FACULTATEA DE AUTOMATICĂ ŞI CALCULATOARE. Eng. Cristian Marian Vicaş PHD THESIS. Abstract

Transcription

1 FACULTATEA DE AUTOMATICĂ ŞI CALCULATOARE Eng. Cristian Marian Vicaş PHD THESIS Abstract IMPROVING THE DIAGNOSIC VALUE OF THE MEDICAL IMAGING BY THE MEANS OF IMAGE PROCESSING TECHNIQUES Advisor, Prof. Dr. Eng. Sergiu Nedevschi 2011

2 Abstract The main purpose of this thesis is to increase the diagnostic value of the medical imaging examinations by researching and developing non-invasive diagnostic techniques. The instruments that are used in this process are those found in image processing and computer vision areas. This paper develops specific methodologies for evaluating supervised learning algorithms and will propose certain algorithms and methods that automatically detect certain features of ultrasound images, features that will be used in diagnosis automation. Also, in this paper we propose a model of liver tissue that allows detailed examination of various factors that contribute in a positive or negativeway to the quality of diffuse pathologies diagnostic. This paper has a multidisciplinary character. The methods, the algorithms and the methodologies developed here are specific to computer vision. Practical applications, however, are focused mainly in the medical field, with emphasis on developing tools that are usefull in clinical practice. The methods that I proposed in this paper are compared with established methods in literature. Rigorous validation proves that the methods proposed here are robust and able to fulfill the purpose for which they were developed. Even if the proposed methods are highly theoretical, practical applications may be found in the field of non invasive diagnostic examinations in the context of diffuse liver pathologies. I investigated extensively the methodology that is traditionally accepted by the scientific community and I clearly showed the contributions and benefits of the proposed methods. The methods and the algorithms are implemented rigorously and are validated and tested according to procedures accepted in the literature. The thesis is structured as follows: Chapter 1 is an introduction to the study in which I justify my choice for the research. Chapter 2 is a concise review of the literature focused on the diagnosis of diffuse liver diseases using textural analysis. Here is presented the accepted methodology and its results. This chapter also shows the perception of textural analysis in the medical community. It identifies the main limitations of the encountered approaches and the research directions that can be improved. Chapter 3 is devoted entirely to the processes and phenomena that allow the creation of ultrasound imaging in mode B. In addition to physical phenomena, the underlying processes are shown and explained. I also presented all the steps of the creation going from raw signal to the image shown on the monitor. The discussions are focused on characteristics of the ultrasound images. I presented here also the formation of the most common types of artifacts. Knowledge of these phenomena is particularly important when one wants to establish an ultrasound image acquisition protocol. Establishing arbitrary values for the machine settings may result in the acquisition of ultrasound images with a lower ratio between useful information and artifacts that are unavoidable in ultrasound examination. Chapter 4 presents a major contribution, namely the liver tissue model. Based on various studies of the phenomena that contribute to the generation of ultrasound imaging and studies on the liver fibrosis model, I developed a tissue model that mimics the structure of the microscopic anatomy of the liver. This tissue model is used to generate ultrasound images that can then be processed using textural analysis. The model allowes also simulation of both the healthy tissue and various degrees of fibrosis. In this chapter I detalied the tissue model, model parameters and pathologies that can be simulated using them. Ultrasound images acquired from patients suffer in general from many shortcomings such as: the presence of artifacts, the variability introduced by the medical examiner, the 2

