Ayush Goyal

Model-Based Vasculature Extraction From Optical Fluorescence Cryomicrotome Images

The aim of this study was to develop a novel method to reconstruct 3-D coronary vasculature from cryomicrotome images, comprised of two distinct sets of data-fluorescent microsphere beads and coronary vasculature. Fluorescent beads and cast injected into the vasculature were separately imaged with different filter settings to obtain the microsphere and vascular data, respectively. To extract the vascular anatomy, light scattering in the tissue was modelled using a point spread function (PSF). The PSF was parametrized by optical tissue excitation and emission attenuation coefficients, which were estimated by fitting simulated images of microspheres convolved with the PSF model to the experimental microsphere images. These parameters were then applied within a new model-based method for vessel radius estimation. Current state-of-the-art radii estimation methods and the proposed model-based method were applied on vessel phantoms. In this validation study, the full-width half-maximum method of radii estimation, when performed on the raw data without correcting for the optical blurring, resulted in 42.9% error on average for the 170 μm vessel. In comparison, the model-based method resulted in 0.6% error on average for the same phantom. Whole-organ porcine coronary vasculature was automatically reconstructed with the new model-based vascular extraction method.

Multi-Scale Parameterisation of a Myocardial Perfusion Model Using Whole-Organ Arterial Networks

A method to extract myocardial coronary permeabilities appropriate to parameterise a continuum porous perfusion model using the underlying anatomical vascular network is developed. Canine and porcine whole-heart discrete arterial models were extracted from high-resolution cryomicrotome vessel image stacks. Five parameterisation methods were considered that are primarily distinguished by the level of anatomical data used in the definition of the permeability and pressure-coupling fields. Continuum multi-compartment porous perfusion model pressure results derived using these parameterisation methods were compared quantitatively via a root-mean-square metric to the Poiseuille pressure solved on the discrete arterial vasculature. The use of anatomical detail to parameterise the porous medium significantly improved the continuum pressure results. The majority of this improvement was attributed to the use of anatomically-derived pressure-coupling fields. It was found that the best results were most reliably obtained by using porosity-scaled isotropic permeabilities and anatomically-derived pressure-coupling fields. This paper presents the first continuum perfusion model where all parameters were derived from the underlying anatomical vascular network.

Automatic Pattern Recognition for Detection of Disease from Blood Drop Stain Obtained with Microfluidic Device

This paper investigates automatic detection of disease from the pattern of dried micro-drop blood stains from a patient’s blood sample. This method has the advantage of being substantially cost-effective, painless and less-invasive, quite effective for disease detection in newborns and the aged. Disease has an effect on the physical properties of blood, which in turn affect the deposition patterns of dried blood micro-droplets. For example, low platelet count will result in thinning of blood, i.e. a change in viscosity, one of the physical properties of blood. Hence, the blood micro-drop stain patterns can be used for diagnosing diseases. This paper presents automatic analysis of the dried micro-drop blood stain patterns using computer vision and pattern recognition algorithms. The patterns of micro-drop blood stains of normal non-diseased individuals are clearly distinguishable from the patterns of micro-drop blood stains of diseased individuals. As a case study, the micro-drop blood stains of patients infected with tuberculosis have been compared to the micro-drop blood stains of normal non-diseased individuals. The paper delves into the underlying physics behind how the deposition pattern of the dried micro-drop blood stain is formed. What has been observed is a thick ring like pattern in the dried micro-drop blood stains of non-diseased individuals and thin contour like lines in the dried micro-drop blood stains of patients with tuberculosis infection. The ring like pattern is due to capillary flow, an outward flow carrying suspended particles to the edge. Concentric rings (caused by inward Marangoni flow) and central deposits are some of the other patterns that were observed in the dried micro-drop blood stain patterns of normal non-diseased individuals.

