Balaji GN | Vellore Institute of Technology (original) (raw)

Papers by Balaji GN

The automatic text extraction system involves intelligent algorithms to identify and extract the ... more The automatic text extraction system involves intelligent algorithms to identify and extract the textual content present in various kinds of images. With the advent of the digital era and the availability of myriad of multimedia contents, it has become extremely important to read and interpret the texts associated with those contents. The automatic extraction of texts would not only serve to infer the semantics of those multimedia documents but also help in effi cient indexing and subsequent retrieval of the same. However, the text differs in size, style, alignment etc. and low resolution of the background of complex images make the problem of text identifi cation a complex one. Hence, the extraction of text data in images has become a challenging fi eld of research in the domain of Image Processing. The main limitation of the existing techniques such as texture-based or connected-component based is that they are unable to provide accurate results with great precision for the applications of text extraction. The proposed Text Extraction System would intelligently read the text regions from various complex images. The design includes various stages like localization, segmentation and fi nally recognition of the textual data in images. For the localization of text, Discrete Wavelength Transform function is used. Then the morphological operations are applied to correctly mark the text regions. After that, the text portion is segmented and recognized by an effi cient system. A big advantage of this system is that the output which is a text data can be stored in a .txt fi le format. Furthermore, modifi cation of the extracted text is also possible. This proposed approach can be used in more advanced and sophisticated applications as it has exhibited better precision rate, effi ciency and recall rate.

Computer aided diagnosis (CAD) systems that guide healthcare professionals to making the correct ... more Computer aided diagnosis (CAD) systems that guide healthcare professionals to making the correct diagnosis are slowly becoming more prevalent throughout the medical field. Bone fractures are a relatively common occurrence. In most developed countries the number of fractures associated with age-related bone loss is increasing rapidly. Regardless of the treating physician’s level of experience, accurate detection and evaluation of femur fractures is often problematic. In this paper a system is proposed which involves image denoising, enhancement and segmentation process for extracting the bone from X - rays. Using the extracted bone statistical pattern recognition and classification techniques are used to identify the diaphyseal femur fracture. The performance of this system is assessed in 100 real patient data with both normal and abnormal conditions. The experimental results reveal that the proposed method can be used as an effective tool for detection of diaphyseal femur fracture automatically.

Ad hoc network is a network which is dynamic in nature where the mobile nodes form a temporary ne... more Ad hoc network is a network which is dynamic in nature where the mobile nodes form a temporary network in the absence of centralized administration. Due to the absence of centralized administrator in network, routing in mobile ad hoc network (MANET) becomes the fundamental issue which minimizes the selection of an optimal path for routing. Certain performance parameters such as latency, overhead, and packet delivery ratio (PDR) are affected adversely for which numerous techniques are advocated that enhances the selection of efficient and stable path. In the present paper, an attempt is made to select the optimal path and compare the results by varying the number of nodes by using knowledge-based learning algorithm. The optimal path will possess the highest average sum of relay nodes and will be considered as the most optimal and reliable path. We also proposed that analysis of throughput and PDR is better as compared to the traditional methods. The simulation is carried out at NS-2 network simulator, which is employed to implement wired and wireless ad hoc simulation. Keywords PDR (packet delivery ratio) Á Relay number Á Throughput AODV Á MANET

In recent days, computer-aided fracture detection system plays a role in aiding both orthopaedici... more In recent days, computer-aided fracture detection system plays a role in aiding both orthopaedician and a radiologist by providing accurate and fast results. In order to detect the fracture automatically, classification of X-ray images should be automated and it becomes the initial step. Therefore, an attempt has been made and a system is presented in this paper, which involves five image processing steps namely, denoising using high boost filter, enhancement using adaptive histogram equalization, statistical feature extraction, and classification using artificial neural network. To classify the given input X-ray images into the categories head, neck, skull, foot, palm, and spine, the probabilistic neural network, backpropagation neural network, and support vector machine classifiers are employed in classifying X-ray images. The results ascertain an overall accuracy of 92.3% in classifying X-ray images and the presented system can be used as an effective tool for X-ray image classification.