3 variability of biopsy diagnosis and the limited amount of patients that can be included in the study. By knowing the anatomical structure of the liver tissue and how the production of ultrasound images occurs we can say that it is possible to develop a model capable to meet the limitations outlined above. The liver tissue model allowes obtaining objective results in the non invasive diagnosis of liver fibrosis. The tissue model introduced here is the most detailed model existing in literature. The developed model and the conclusions drawn based on this model were disseminated in the international scientific community in a paper published in a prestigious journal in the field of medical ultrasound. Chapters 5 is focused on the texture analysis. Textural analysis is the main tool used to improve the diagnostic power of ultrasound imaging. Textural analysis has two fundamental components. The first component is responsible for numerically describing the texture and is performend by textural algorithms. The second component consists of a supervised learning algorithm (classified) working on textural descriptions. In this chapter I presented in detail the methods for describing texture and supervised learning methods that are used in non invasive diagnosis. In this chapter I presented also another important contribution to the field of non invasive diagnosis: a methodology for rigorous training and evaluation of classifiers. This methodology was developed taking into account the peculiarities of medical data (small volume, missing values, noise, etc.). In the same time, the methodology follows a more precise estimation of performance learning schemes minimizing the positive bias. Proposed methods are validated on several synthetic data sets. I highlighted the strengths and weaknesses of the classification schemes used throughout this work. Chapter 6 introduces a significant contribution to the field of computer vision. Here I developed an original method to detect artifacts by localised image anisotropy quantification. This quantification takes place only in a certain spectrum. The method is then used to develop algorithms that allow the replacement of human expert in an important step of textural analysis, namely the establishment of regions of interest. The anisotropy detection method is particularly robust and useful when processing images generated using a monochromatic and coherent radiation. These images will show a speckle texture type due to interference process inherent in ultrasound visualisation. Developing a method for automatically determining the regions of interest in ultrasound images is particularly useful because it eliminaties the variability introduced by human experts in positioning the region of interest. Such an algorithm increases the reproductibility of a non-invasive diagnostis method based on textural analysis. An algorithm for automatically determining the regions of interest must meet the criteria accepted in the literature. The regions of interest must be placed in a certain position relative to the image geometry, so as to capture the liver tissue but do not overlap any anatomical structures and artifacts that might be present in the ultrasound image. Of all the requirements listed, the most difficult to accomplish without resorting to a human expert is related to avoiding artifacts. To achieve this goal it is necessary to detect ultrasound image areas which may contain artifacts or anatomical structures. In this chapter I propose an innovative method to detect artifacts, a method that is based on local characterization of the properties of ultrasound image. This method is based on the fact that liver tissue is a relatively homogeneous with the same macroscopic and microscopic properties, regardless of the studied direction. Therefore, an ultrasound image of a homogeneous region of liver tissue must have the same characteristics regardless of the examined direction. In addition to liver tissue in the liver one can find the liver capsule, biliary tracts, blood vessels etc. These normal anatomical structures, along with other artifacts (shadows cost, reverberations, etc.) change the local image properties. Moreover, the 3

4 properties do not change the same for all directions, because artifacts usually have a linear symmetry. The presence of an artifact in a local region of the image will generate a degree of anisotropy. Detection and characterization of image regions with a higher degree of anisotropy will immediately detect artifacts. The physics and the signal processing of the returned echo implies the existence of an inherent anisotropy of the image. This complicates the detection of artifacts because, implicitly, all the ultrasound images have a significant degree of anisotropy. Moreover, some of the other properties of the image (function PSF) depend on the position relative to the transducer. This dependence complicates the anisotropy model. A very important aspect that is to be noted is that the intrinsic anisotropy is manifested in high frequency register (frequency in the meaning of Fourier frequency, not ultrasonic signal). The original method developed here characterizes the anisotropy localised in space and in the frequency domain. The proposed method can