Ramakrishnan Swaminathan

Analysis of needle electromyography signal is used for the differentiation of neuropathy and myopathy condition from the normal. Amplitude based features such as root mean square and mean absolute value are used to differentiate between normal and pathological signals. Tunable-Q wavelet transform is used to decompose the frequency bands of the signal. Further, the same set of features are used to analyse each frequency bands. The results show that the proposed approach is able to distinguish between normal and different pathological electromyography signals better than the conventional time domain analysis. It is also observed that myopathy and neuropathy signals are comprised of high frequency components than low frequency components as compared to normal signal. The proposed method yields a higher significance with a p-value <0.05 between normal and each pathological signal such as neuropathy and myopathy.

Analysis of blood vessel mechanics in normal and diseased conditions is essential for disease research, medical device design and treatment planning. In this work, 2D finite element models of normal vessel and atherosclerotic vessels with 50% and 90% plaque deposition were developed and were meshed using Delaunay triangulation method. The transient analysis was performed and the parameters such as total displacement, Von Mises stress and strain energy density were analyzed for normal and atherosclerotic vessels. Results demonstrate that an inverse relation exists between the considered mechanical parameters over the vessel surface and the percentage of plaque deposited on the inner vessel wall. It was further observed that the total displacement and Von Mises stress decrease nonlinearly with increasing plaque percentage. Whereas, the strain energy density decreases almost linearly with increase in plaque deposition. In this paper, the objectives of the study, methodology and significant observations are presented. © 2011 Springer-Verlag.

In this work, sub anatomic regions of brain are studied using diffusion tensor images. Images of normal controls (NC) and Alzheimer disease (AD) subjects were obtained from ADNI database. Volume of four regions (Thalamus, Posterior cingulate, Temporal lobe and Hypothalamus) were estimated by implementing voxel analysis and atlas based approach. All the regions showed a reduction in volume. The variation in volume was found to be significant in Temporal lobe. A classification approach using Naive Bayes algorithm was used for estimated volume feature. A accuracy of 85% was obtained with sensitivity and specificity of 85% and 80% respectively.

Alzheimer's disease (AD) is a leading cause of dementia in elderly adults. In this, the white matter (WM) tracts in brain are disintegrated leading to loss of important cognitive functionality. Recent analysis have shown that early diagnosis of AD is still a challenging task. Although several reports are available, tractography remains the most promising and clinically relevant method for in-vivo study of WM tracts. In tractography, continuous WM pathways are reconstructed from voxel based models of discrete fiber orientation generated using diffusion tensor images. In this work an attempt has been made to classify AD using average length of tracts, a significant feature extracted from tractographic brain maps. The diffusion weighted images for AD and matched controls were obtained from ADNI, an international open access repository for Alzheimer's study. Data from equal number of AD and controls were used for this study. Fiber tracking was performed for the whole brain using tract based spatial statistics algorithm. ICBM Mori Labels 1 atlas provided in the Network Analysis option of ExploreDTI was used to divide the WM into 48 anatomical regions. Classification was performed using random forest, random tree and decision stumps, and their performance indices were compared. The results show that all the classifiers are able to classify AD and controls using the extracted feature. An accuracy of 78.4% is obtained using decision stumps. Random forest and random tree provide an increased accuracy of 96% and 97% respectively. The precision and recall is also found to be higher for random forest and random tree as compared to decision stumps. These results suggest that random forest and random tree are suitable for classification of AD and controls using average tract length as a feature. In this paper, the introduction, objectives, materials and methods, results and discussions and conclusions are presented in detail. Copyright 2014, ISA All Rights Reserved.

Low signal to noise ratio and low contrast are the major limitations for segmentation, image analysis and interpretation of medical thermal images. In this work, an attempt has been made to improve and preserve the inter-regions edges by effectively removing the noise without blurring and hence, to extract the breast tissues from infrared images using level sets based on improved edge information. Gaussian filter is a linear, homogenous diffusion process that performs smoothing operation at each location that blurs the edge information resulting in difficulty of detection and localization of edges. To avoid smoothing across boundaries, an anisotropic diffusion based smoothing filter is used. This enables smoothing within the region by preserving sharp region boundaries. The performance improvement of the diffusion filter is verified and validated by extracting the breast tissues. The segmentation of regions of interests (ROIs) is performed by evolving the initial level set function based on this improved edge information. The extracted ROIs are compared with the corresponding four sets of ground truth images. The results show a good agreement of the segmented ROIs with ground truths. Further, the performance of the segmentation method is analyzed across inter person variations by calculating quantitative measures based on overlap and the statistics of regional similarities. It is observed that the segmentation method could able delineate the accurate regions of interest irrespective to the limitations of thermal images such as lack of clear edges. Average accuracy of 98% of regional similarity is obtained between segmented ROIs and ground truth images. Therefore, the enhanced edge detail seems to be useful to improve the performance of segmentation algorithm which could be used during breast cancer screening for early detection of tumor. © 2014 Elsevier Ltd.

