Sanika S. Patankar

Optimized MFR & automated local entropy thresholding for retinal blood vessel extraction

Retinal blood vessel extraction plays important role in diagnosis of many diseases such as diabetic retinopathy (DR), hypertension, glaucoma and arteriosclerosis. In this paper optimized matched filter response is used to enhance the blood vessel followed by local entropy thresholding used to segment the vessels automatically. First optimized matched filter are applied to the retinal images to enhance vessels then we used their corresponding co-occurrence matrix & automatic to find local entropy thresholding that used for segmented the blood vessel in retinal image. The results shows that automated local entropy thresholding more successful compare to other methods in our proposed matched have 95.86 % average accuracy.

Features based classification of hard exudates in retinal images

Diabetes mellitus is a major disease spread all across the globe. Long-time diabetes mellitus causes the complication in the retina called Diabetic Retinopathy (DR), which results in visual loss and sometimes blindness. In this paper, we discuss a simple and effective algorithm for segmentation of the optic disk (OD) and bright lesions such as hard exudates from color retinal images. Color fundus images are enhanced using brightness transform function. Morphological operator along with the Circular Hough Transform (CHT) is used for optic disk segmentation. Further, local mean and entropy based region growing technique is applied in order to classify exudate - non-exudate pixels in retinal images. The performance of the proposed algorithm has been tested on publicly available standard Messidor database images with varied disease levels and non-uniform illumination. Experimentation yields 94% success rate for localization of the optic disk, 99% accuracy of classification of exudate - non-exudate pixels and subject level accuracy is found to be 93% and 67% in identifying the abnormal (with exudates) and normal (without exudates) images respectively.

Gear tooth fault detection by autoregressive modelling

Gears are important element in a variety of industrial applications. An unexpected failure of the gear may cause significant economic losses. For that reason, fault diagnosis in gears has been the subject of intensive research. Vibration signal analysis has been widely used in the fault detection of rotation machinery. This paper present a gear tooth fault diagnosis technique of Autoregressive (AR) modeling of vibration signals. AR model coefficient is been determined by Yule-Walker equation with Levision-Durbin recursive algorithm. The model order is an essential part and is calculated by Akaike Information Criteria. The vibration signal of normal and faulty gear is been modeled and frequency response of AR model of the faulty gear is been compared with the AR model of the normal gear. The changes in the frequency spectrum indicate the fault.

Condition monitoring of rotary machinery using Continuous Wavelets

Condition monitoring has significant importance in manufacturing industry. Avoiding production loss and minimizing the probability of occurance of calamitous machine failure, based on updated information acquired from machine status on-line is the aim of condition monitoring. This paper discusses various vibration signal analysis techniques. The experimentation has been carried out using a mechanical setup consisting of rotary machine. The setup has a provision of introducing fault (uncertainty) by way of using gear with broken tooth. The effect of uncertainty (introduced in the vibration signal because of gear with broken tooth) is analyzed using Fourier Transform and Continuous Wavelet Transform (with Daubechies having three vanishing moments and Mexican Hat basis functions). From the experimental results, it is observed that the uncertainty due to broken tooth has been significantly detected by Continuous Wavelet Transform using Mexican Hat basis function as compared to Fourier Transform.

Intensity features based classification of hard exudates in retinal images

A major cause of blindness is diabetic retinopathy, which is found in the people who suffer from diabetes, which can be detected through a screening process. Hard exudates are one of the signs of diabetic retinopathy, which caused due to breakdown of retinal blood vessels. This paper presents a method for classification of hard exudates using 6-Dimensional intensity based features. The exudates and non-exudates (background) classification is performed using the Euclidean distance classifier. The proposed method is tested against publicly available databases such as DIARETDB1, e-ophtha EX, MESSIDOR. The proposed algorithm demonstrates maximum subject level accuracy of 96.92% on DIARETDB1.

Orthogonal moments for determining correspondence between vessel bifurcations for retinal image registration