discriminate between the anisotropy given by a natural artifact that is present in human tissue and intrinsic image anisotropy, generated by the acquisition processes. The proposed method uses a band pass filter in order to determine the degree of anisotropy by characterizing the energy distribution according to the orientation. Original contributions ensure that the method is not sensitive to noise and that the method is rotation invariant. Anisotropy detection method proposed here is an important contribution to the field of computer vision because it is a very robust method when applied to images generated using a coherent radiation. Practical applications of anisotropy detection make a major contribution to the non invasive diagnosis because they allow automatic establishment of regions of interest on the ultrasound image, with all the benefits that occur. The results obtained by applying proposed method on a large batch of images (over 10,000) are similar to results obtained by employing a human expert. Chapter 7 is devoted to experiments and practical applications to non-invasive diagnosis. The chapter begins with a short presentation of the system arhitecture and the main modules. I presented the patients (over 750), the ultrasound images and the medical features recorded for each patient. In this section I presented another original contribution to the field of non invasive diagnostic method, an algorithm that is able to discriminate between two common types of ultrasound images. The large volume of ultrasound images required to develop a method for automatic filtering and labeling of useful images. Traditionally, the ultrasound image labelling was a human expert task. The performance of the proposed algorithm allowed me to free the human expert of this task. The chapter continues with comprehensive studies and experiments on non invasive detection and diagnosis of the two diseases: steatosis and hepatic fibrosis with emphasis on the performance of texture analysis system in these pathologies. Here I show the limitations of the textural analysis tool in fibrosis diagnosis and I study the impact induced by the expert quality and the impact of different parameters of the normal liver tissue on the diagnosis performance. In this chapter I presented also an original method for non-invasive detection and diagnosis of fibrosis, using algorithms developed in previous chapters. This chapter concluded with an extensive commentary on non-invasive diagnosis of diffuse liver pathologies. I should mention that research carried out in this work led to the creation of two software tools that allow the detection and diagnosis of diffuse pathologies, tools currently used in clinical practice of radiologists. Chapter 8 concludes the paper. 4

5 Present paper brings important contributions to the field of computer vision and non invasive diagnosis. The proposed method of local anisotropy resolution dependent quantification is very robust and its applications in artifact detection automate the diagnosis in both fibrosis and steatosis diseases. Resolution dependent anisotropy detection is a fundamental contribution to the field of computer vision. This method is not limited to ultrasound images and it can be successfully applied in other imaging areas, where a coherent radiation is used, and the captured images have the speckle appearance. In this paper we evaluated the possibility of non-invasive detection of diffuse liver pathologies using domain-specific image processing and computer vision methods. I showed the possibilities and limitations of current methods presented in literature. I proposed a liver tissue model to assess the impact of various factors on fibrosis detection. These factors can be quantified precisely in the model, a fact that can not happen if images are from actual patients. Proposed methods in computer vision have been applied successfully to improve diagnosis performance of ultrasound imaging. The results and contributions to non invasive diagnosis are highly relevant due to the large volume of patients and the use of liver biopsy as diagnostic reference. The most important personal contributions to the image processing and non invasive diagosis are summarized in the following: A thorough study of scientific literature dedicated to non invasive diagnosis. The study was focused not only on the computer vision field but to the medical field also. A comprenhesive description on how an ultrasound image is generated. I presented the processes and the artifacts that are inheritent to ultrasound imaging. I proposed a methodology to train and evaluate supervised learning algorithms. This methodology ensures that the results have a high degree of relevance and