Intramural spatial variation of optical tissue properties measured with fluorescence microsphere images of porcine cardiac tissue

This proceeding studies the optical fluorescence images of a porcine heart filled with microspheres of two colors, carmine and red. A significant difference in the total optical tissue attenuation coefficient was observed between excitation and emission for both carmine (excitation – 13±4(1/mm) and emission – 9.4±3(1/mm)) and red (excitation -29±5(1/mm) and emission – 25±5(1/mm)), indicating that optical tissue properties can change significantly for a small change in light wavelength. The above-mentioned large ranges of variation observed in the tissue attenuation coefficient for excitation and emission (both for carmine and red) suggest significant intramural variation of optical properties across the entire organ. Patterns of global spatial variation in optical attenuation properties in tissue across the entire organ were observed. A novel method using fluorescence microsphere images is presented for measurement of the tissue attenuation’s intramural variation across an entire organ.

Abandoned Object Detection and Tracking Using CCTV Camera

With the increase in crime and terror rate globally, automated video surveillance, is the need of the hour. Surveillance along with the detection and tracking has become extremely important. Human detection and tracking is ideal, but the random nature of human movement makes it extremely difficult to track and classify as suspicious activities. The primary objective of this is to detect the suspiciously abandoned object recorded by the closed-circuit television cameras (CCTV). The main aim of this project is to ease the load on the controller at the main CCTV station by generating and alarm, whenever there is a detection of an abandoned object. To solve the problem, we first proceeded by the background subtraction such that we obtain the foreground image. Further, we calculated the inter-pixel distance and used area-based thresholding so as to differentiate between the person and the object. The object will further be tracked for a previously set time, which will help the system to decide whether or not the object is abandoned or not. Such a system that can ease the load on single CCTV controller can be deployed in places which require high discipline and security and are more prone to suspicious activities like Airports, Metro station, Railway Stations, entrances and exits of buildings, ATMs, and similar public places.

Image-based fuzzy c-means clustering and connected component labeling subsecond fast fully automatic complete cardiac cycle left ventricle segmentation in multi frame cardiac MRI images

A rapid method for left ventricle extraction from MRI images of cardiac patients is presented in this research. This facilitates cardiologists to critically assess the cardiac function or dysfunction in a patient in terms of their left ventricles performance, measured as its ejection fraction. Fuzzy c-means based pixel clustering is used for automatic segmentation. The left ventricle in all frames in the complete cardiac heartbeat cycle are extracted after being automatically loaded and segmented. In each image, pixels are grouped into two clusters - foreground and background. After the clustering, connected component analysis labels the pixels into connected regions. The left ventricle region is heuristically selected based on the distance from the image center and eccentricity. This novel original pixel clustering with labeling approach avoids manual initialization or user intervention. This method fully automatically extracts the left ventricle with more accuracy than manual tracing on all slices in the MRI images of the complete cardiac heartbeat cycle. The average computational processing speed per frame is 0.6 seconds, making it much more efficient than level sets, active contours, or other deformable methods, which need many iterations for the evolution of the snake or contour. Accuracy of the automated method presented herein was validated against manual tracing-based extraction. After performing the comparison on four MRI frames, it was found that an average correlation coefficient of 0.95 between the automatic and manual left ventricle segmented boundaries was higher than an average correlation coefficient of 0.85 between two manual tracing-based segmentations of the same.

Automatic disease screening method using image processing for dried blood microfluidic drop stain pattern recognition.

Abstract This paper examines programmed automatic recognition of infection from samples of dried stains of micro-scale drops of patient blood. This technique has the upside of being low-cost and less-intrusive and not requiring puncturing the patient with a needle for drawing blood, which is especially critical for infants and the matured. It also does not require expensive pathological blood test laboratory equipment. The method is shown in this work to be successful for ailment identification in patients suffering from tuberculosis and anaemia. Illness affects the physical properties of blood, which thus influence the samples of dried micro-scale blood drop stains. For instance, if a patient has a severe drop in platelet count, which is often the case of dengue or malaria patients, the blood’s physical property of viscosity drops substantially, i.e. the blood is thinner. Thus, the blood micro-scale drop stain samples can be utilised for diagnosing maladies. This paper presents programmed automatic examination of the dried micro-scale drop blood stain designs utilising an algorithm based on pattern recognition. The samples of micro-scale blood drop stains of ordinary non-infected people are clearly recognisable as well as the samples of micro-scale blood drop stains of sick people, due to key distinguishing features. As a contextual analysis, the micro-scale blood drop stains of patients infected with tuberculosis have been contrasted with the micro-scale blood drop stains of typical normal healthy people. The paper dives into the fundamental flow mechanics behind how the samples of the dried micro-scale blood drop stain is shaped. What has been found is a thick ring like feature in the dried micro-scale blood drop stains of non-ailing people and thin shape like lines in the dried micro-scale blood drop stains of patients with anaemia or tuberculosis disease. The ring like feature at the periphery is caused by an outward stream conveying suspended particles to the edge. Concentric rings (brought on by internal Marangoni flow) and deposition in the centre of the stain are patterns that were found in the dried micro-scale drop blood stain samples of ordinary healthy people.