In the last decades,-the advancements in computer aided diagnosis (CAD systems, enables the medi... more In the last decades,-the advancements in computer aided diagnosis (CAD systems, enables
the medical practitioners in delivering timely treatments by interpreting the medical images
in short duration. Analyzing X-ray images is one of the pivotal task of CADe systems. This
paper presents, two new methods to effectively detect and locate the fracture in digital X-ray
images. The two methods include:(i) Hough transform based fracture detection (HTBFD) an
unsupervised learning approach where, fuzzy c-means thresholding, and edge detection
methods are used to obtain the bone boundaries. Finally, The hough transform is utilized to
detect the fracture.(ii) Gradient feature based fracture detection (GFBFD), a supervised
learning approach where, Gradient features are extracted by sub-window search. Based on
the region of extraction, features are labelled as a fracture/non fracture. Finally, fractures …

Valvular regurgitation (VR) is considered to be the mainreason behindmorbidity and mortality amon... more Valvular regurgitation (VR) is considered to be the mainreason behindmorbidity and mortality among cardiac patients. Although physical examination is enough for a clinician to find out the
presenceof regurgitation, diagnostic methods are necessary to assessthe seriousness of VR and the changes in cardiac chambers as a resultof the volume overload. Recently, echocardiography with Doppler proved to be the most useful to have the noninvasive recognition and assessment of severity besides etiology of the regurgitation of the valves. The measurements of the regurgitation help in assessingthe progressof the disease, which is criticalin determiningthe correcttime for surgical treatment or any particulartreatment. Doppler echocardiography plays the vital role in giving valuable information on the severity of VR. Today, in clinical cardiology a very high quantification
precision is needed for medical application, which is provided by the color Doppler echocardiographic images. This articlereviews several comprehensive methods that are presented in the literature to assessand quantify mitral regurgitation and aortic regurgitation through twodimensional (2D) color Doppler echocardiographic images, which is the resultof proximal flow convergence method.

In recent days, computer-aided fracture detection system plays a role in aiding both orthopaedici... more In recent days, computer-aided fracture detection system plays a role in
aiding both orthopaedician and a radiologist by providing accurate and fast results. In order to detect the fracture automatically, classification of X-ray images should be automated and it becomes the initial step. Therefore, an attempt has been made and a system is presented in this paper, which involves five image processing steps namely, denoising using high boost filter, enhancement using adaptive histogram
equalization, statistical feature extraction, and classification using artificial neural network. To classify the given input X-ray images into the categories head, neck, skull, foot, palm, and spine, the probabilistic neural network, backpropagation neural network, and support vector machine classifiers are employed in classifying X-ray images. The results ascertain an overall accuracy of 92.3% in classifying X-ray images and the presented system can be used as an effective tool for X-ray image classification.

With the extensive use of credit cards, fraud appears as a major issue in the credit card busines... more With the extensive use of credit cards, fraud appears as a major issue in the credit card business. It is hard to have some figures on the impact of fraud, since companies and banks do not like to disclose the amount of losses due to frauds. At the same time, public data are scarcely available for confidentiality issues, leaving unanswered many questions about what is the best strategy. Another problem in credit-card fraud loss estimation is that we can measure the loss of only those frauds that have been detected, and it is not possible to assess the size of unreported/undetected frauds. Fraud patterns are changing rapidly where fraud detection needs to be re-evaluated from a reactive to a proactive approach. In recent years, machine learning has gained lot of popularity in image analysis, natural language processing and speech recognition. In this regard, implementation of efficient fraud detection algorithms using machine-learning techniques is key for reducing these losses, and to assist fraud investigators. In this paper logistic regression, based machine learning approach is utilized to detect credit card fraud. The results show logistic regression based approaches outperforms with the highest accuracy and it can be effectively used for fraud investigators.

Major heart diseases like heart muscle damage and valvular problems are diagnosed using echocardi... more Major heart diseases like heart muscle damage and valvular problems are diagnosed using echocardio-
gram. Since the echocardiogram is an image or sequence of images with less information the cardiologist
spends more time to predict or to make decision. Automating the detection and diagnosis of dilated car-
diomyopathy (DCM) and hypertrophic cardiomyopathy (HCM) is a key enabling technology in computer
aided diagnosis systems. In this paper, a system is proposed to automatically detect and diagnose dilated
cardiomyopathy (DCM) and hypertrophic cardiomyopathy (HCM). This system performs denoising,
enhancement, before left ventricular segmentation is carried out in the individual frames. Using the seg-
mented left ventricle, the LV parameters like volume and ejection fraction (EF) are calculated and also the
end-diastolic LV is extracted. The PCA and DCT features are obtained from the extracted end-diastolic LV
and the classifiers BPNN, SVM and combined K-NN are used to classify the normal hearts, hearts affected
with DCM and hearts affected with HCM. The PCA feature with BPNN classifier gives a highest overall
accuracy of 92.04% in classifying normal and abnormal hearts. Experiments over 60 echocardiogram
videos expose that the proposed system can be effectively utilized to detect and diagnose DCM and HCM.