objectivity. I proposed a liver tissue model. This model is the most detalied one that exists in literature. The model was used to objectively study the impact of liver fibrosis on ultrasound image and to assess the discrimination power of the texure analysis versus the human vision. The model addresses all the limitations imposed by the actual patient images. The resolution dependent anisotropy quantification is a powerful and robust algorithm. The proposed method was rigurously tested and validated. This method was designed to handle the shortcomings of ultrasound images. In the same time, this method is very useful in other imaging areas, where a coherent radiation is employed. Artifact detection is possible using anisotropy quantification snd this is another important contribution to the non invasive diagnosis field. The automatisation of textural analysis tool ensures reproductibility and objective results for these non invasive diagnosis tools. The experiments presented in this thesis include a large number of patients. Liver biopsy was used as a reference diagnosis. These facts, along with a rigurous evaluation methodology ensured that the obtained experimental results have a high impact in the medical field. This paper proposes two non invasive diagnosis tools: one for detecting and grading liver steatosis and one for detecting and staging of liver fibrosis. They are at the disposal of medical clinicians and are used currently in clinical practice. 5

6 Selected references [1] S. Bonekamp, I. Kamel, S. Solga, and J. Clark, "Can imaging modalities diagnose and stage hepatic fibrosis and cirrhosis accurately?," Journal of Hepatology, vol. 50, pp , Jan [2] P. Bedossa, D. Dargere, and V. Paradis, "Sampling variability of liver fibrosis in chronic hepatitis C," Hepatology, vol. 38, pp , [3] T. Yamaguchi and H. Hachiya. (2010 July Proposal of a parametric imaging method for quantitative diagnosis of liver fibrosis. Journal of Medical Ultrasonics, 12. Available: [4] J. Jan, Medical image processing, reconstruction, and restoration: concepts and methods. Boca Raton,FL: Taylor & Francis Group, [5] D. Boukerroui, J. A. Noble, and M. Brady, "On The Selection Of Band-Pass Quadrature Filters," in Frontiers in robotics research, M. A. Denket, Ed., ed New York: Nova Science Publishers, 2006, pp [6] Witten and E. Frank, Data Mining: Practical machine learning tools and techniques: Morgan Kaufmann Pub, [7] Y. Xia, D. G. Feng, and R. C. Zhao, "Morphology-based multifractal estimation for texture segmentation," Ieee Transactions on Image Processing, vol. 15, pp , Mar Published papers ISI journals, with impact factor [1] C. Vicas, M. Lupsor, R. Badea, and S. Nedevschi, "Usefulness of Textural Analysis as a Tool for Non- Invasive Liver Fibrosis Staging, " Journal of Medical Ultrasonics, 2011, DOI: /s x BDI indexed journals (Mathematical Reviews) [2] C. Vicas, M. Lupsor, M. Socaciu, R. Badea, and S. Nedevschi, "Liver Fibrosis detection by the means of texture analysis. Limitations and further development directions," Automat. Comput. Appl. Math., vol. 19, pp , [3] C. Vicas, S. Nedevschi, M. Lupsor, and R. Badea, "Fibrosis detection using ultrasound and serological markers as features for additive logistic regression.," Automat. Comput. Appl. Math., vol. 18, pp , [4] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, H. Stefãnescu, M. Socaciu, C. Radu, D. Crisan, and M. Gigorescu, "The performance of attenuation coefficient computed on the ultrasound image in quantifying liver steatosis in diffuse liver diseases patients," Automat. Comput, Appl. Math., vol. 18, pp , [5] S. Nedevschi, C. Vicas, M. Lupsor, R. Badea, and M. Grigorescu, "The employment of textural and non textural image analysis algorithms in assessing the diffuse liver diseases," Automat. Comput, Appl. Math., vol. 17, pp , [6] R. Badea, M. Lupsor, H. Stefanescu, C. Vicas, D. Mitrea, M. Socaciu, and S. Nedevschi, "Integrated procedures of evaluating the hepatic morphology, for the non-invasive detection and quantification of steatosis and fibrosis based on the ultrasound diagnosis," Automat. Comput, Appl. Math., vol. 17, pp. 1-11, [7] C. Vicas, S. Nedevschi, M. Lupsor, R. Badea, and M. Grigorescu, "Steatohepatitis Detection from Ultrasound Images Using Attenuation and Backscattering Coefficients," Automat. Comput, Appl. Math., vol. 16, pp , [8] C. Vicas, S. Nedevschi, M. Lupsor, and R. Badea, "Fibrosis detection from ultrasound imaging. The influence of necro-inflammatory activity and steatosis over the detection rates," Automat. Comput, Appl. Math., vol. 16, pp , ISI indexed conferences [9] C. Vicas, S. Nedevschi, M. Lupsor, and R. Badea, "Automatic detection of liver capsule using Gabor filters. Applications in steatosis quantification," IEEE Proc.Intelligent Computer Comm. Process., pp , Aug [10] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefanescu, C. Radu, D. Crisan, Z. Sparchez, A. Serban, and H. Branda, "Ultrasonographic diagnosis of nonalcoholic steatohepatitis based on the quantitative evaluation of the ultrasound beam behavior into the liver," Proc. Int. Conf. Automation, Quality and Testing, Robotics, pp ,