Image-based pixel clustering and connected component labeling in left ventricle segmentation of cardiac MR images

This research demonstrates a completely automated sub-second fast technique for left ventricle (LV) segmentation from clinical cardiac MRI images for the crucial assessment of left ventricular dysfunction as a measure of cardiac diseases. In this work left ventricle segmentation is achieved using the combination of fuzzy c-means which is a pixel based classification method and connected component labeling. This strategic combination obviates user intervention and problem of seed point initialization as it automatically segments the LV accurately on all frames in the complete cardiac cycle in multi-frame MRI. The both methods complement each other such that it achieves sub-second fast computational speed of 0.7 seconds on average per frame. Thus this techniques computational time for left ventricle segmentation is much faster than iteration based methods. The accuracy of the automatic segmentation technique was tested against manual segmentation on the basis of correlation coefficient. The value of correlation coefficient between the automatic and manually traced LV boundaries was 0.932 which can be considered clinically significant.

Fast fully automatic multiframe segmentation of left ventricle in cardiac MRI images using local adaptive k-means clustering and connected component labeling

This paper presents a sub-second fast fully automatic method for segmentation of the left ventricle (LV) from cardiac MRI images, which plays a vital role in the diagnosis of left ventricular function for the assessment of cardiac disease in a patient. In this paper the segmentation of the left ventricle using local adaptive k-means clustering and connected components is achieved fully automatically. The segmentation is carried out on multi frame MRI. Adaptive k-means is used to cluster the pixels into groups based on their intensities in order to separate the foreground (ventricle) pixels from the background pixels. Connected component labeling is used to group the pixels into regions based on their connectivity in order to segment the LV pixel region from the other regions of the MRI image. This novel combined method eliminates the problem of initialization and iteration and it segments the LV accurately on multi frame MRI with sub-second fast computational times in the range of 0.01-0.1 seconds per frame. Thus this method achieves left ventricle segmentation for one frame in sub-second duration, much less than the time required for a single iteration in deformable model methods such as level sets and active contours. The automatic segmentations accuracy was also validated on two frames as the correlation coefficient between the automatic and manually traced LV boundaries (0.992 for frame 1 and 0.993 for frame 2) was found to be higher than the correlation coefficient between two manually traced LV boundaries (0.984 for frame 1 and 0.900 for frame 2) for the same frame.

Image based sub-second fast fully automatic complete cardiac cycle left ventricle segmentation in multi frame cardiac MRI images using pixel clustering and labelling

This research presents a fully automatic sub-second fast method for left ventricle (LV) segmentation from clinical cardiac MRI images based on fast continuous max flow graph cuts and connected component labeling. The motivation for LV segmentation is to measure cardiac disease in a patient based on left ventricular function. This novel classification scheme of graph cuts labeling removes the need for manual segmentation and initialization with a seed point, since it automatically accurately extracts the LV in all slices of the full cardiac cycle in multi-frame MRI. This LV segmentation method achieves a sub-second fast computational time of 0.67 seconds on average per frame. The validity of the graph cuts labeling based automatic segmentation technique was verified by comparison with manual segmentation. Medical parameters like End Systolic Volume (ESV), End Diastolic Volume (EDV) and Ejection Fraction (EF) were calculated both automatically and manually and compared for accuracy.