Automating the X-ray view identification is the first step in automating the detection and diagno... more Automating the X-ray view identification is the
first step in automating the detection and diagnosis of
fractures in bones. In this paper, an attempt has been made
to classify the anterio-posterior (AP) and lateral (LAT) views
of leg X-rays. Two methods namely model based and
template based is proposed to classify the AP and LAT views.
In the model based method the X-rays are preprocessed,
and then the histogram and statistical features are
extracted. The support vector machine and probabilistic
neural network were employed to classify the views. In the
template based method the speed up robust features (SURF)
is used for classification. SURF is effective in collecting more
class-specific information and robust in dealing with partial
occlusion and viewpoint changes. To authenticate the
generalizability and robustness, the proposed methods are
tested on a dataset of 50 X-ray images, and among the two,
SURF achieves a higher classification rate of 91.8%.

Automating the detection and diagnosis of cardiovascular diseases using echocardiogram sequences ... more Automating the detection and diagnosis of cardiovascular diseases using echocardiogram sequences is a challenging task because of the presence of speckle noise, less information and movement of chambers. In this paper an attempt has been made to classify the normal hearts, and hearts affected by dilated cardiomyopathy (DCM) and hypertrophic cardiomyopathy (HCM) by automating the segmentation of left ventricle (LV). The segmented LV from the diastolic frames of echocardiogram sequences alone is used for extracting features. The statistical features and Zernike moment features are obtained from extracted diastolic LV and classified using the classifiers namely support vector machine (SVM), back propagation neural network (BPNN) and probabilistic neural network (PNN). An intensive examination over 60 echocardiogram sequences reveals that the proposed method performs well in classifying normal hearts and hearts affected by DCM and HCM. Among the classifiers used the BPNN classifier with the combination of Zernike moment features gave an highest accuracy of 92.08 %.

Magnetic Resonance imaging is one of the widely used brain imaging modality to visualize and asse... more Magnetic Resonance imaging is one of the widely used
brain imaging modality to visualize and assess the brain
anatomy and its function. It provides a much greater
contrast between the different soft tissues of the body than
computed tomography (CT) does. In this paper an
approach for segmentation of brain tissues from T1 –
MRI images is proposed. To exemplify the approach, a
system is presented which involves image denoising,
contrast enhancement and segmentation of brain tissues
namely gray matter (GM), white matter (WM) and
cerebrospinal fluid (CSF). The segmentation is carried
out using combination of Expectation Maximization (EM)
and Gaussian Mixture Model (GMM). From the
experimental results it is observed that the brain tissue is
segmented accurately using EM-GMM.

Automatic classification cardiac views is the first step to automate wall motion analysis, comput... more Automatic classification cardiac views is the first step to automate wall motion analysis, computer aided disease diagnosis, measurement computation etc. In this paper a fully automatic classification of cardiac view in echocardiogram is proposed. The system is built based on a machine learning approach which characterizes two features 1) Histogram features and 2) Statistical features. In this system four standard views parasternal short axis (PSAX), parasternal long axis (PLAX), apical two chamber (A2C) and apical four chamber (A4C) views are classified. Experiments over 200 echocardiogram images show that the proposed method with an accuracy of 87.5% can be effectively used in cardiac view classification.

Automating cardiac view classification is the first step for automating computer aided cardiac di... more Automating cardiac view classification is the first step for automating computer aided cardiac disease
diagnosis. In this paper automatic cardiac view classification system is proposed. Methods: This system attempts to
classify four standard cardiac views in echocardiogram namely Parasternal Long Axis (PLAX), Parasternal Short Axis
(PSAX), Apical Four Chamber (A4C), and Apical Two Chamber (A2C) views automatically using Speed Up Robust Features
(SURF). Conclusion: The Speed Up Robust Features is effective in collecting more class-specific information and ro-bust
in dealing with partial occlusion and viewpoint changes. To authenticate the generalizability and robustness, the proposed
system is tested on a dataset of 200 echocardiogram images which achieve a classification rate of 90.7%.

