longitudinal B-mode images (100 normal CCAs, 100 CCAs with increased intima-media thickness, 60 stable plaques, and 40 echolucent plaques) were acquired using three different (GE, Siemens, and Biosound) OEM ultrasound image scanners. The algorithm processed each image to delineate the region of interest containing the CCA. Output of the algorithm are three segmentation lines representing (a) distal (far) and (b) near adventitia layers, and (c) lumen of the CCA. Three operators qualitatively scored the ACTs. Results. The CCA was correctly automatically traced in all the 300 B-mode images. The performance was independent on the image scanner used to acquire the image or on the type of the CCA (healthy versus pathologic). Eight ACTs out of 300 received a poor score after visual inspection due to an automated adventitia tracing that did not correctly follow the CCA wall in a small portion of the image. Conclusions. The proposed algorithm is robust in ACTs of CCA since it is independent of scanner and normal/abnormal wall. This approach could constitute a general basis for a completely automated segmentation procedure.