In this study, an attempt has been made to differentiate Novel Coronavirus-2019 (COVID-19) conditions from healthy subjects in Chest radiographs using a simplified end-to-end Convolutional Neural Network (CNN) model and occlusion sensitivity maps. Early detection and faster automated screening of the COVID-19 patients is essential. For this, the images are considered from publicly available datasets. Significant biomarkers representing critical image features are extracted from CNN by experimentally investigating on cross-validation methods and hyperparameter settings. The performance of the network is evaluated using standard metrics. Perturbation based occlusion sensitivity maps are employed on the features obtained from the classification model to visualise the localization of abnormal areas. Results demonstrate that the simplified CNN model with optimised parameters is able to extract significant features with a sensitivity of 97.35% and F-measure of 96.71% to detect COVID-19 images. The algorithm achieves an Area Under the Curve-Receiver Operating Characteristic score of 99.4% with Matthews correlation coefficient of 0.93. High value of Diagnostic odds ratio is also obtained. Occlusion sensitivity maps provide precise localization of abnormal regions by identifying COVID-19 conditions. As early detection through chest radiographic images are useful for automated screening of the disease, this method appears to be clinically relevant in providing a visual diagnostic solution using a simplified and efficient model.

In this work, isometric contraction of triceps brachii muscles are analyzed using surface electromyography (sEMG) signals and multifractals. The signals are recorded from 20 healthy adult volunteers at three different angles of elbow flexions namely 30°, 60° and 90° using standard experimental protocol. The recorded signals are preprocessed and subjected to Multifractal Detrended Fluctuation Analysis (MF-DFA). The three extracted MF-DFA features namely, peak exponent event, rare exponent event and smooth exponent event, are used for further analysis. The results show that multifractal spectrum response are distinct for all the three angles of flexion. It is observed that multifractal spectrum is shifting from higher order exponents to lower order with the increase in the angle of flexion. The features are statistically highly significant between 30° and 90° flexion. The smooth exponent is found to be statistically highly significant for both 60°-90° and 30°-90° flexion. It appears that this method of multifractal technique is an useful approach to understand muscle contractions in fatigue and in varied clinical conditions. © 2014 MIPRO.

In this work, an attempt has been made to analyze surface electromyography signals (sEMG) by estimating maximum singular value. sEMG signals are recorded from biceps brachii muscles of 50 healthy volunteers during repetitive elbow flexion and extension exercise. Maximum singular values are estimated from the signals. The results show a decrease in MSV at the point of first muscle discomfort experienced by subjects. For most of the subjects, the point of first discomfort occur in fourth and fifth regions of the time axis. It appears that this method can be used to analyze progress of muscle condition towards fatigue.

In this work, an automated analysis of trabecular architecture is carried out using empirical mode decomposition and Support Vector Machines (SVM). The trabecular regions of normal and abnormal human femur bone images (N=40) recorded under standard condition are used for this study. The compressive region in the images are delineated and decomposed into their corresponding intrinsic mode functions using Bi-dimensional Empirical Mode Decomposition (BEMD) method. The characteristic feature vectors are extracted and further subjected to classification using SVM. Results show that BEMD analysis combined with SVM could differentiate normal and abnormal images. As the strength of the bone depends on architectural variation in addition to bone mass, this study seems to be clinically useful. © 2012 IEEE.

Background and Objective: Computer aided diagnostics of Pulmonary Tuberculosis in chest radiographs relies on the differentiation of subtle and non-specific alterations in the images. In this study, an attempt has been made to identify and classify Tuberculosis conditions from healthy subjects in chest radiographs using integrated local feature descriptors and variants of extreme learning machine. Methods: Lung fields in the chest images are segmented using Reaction Diffusion Level Set method. Local feature descriptors such as Median Robust Extended Local Binary Patterns and Gradient Local Ternary Patterns are extracted. Extreme Learning Machine (ELM) and Online Sequential ELM (OSELM) classifiers are employed to identify Tuberculosis conditions and, their performances are analysed using standard metrics. Results: Results show that the adopted segmentation method is able to delineate lung fields in both healthy and Tuberculosis images. Extracted features are statistically significant even in images with inter and intra subject variability. Sigmoid activation function yields accuracy and sensitivity values greater than 98% for both the classifiers. Highest sensitivity is observed with OSELM for minimal significant features in detecting Tuberculosis images. Conclusion: As ELM based method is able to differentiate the subtle changes in inter and intra subject variations of chest X-ray images, the proposed methodology seems to be useful for computer-based detection of Pulmonary Tuberculosis.