In this work, an approach for heart muscle damage detection from echocardiography sequences is pr... more In this work, an approach for heart muscle damage detection from echocardiography sequences is proposed. To exemplify
the approach, a system is presented which involves image denoising and enhancement and segmentation of the
left ventricle (LV) for extracting the heart wall boundaries. Using the heart wall boundaries global LV parameters are
calculated followed by statistical pattern recognition and classification to identify the heart muscle damage or myocardial
ischemia (MI). The performance of this algorithm is assessed in 60 real patient data with both normal and abnormal
conditions. The experimental results reveal that the proposed method can be used as an effective tool for
detection of heart muscle damage or MI automatically.

In this paper automatic detection of mitral regurgitation jet in color Doppler echocardiogram vid... more In this paper automatic detection of mitral regurgitation jet in color Doppler echocardiogram videos is proposed. Here the main objective is to segment the jet area during Mitral Regurgitation and track it for the convenience of cardiologist. The severity of Mitral Regurgitation (MR) is directly related to the jet area. The measurements of the regurgitation help in evaluating the exact advancement of the disease which is crucial in deciding the opportune time for surgical treatment or any specific treatment. The K-Means clustering is used to segment the jet area.

In this paper a novel and robust automatic LV segmentation by measuring the properties of each co... more In this paper a novel and robust automatic LV segmentation by measuring the properties of each connected components in the echocardiogram images and a cardiac abnormality detection method based on ejection fraction is proposed. Starting from echocardiogram videos of normal and abnormal hearts, the left ventricle is first segmented using connected component labeling and from the segmented LV region the proposed algorithm is used to calculate the left ventricle diameter. The diameter derived is used to calculate the various LV parameters. In each heart beat or cardiac cycle, the volumetric fraction of blood pumped out of the left ventricle (LV) and the ejection fraction (EF) were calculated based on which the cardiac abnormality is decided. The proposed method gave an accuracy of 93.3% and it can be used as an effective tool to segment left ventricle boundary and for classifying the heart as either normal or abnormal.

In this paper an efficient cardiac view classification of echocardiogram is proposed. A cardiac c... more In this paper an efficient cardiac view classification of echocardiogram is proposed. A cardiac cycle
consists of two phases systolic and diastolic. The systolic is the contraction and diastolic is relaxation and
filling. From the given video sequences only the diastolic frames are extracted and it is utilized for
determining the view of the echocardiogram. The Echocardiogram image are first prepared to reduce noise
and to enhance the contrast of the image then mathematical morphology is used to highlight the cardiac
cavity before segmentation using Connected Components Labeling (CCL) is carried out. We classify three
standard cardiac views namely parasternal short axis (PSAX), apical two chamber (A2C) and four chamber
(A4C) views. Experiments over 200 echocardiogram images show that the proposed method ascertains
94.56% of accuracy in cardiac view classification.

Echocardiogram is one of the easiest ad widely employed methods that uses ultrasound to evaluate ... more Echocardiogram is one of the easiest ad widely employed methods that uses ultrasound to evaluate heart muscle, heart valves, and risk for heart disease. Heart failure (HF) can result from any structural or functional cardiac disorder that impairs the ability of the ventricle to fill with or eject blood. Echocardiography represents "the gold standard" in the assessment of left ventricle LV systolic and diastolic dysfunction. Left ventricular dimensions, volumes and wall thicknesses are echocardiographic measurements that are widely used in clinical practice and research. To obtain accurate linear measurements from the echocardiography accurate segmentation of the LV is essential. This paper proposes a method to segment the left ventricular border automatically on the 3-dimensional (2D+t) echocardiogram, where 't' is the time. The 2D image is obtained by extracting the frames from the video of echocardiogram which is further processed to detect the edges of the left ventricle and finally the edge detected frames are converted back into video which will help the cardiologist to visualize the left ventricle in motion. The obtained results are efficient and can be utilized for the purpose of detecting various medical parameters.

In recent days, computer-aided fracture detection system plays a role in aiding both orthopaedici... more In recent days, computer-aided fracture detection system plays a role in
aiding both orthopaedician and a radiologist by providing accurate and fast results. In order to detect the fracture automatically, classification of X-ray images should be automated and it becomes the initial step. Therefore, an attempt has been made and a system is presented in this paper, which involves five image processing steps namely, denoising using high boost filter, enhancement using adaptive histogram
equalization, statistical feature extraction, and classification using artificial neural network. To classify the given input X-ray images into the categories head, neck, skull, foot, palm, and spine, the probabilistic neural network, backpropagation neural network, and support vector machine classifiers are employed in classifying X-ray images. The results ascertain an overall accuracy of 92.3% in classifying X-ray images and the presented system can be used as an effective tool for X-ray image classification.