7 [11] M. Grigorescu, C. Radu, M. Lupsor, C. Vicas, S. Nedevschi, R. Badea, Z. Sparchez, D. Crisan, and A. Serban, "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," Proc. Int. Conf. Automation, Quality and Testing, Robotics, vol. 3, pp BDI conferences (indexed IEEEXplore, INSPEC, SCOPUS) [12] C. Vicas, S. Nedevschi, M. Lupsor, and R. Badea, "Detection and Staging of Liver Fibrosis Using Additive Logistic Models," in Eleventh IASTED Int. Conf. Artificial Intelligence and Applications, Innsbruck, Austria, 2011, pp [13] C. Vicas and S. Nedevschi, "Performance Evaluation of the Textural Analysis Algorithms in Liver Fibrosis Detection Using Ultrasound Software Phantoms," in 2010 IEEE 6th Int. Conf. Intelligent Computer Communication and Processing, Cluj-Napoca, 2010, pp [14] C. Vicas, M. Lupsor, R. Badea, and S. Nedevschi, "Detection of anatomical structures on ultrasound liver images using Gabor filters " in Automation Quality and Testing Robotics (AQTR), 2010 IEEE International Conference on Cluj Napoca, 2010, pp [15] C. Radu, M. Grigorescu, M. Lupsor, C. Vicas, S. Nedevschi, R. Badea, M. D. Grigorescu, Z. Sparchez, D. Crisan, and D. Feier, "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 " in Automation Quality and Testing Robotics (AQTR), 2010 IEEE International Conference on Cluj Napoca, 2010, pp [16] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, C. Radu, H. Stefanescu, and D. Crisan, "Non-invasive steatosis assessment in NASH through the computerized processing of ultrasound images: Attenuation versus textural parameters " in Automation Quality and Testing Robotics (AQTR), 2010 IEEE International Conference on 2010, pp Other international conferences [17] S. Nedevschi, C. Vicas, D. Mitrea, M. Lupsor, M. Grigorescu, and R. Badea, "Usefulness Of Attenuation And Backscattering Coefficients In Investigating Complex Nonalcoholic Steatohepatitis From Ultrasound Images. Preliminary Results," Proc. 19 th Biosignal Conference, Brno, p. paper ID 126, Jun [18] M. Lupsor, S. Nedevschi, C. Vicas, R. Badea, M. Grigorescu, S. Tripon, H. Stefanescu, C. Radu, A. Serban, and T. Suteu, "Estimating the fibrosis stage in the human liver tissue using image processing methods on ultrasonographic images. Preliminary results," Proc. EMMIT, 3rd Ed., pp , Romanian journals, graded B,C,D by CNCSIS [19] R. Badea, M. Lupsor, H. Stefanescu, S. Nedevschi, C. Vicas, and D. Mitrea, "Techniques for optimizing the ultrasound method used for accurate discovery and non-invasive evaluation of steatosis, fibrosis and liver restructuring SIDEF," Proceedings of ICCP nd Workshop on Computers in Medical Diagnosis, Technical University of Cluj-Napoca, pp. 1-9, Aug [20] S. Nedevschi, C. Vicas, M. Lupsor, R. Badea, and M. Grigorescu, "Usefulness of image processing techniques in assessing the diffuse liver diseases," Proc. ICCP 2008, 2nd Workshop Computers in Medical Diagnosis, Cluj-Napoca, pp , Aug [21] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefanescu, Z. Sparchez, H. Branda, C. Radu, D. Crisan, and A. Serban, "The diagnosis performance of the attenuation coefficient computed on the ultrasonic image in comparison with the classical ultrasonographic examination, to quantify the hepatic fat content in nonalcoholic steatohepatitis patients," Romanian Journal of Hepatology, vol. 4, pp , [22] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefanescu, Z. Sparchez, C. Radu, D. Crisan, and A. Serban, "Detection of steatosis in chronic hepatitis C, based on the evaluation of the attenuation coefficient computed on the ultrasound image," Medical Ultrasonography, vol. 10, pp , [23] M. Lupsor, R. Badea, S. Nedevschi, C. Vicas, S. Tripon, H. Stefanescu, C. Radu, and M. Grigorescu, "The assessment of liver fibrosis using the computerized analysis of ultrasonographic images. Is the virtual biopsy appearing as an option?," Acta Electrotehnica, vol. 48, pp , Book chapters [24] M. Lupsor, R. Badea, D. Mitrea, C. Vicas, and S. Nedevschi, "Evaluarea şi caracterizarea steatozei, fibrozei şi restructurării parenchimului hepatic cu ajutorul ultrasonografiei şi a metodelor computerizate de analiză a imaginii," in Actualităţi în patologia hepatică, M. Grigorescu and M. Beuran, Eds., ed Cluj Napoca: Editura Medicală Universitară Iuliu Haţieganu, 2008, pp