Retinal image registration is a necessary step in diagnosis and monitoring of Diabetes Retinopathy (DR), which is one of the leading causes of blindness. Long term diabetes affects the retinal blood vessels and capillaries eventually causing blindness. This progressive damage to retina and subsequent blindness can be prevented by periodic retinal screening. The extent of damage caused by DR can be assessed by comparing retinal images captured during periodic retinal screenings. During image acquisition at the time of periodic screenings translation, rotation and scale (TRS) are introduced in the retinal images. Therefore retinal image registration is an essential step in automated system for screening, diagnosis, treatment and evaluation of DR. This paper presents an algorithm for registration of retinal images using orthogonal moment invariants as features for determining the correspondence between the dominant points (vessel bifurcations) in the reference and test retinal images. As orthogonal moments are invariant to TRS; moment invariants features around a vessel bifurcation are unaltered due to TRS and can be used to determine the correspondence between reference and test retinal images. The vessel bifurcation points are located in segmented, thinned (mono pixel vessel width) retinal images and labeled in corresponding grayscale retinal images. The correspondence between vessel bifurcations in reference and test retinal image is established based on moment invariants features. Further the TRS in test retinal image with respect to reference retinal image is estimated using similarity transformation. The test retinal image is aligned with reference retinal image using the estimated registration parameters. The accuracy of registration is evaluated in terms of mean error and standard deviation of the labeled vessel bifurcation points in the aligned images. The experimentation is carried out on DRIVE database, STARE database, VARIA database and database provided by local government hospital in Pune, India. The experimental results exhibit effectiveness of the proposed algorithm for registration of retinal images.

Gradient features and optimal thresholding for retinal blood vessel segmentation

Retinal blood vessel segmentation is a fundamental step in diagnosis, screening, treatment and evaluation of diabetes retinopathy. Manual segmentation of retinal blood vessels is a long and tedious task and requires trained graders, thus automatic segmentation of retinal blood vessels is a fundamental step in development of computer based diagnostic system for diabetic retinopathy. This paper presents a method for segmentation of retinal blood vessels based on gradient between vessel pixels and background pixels. Due to the intensity variation between retinal blood vessels and background, gradient features of vessel and non vessel pixels can be used for segmentation. Green component of the input retinal image is extracted and first order gradient features are computed using 3 × 3 gradient kernel. The magnitude of the gradient is observed to be maximum at the blood vessels due to intensity variations between vessel and non vessel pixels. Optimal thresholding is then performed on gradient features and retinal blood vessels are segmented. Median filtering is used to reduce salt and pepper noise and length filtering is used to remove isolated pixels. The algorithm is tested on publicly available DRIVE database. The overall accuracy of 92.04 %, sensitivity of 82.44 % and specificity of 95.10 % is observed.

Modified matched filter kernel for classification of hard exudate

Diabetic Retinopathy is an abnormality of eye in which the retina of patient is affected due to an increasing amount of insulin in blood. The symptoms can distort or blur the patients vision and thus lead blindness. For automatic detection of exudates we first have to differentiate intensity levels of exudate and non-exudate pixels. Matched filter is used for same. Tuning of matched filter is an important criteria which is presented in this paper. This paper contains how to tune and modify matched filter response for easily segmentation of Hard Exudates. It also contains graphical experimented results for different values of sigma and how accuracy of the algorithm varies with it. Experimentation gives 99.62% accuracy of classification of exudate — non-exudate pixels and subject level accuracy is found to be 93.75% in identifying the abnormal (with exudates) and normal (without exudates) images respectively.

Directional gradient based registration of retinal images

Registration of retinal images provided by different modalities is required to facilitate diagnosis of the retina specifically in context with Diabetes Retinopathy (DR). Temporal registration is necessary in order to follow the various stages DR, whereas multimodal registration allows us to improve the identification of some lesions or to compare pieces of information gathered from different sources. This paper presents an algorithm for temporal and/or multimodal registration of retinal images based on directional gradient of salient vessel bifurcation points. Gradient along 0, +45 and 90 degrees around bifurcation point is invariant to rotation. We have validated this by using Fourier transform. This enables to estimate the correspondence of bifurcation points between the image pairs to be aligned. The retinal vessel tree is first segmented and vessel bifurcation points are located. A feature vector comprising of directional gradients along 0, +45 and 90 degrees around labeled bifurcation point in its 3 × 3 neighborhood is then computed in order to determine correspondence. The rotation is estimated from the coordinates of the matched vessel bifurcation points. The algorithm is tested on publically available DRIVE and STARE database. The overall registration error is observed to be 0.8% and 0.7% for STARE and DRIVE database respectively.

Determination of deformation of composite material using harmonic analysis

Composite materials have beneficial uses in Aerospace, Structural engineering particularly in weight reduction in the finished part. For that reason, harmonic analysis has been the topic of intensive research. This paper aims to carry out an algorithm for natural frequency and deformation of composite plate by vibration signal analysis. The vibration signal is measured using piezoelectric type accelerometer. The ARX model and FFT of the signal is simulated using MATLAB. Natural frequency is used to calculate the deformation of the plate. The results are validated with finite element method.