C. Vicas

Automatic detection of liver capsule using Gabor Filters. Applications in steatosis quantification

In this paper we present a new method of detecting hepatic capsule and other large structures from liver ultrasound images. Local texture anisotropy is evaluated using a specially designed bank of Gabor filters. We show that the proposed method is robust with respect to its parameters and is well suited for processing liver ultrasound images. Two human experts evaluate the method in conjunction with Canny edge detector and a method based on phase congruency. A practical application in quantifying the diffuse liver steatosis is presented. We show that proposed method can successfully replace the human expert in establishing the regions of interest used in steatosis quantification.

Non-invasive steatosis assessment in NASH through the computerized processing of ultrasound images: Attenuation versus textural parameters

Ultrasonography is a simple method in diagnosing nonalcoholic steatohepatitis (NASH), providing useful information, but it is subjective and does not accurately differentiate between steatosis grades. The computerized processing of the data that comprises the ultrasonic image (CPU) might transform ultra-sonography into an objective examination. CPU can be achieved either by methods based on the study of parenchymal echogenicity and on the attenuation of the ultrasounds (attenuation and back-scattering coefficients), or by methods based on the quantification of some textural parameters. In the present paper we set out to compare the performance of the attenuation coefficient (AC) and the textural parameters derived from the GLCM matrix 96 NASH patients and 24 healthy subjects were prospectively included in this study. We found a strong correlation between the AC and steatosis and a weak, but statistically significant one, with balooning and lobular inflammation, but not with fibrosis. The multivariate analysis showed, however, that only steatosis influences independently the AC. Of the analyzed textural parameters, only the GLCM entropy correlated weakly, but significantly, with the steatosis degree. Our study proves that the use of the attentuation coefficient computed on the ultrasonographic image can help differentiate healthy from NASH patients, as well as discriminate between various degrees of fatty load. The attenuation coefficient performs better than the textural parameters derived from the GLCM matrix. However, only GLCM entropy, of all textural parameters tested, correlates with steatosis, and even then, only for the differentiation normal vs NASH, not between steatosis grades.

Influence of Expert-Dependent Variability over the Performance of Noninvasive Fibrosis Assessment in Patients with Chronic Hepatitis C by Means of Texture Analysis

Texture analysis is viewed as a method to enhance the diagnosis power of classical B-mode ultrasound image. The present paper aims to evaluate and eliminate the dependence between the human expert and the performance of such a texture analysis system in predicting the cirrhosis in chronic hepatitis C patients. 125 consecutive chronic hepatitis C patients were included in this study. Ultrasound images were acquired from each patient and four human experts established regions of interest. Textural analysis tool was evaluated. The performance of this approach depends highly on the human expert that establishes the regions of interest (P < 0.05). The novel algorithm that automatically establishes regions of interest can be compared with a trained radiologist. In classical form met in the literature, the noninvasive diagnosis through texture analysis has limited utility in clinical practice. The automatic ROI establishment tool is very useful in eliminating the expert-dependent variability.

Comparison between attenuation coefficient computed on the ultrasound image and a biological marker, adiponectin, in the diagnosis of steatosis in non-alcoholic fatty liver disease

The study aimed to identify the presence of steatosis and differentiate between insignificant and significant steatosis in non-alcoholic fatty liver disease (NAFLD). Thirty three patients with NAFLD proven by liver biopsy and evaluated in order to stratify the degree of steatosis were studied. The computerized processing of the data of the ultrasonic image (attenuation coefficient - LS7 slope), and a biological marker (adiponectin) were evaluated. Predictive value and AUROC curves were used to assess the accuracy of the results. Adiponectin levels was found to differentiate between controls (13435.4plusmn741.5 ng/ml) and NAFLD (4552.9plusmn2473 ng/ml) but also between insignificant (9348.4plusmn7368 ng/ml) and significant steatosis (4501plusmn2389.6 ng/ml). The AUROC for adiponectin, was 0.899 and for LS7 slope 0.952 with sensitivity, specificity, positive predictive value and negative predictive value of 84.8; 93.3; 96.5; 73.7, respectively 84.8;100;90.3 and 70.6 for the diagnosis of hepatic steatosis. The simultaneous utilization of these methods could substantial improve the identification of steatosis in NAFLD and the assessment of its grade.

Ultrasonographic diagnosis of nonalcoholic steatohepatitis based on the quantitative evaluation of the ultrasound beam behavior into the liver

Ultrasonography is a simple method in diagnosing nonalcoholic steatohepatitis (NASH), providing useful information, but it is subjective and does not accurately differentiate between steatosis grades. The computerized processing of the data that comprises the ultrasonic image might transform ultrasonography into an objective examination. We aim to study the performance of the quantitative evaluation of the ultrasound beam behavior into the liver in diagnosing NASH and in establishing the fatty load grade.