8 Abstracts and oral presentations [25] C. Radu, M. Grigorescu, M. Lupsor, C. Vicas, S. Nedevschi, R. Badea, Z. Sparchez, D. Crisan, and A. Serban, "Performanţa determinării coeficientului de atenuarea al ultrasunetelor comparativ cu un marker biologic (adiponectina) pentru diagnosticul steatozei în FGNA.," J Gastrointestin Liver Dis., vol. 18, p. 82, [26] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, H. Stefãnescu, M. Grigorescu, Z. Sparchez, H. Branda, C. Radu, D. Crisan, A. Serban, and A. Maniu, "Performanţa determinării coeficientului de atenuare al ultrasunetelor comparativ cu examinarea ultrasonografică uzuală pentru cuantificarea steatozei la pacienţii cu steatohepatită nonalcoolică.," Medical Ultrasonography, vol. 11, pp , [27] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, H. Stefãnescu, M. Grigorescu, M. Radu, D. Crisan, Z. Sparchez, H. Branda, A. Serban, and A. Maniu, "The diagnosis performance of the attenuation coefficient computed on the ultrasonic image in comparison with the classical ultrasonographic examination, to quantify the hepatic fat content in nonalcoholic steatohepatitis patients.," Proceedings of EASL Special Conference on NAFLD/NASH and related metabolic disease, p. 169, [28] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefãnescu, C. Radu, D. Crisan, M. Socaciu, Z. Sparchez, and A. Serban, "The performance of attenuation coefficient computed on the ultrasound image in quantifying liver steatosis in nonalcoholic steatohepatitis patients.," GUT, vol. 58, p. A197, [29] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefãnescu, C. Radu, D. Crisan, M. Socaciu, Z. Sparchez, H. Branda, and A. Serban, "Cuantificarea steatozei in hepatita cronică virală C prin calcularea coeficientului de atenuare al ultrasunetelor pe imaginea ecografică.," J Gastrointestin Liver Dis., vol. 18, p. 16, [30] M. Lupsor, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefãnescu, Z. Sparchez, A. Serban, H. Branda, M. Florea, and R. Badea, "Steatohepatitis Detection from Ultrasound Images Using Attenuation and Backscattering Coefficients.," Ultraschall Med, vol. 29, p. S8, [31] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefãnescu, Z. Sparchez, C. Radu, D. Crisan, A. Serban, and M. Socaciu, "The performance of attenuation coefficient computed on the ultrasound image in quantifying liver steatosis in chronic hepatitis C patients.," GUT vol. 57, p. A45, [32] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefãnescu, Z. Sparchez, C. Radu, D. Crisan, A. Serban, and M. Socaciu, "Performanţa diagnostica a coeficientului de atenuare calculat pe imaginea ecografică în detecţia steatozei din hepatita cronica virală C. Rezultate pe un lot de 524 pacienţi.," Romanian Journal of Hepatology, vol. 4, pp , [33] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefãnescu, Z. Sparchez, C. Radu, D. Crisan, A. Serban, and H. Branda, "Diagnosis value of the attenuation coefficient computed from ultrasound image for predicting the presence and severity of nonalcoholic steatohepatitis.," Ultraschall Med, vol. 29, pp. S11-S12, [34] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, M. Grigorescu, H. Stefãnescu, Z. Sparchez, C. Radu, D. Crisan, and A. Serban, "Detection of steatosis in chronic hepatitis C, based on the evaluation of the attenuation coefficient computed on the ultrasound image.," J Gastrointestin Liver Dis., vol. 17, pp , [35] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, S. Tripon, H. Stefanescu, T. Suteu, C. Radu, and M. Grigorescu, "Estimarea stadiului fibrozei în hepatita cronică virală C folosind metode computerizate de analiza a imaginii ecografice. Rezultate preliminare.," J Gastrointestin Liver Dis., vol. 16, pp , [36] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, H. Stefanescu, and M. Grigorescu, "Spre biopsia hepatică virtuală. Detectarea ultrasonografică a steatohepatitei, folosind coeficienţii de atenuare şi reflexie rezultate preliminare.," Romanian Journal of Hepatology, vol. 3, pp , [37] M. Lupsor, R. Badea, C. Vicas, S. Nedevschi, H. Stefãnescu, S. tripon, T. Suteu, C. Radu, and M. Grigorescu, "Estimating the fibrosis stage in chronic hepatits C patients using image processing methods on ultrasonographic images. Preliminary results.," Ultraschall Med, vol. 28, pp. S1-S74, Citations The paper: C. Radu, M. Grigorescu, M. Lupsor, C. Vicas, S. Nedevschi, R. Badea, et al., "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," in Automation Quality and Testing Robotics (AQTR), 2010 IEEE International Conference on Cluj Napoca, 2010, pp Was cited by: S. İçer, A. Coşkun, and T. İkizceli, "Quantitative Grading Using Grey Relational Analysis on Ultrasonographic Images of a Fatty Liver," Journal of Medical Systems, pp. 1-8, 2011, DOI /s z 8