Non-invasive Steatosis Assessment through the Computerized Processing of Ultrasound Images: Attenuation versus First Order Texture Parameters

Steatosis is a frequent histological finding in patients with chronic hepatitis C virus (VHC) infection. Usual ultrasonography (US) cannot accurately detect the steatosis grade, nor can it always discriminate between steatosis and fibrosis. An improvement of usual US examination is currently under research. A possible approach might be the computerized processing of the data comprised in the US image. In the present paper we set out to compare the performance of two computerized methods for the steatosis assessment on the US images: the attenuation coefficient and the first order textural parameters (FO): Mean, Standard Deviation and Skewness. The attenuation coefficient correlated significantly with steatosis (r=-0.444, p<0.0001), but not with fibrosis (r=-0.046, p=0.395) or necroinflammatory activity (r=-0056, p=0.211). Of the FO parameters, only the FO mean correlated significantly with steatosis (r=0.300, p<0.0001), but also with necroinflammatory activity (r=0.128, p=0004). The present study proves that, in patients having chronic hepatitis C, the attenuation coefficient, but also the FO mean, can discriminate between different steatosis grades; however, the attenuation coefficient has a better performance than the FO mean, being influenced only by steatosis, not by fibrosis or necroinflammatory activity. The area under the ROC curve is significantly better for the attenuation coefficient as compared to the FO mean for the prediction of steatosis regardless of the grade (0.741 vs 0.652, p=0.001), as well as for the prediction of moderate/severe steatosis (0.791 vs 0.719, p=0.043).

Detection of anatomical structures on ultrasound liver images using Gabor filters

In this paper we present a novel method of detecting anatomical structures from liver ultrasound images. Local texture anisotropy is evaluated using a specially designed bank of Gabor filters. We show that the proposed method is well suited for processing liver ultrasound images and it is robust with respect to its parameters. A practical application of this method is then presented. Image features are computed on detected structures. These features are then used in conjunction with a classifier (Support Vector Machines) to discriminate between left and right lobe liver images. This classifier is used to label ultrasound images. The labeling algorithm obtained a very low error rate (<1%). It can successfully replace the human expert in image labeling.

The diagnostic performance of attenuation coefficient computed on the ultrasound image compared to a biochemical marker — SteatoTest — for steatosis quantification in non-alcoholic fatty liver disease

The aim of this study was to establish the diagnostic performance and comparison of a biological marker (SteatoTest) and an innovative imagistic parameter (the attenuation coefficient-AC) to identify hepatic steatosis and to assess its degree. Seventy seven patients with NAFLD morphologicaly proven by liver biopsy were prospectively studied, compared with 16 healthy subjects. The patients were stratify according to the degree of hepatic steatosis. The computerized processing of the data offered by ultrasonic image (AC) and a blood test (SteatoTest) were determined. The diagnostic value for each method was assessed using sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV) and the area under the receiving operating characteristics curve (AUROC). SteatoTest was found to differentiate between controls (0.173 ± 0.079) and NAFLD controls (0.744 ± 0.166) but failed to discriminate the degree of steatosis. AC had a good diagnostic value in detection of steatosis (controls (0.320 ± 0.502) and NAFLD (−0.122 ± 0.063) and could also discriminates between grades of steatosis. In conclusions our finding suggest that AC and Steat-Test could be used for detection of steatosis. A combination of these parameters might increase the diagnostic performance.

Detecting Curvilinear Features Using Structure Tensors

Few published articles on curvilinear structures exist compared with works on detecting lines or corners with high accuracy. In medical ultrasound imaging, the structures that need to be detected appear as a collection of microstructures correlated along a path. In this paper, we investigated techniques that extract meaningful low-level information for curvilinear structures, using techniques based on structure tensor. We proposed a novel structure tensor enhancement inspired by bilateral filtering. We compared the proposed approach with five state-of-the-art curvilinear structure detectors. We tested the algorithms against simulated images with known ground truth and real images from three different domains (medical ultrasound, scanning electron microscope, and astronomy). For the real images, we employed experts to delineate the ground truth for each domain. Techniques borrowed from machine learning robustly assessed the performance of the methods (area under curve and cross validation). As a practical application, we used the proposed method to label a set of 5000 ultrasound images. We conclude that the proposed tensor-based approach outperforms the state-of-the-art methods in providing magnitude and orientation information for curvilinear structures. The evaluation methodology ensures that the employed feature-detection method will yield reproducible performance on new, unseen images. We published all the implemented methods as open-source software.

DETECTION AND STAGING OF LIVER FIBROSIS USING ADDITIVE LOGISTIC MODELS

Fibrosis and cirrhosis are the main complications of chronic liver diseases. At present, liver biopsy is the golden standard for evaluating liver fibrosis. However, this is an invasive procedure, hence the interest in developing non–invasive approaches. The present study identifies novel possibilities for non-invasive fibrosis evaluation. We included 591 hepatitis C patients. Fibrosis was assessed using the Metavir score. A number of 93 features were obtained from each patient using B-mode ultrasound, Doppler ultrasound, transient elastography and common biochemical and cytological measurements. The patients were grouped according to fibrosis stages and additive logistic regression models were built. Crossvalidation along with Area Under Curve (AUROC) was used to measure the classification performance. The AUROC of 0.90 was recorded when discriminating between fibrosis stage ≤3 and fibrosis stage 4.