Sergio Murillo

Multiscale AM-FM Methods for Diabetic Retinopathy Lesion Detection

In this paper, we propose the use of multiscale amplitude-modulation-frequency-modulation (AM-FM) methods for discriminating between normal and pathological retinal images. The method presented in this paper is tested using standard images from the early treatment diabetic retinopathy study. We use 120 regions of 40 × 40 pixels containing four types of lesions commonly associated with diabetic retinopathy (DR) and two types of normal retinal regions that were manually selected by a trained analyst. The region types included microaneurysms, exudates, neovascularization on the retina, hemorrhages, normal retinal background, and normal vessels patterns. The cumulative distribution functions of the instantaneous amplitude, the instantaneous frequency magnitude, and the relative instantaneous frequency angle from multiple scales are used as texture feature vectors. We use distance metrics between the extracted feature vectors to measure interstructure similarity. Our results demonstrate a statistical differentiation of normal retinal structures and pathological lesions based on AM-FM features. We further demonstrate our AM-FM methodology by applying it to classification of retinal images from the MESSIDOR database. Overall, the proposed methodology shows significant capability for use in automatic DR screening.

An AM-FM model for Motion Estimation in Atherosclerotic Plaque Videos

We present new multidimensional amplitude-modulation frequency-modulation (AM-FM) methods for motion estimation. For a single AM-FM component we show that the optical flow constraint leads to separate equations for amplitude modulation (AM) and frequency modulation (FM). We compare our approach with phase-based estimation developed by Fleet and Jepson and also the original optical flow method by Horn and Schunck. An advantage of the proposed method is that it provides for dense estimates that remain accurate over the entire video. We also present preliminary results on atherosclerotic plaques videos where the AM method appears to work best.

Atherosclerotic Plaque Motion Analysis from Ultrasound Videos

In this paper, we describe a collection of algorithms that are used to provide motion trajectory estimates from atherosclerotic plaque ultrasound videos. Our approach is based on the use of four different optical flow methods to estimate motion vectors (Horn and Schunk, Lucas, Nagel and Uras). To estimate the optimal motion estimation parameters, we perform hundreds of experiments on a Linux cluster, and further validate the results using synthetic simulations. Following motion estimation, we compute pixel motion trajectories over the plaque regions and vessel walls. Pixel trajectories are then used to assess plaque deformation.

Independent Component Analysis for Vision-inspired Classification of Retinal Images with Age-related Macular Degeneration

The purpose of this paper is to present a novel approach for extracting image-based features for classifying age-related macular degeneration (AMD) in digital retinal images. 100 retinal images were classified by an ophthalmologist into 12 categories based on the visual characteristics of the disease. Independent Component Analysis (ICA) was used to extract features at different spatial scales to be used as input to a classifier. The classification used a type of regression, partial least squares. In this experiment ICA replicated the ophthalmologists visual classification by correctly assigning all 12 images from two of the classes.

Detection and phenotyping of retinal disease using AM-FM processing for feature extraction

We present the application of an Amplitude-Modulation Frequency-Modulation (AM-FM) method for extracting potentially relevant features towards the classification of diseased retinas from healthy retinas. In terms of AM-FM features, we use histograms of the instantaneous amplitude, the angle of the instantaneous frequency and the magnitude of the instantaneous frequency extracted over different frequency scales. To classify the AM-FM features, we use a combination of a clustering method and Partial Least Squares (PLS). Using 18 images from each of the four risk levels, three experiments were performed to test the algorithms ability to differentiate the controls (Risk 0) from each of the three levels of pathology, i.e. Risk 1, Risk 2, and Risk 3. For Risk 0 versus Risk 3 an area under the receiver operating system (AROC) of 0.99 was achieved with a best sensitivity of 100% and a specificity of 95%. For Risk 0 versus Risk 2, the AROC was 0.96 with 94% sensitivity and 85% specificity. For Risk 0 versus Risk 1, the AROC was 0.93 and a sensitivity/specificity of 94%/67%.

Multi-scale AM-FM motion analysis of ultrasound videos of carotid artery plaques

An estimated 82 million American adults have one or more type of cardiovascular diseases (CVD). CVD is the leading cause of death (1 of every 3 deaths) in the United States. When considered separately from other CVDs, stroke ranks third among all causes of death behind diseases of the heart and cancer. Stroke accounts for 1 out of every 18 deaths and is the leading cause of serious long-term disability in the United States. Motion estimation of ultrasound videos (US) of carotid artery (CA) plaques provides important information regarding plaque deformation that should be considered for distinguishing between symptomatic and asymptomatic plaques. In this paper, we present the development of verifiable methods for the estimation of plaque motion. Our methodology is tested on a set of 34 (5 symptomatic and 29 asymptomatic) ultrasound videos of carotid artery plaques. Plaque and wall motion analysis provides information about plaque instability and is used in an attempt to differentiate between symptomatic and asymptomatic cases. The final goal for motion estimation and analysis is to identify pathological conditions that can be detected from motion changes due to changes in tissue stiffness.

A review of motion estimation methods for non-invasive ultrasound motion and emerging strain imaging methods of carotid artery plaques

Non-invasive ultrasound imaging of carotid plaques is used in routine clinical evaluation of atherosclerosis and stroke. Strain imaging of the atherosclerotic plaques represents a very promising, emerging application of ultrasound imaging because it can assess plaque vulnerability without the risks of intervention. The promise of strain imaging includes the development of advanced diagnostic tools that can be used to predict plaque rupture. This review paper presents the fundamental assumptions and methods that can be used to extract motion estimates from ultrasound images. In terms of assumptions, both the constant and non-constant brightness models are presented. The most commonly used energy functionals are presented, along with both local and global solutions. Motion and strain imaging examples are also provided to demonstrate the methods. Our goal is that this review will facilitate the development of new, reliable strain imaging methods that can be used to assess the risk of plaque rupture.

Toward comprehensive detection of sight threatening retinal disease using a multiscale AM-FM methodology

In the United States and most of the western world, the leading causes of vision impairment and blindness are age-related macular degeneration (AMD), diabetic retinopathy (DR), and glaucoma. In the last decade, research in automatic detection of retinal lesions associated with eye diseases has produced several automatic systems for detection and screening of AMD, DR, and glaucoma. However. advanced, sight-threatening stages of DR and AMD can present with lesions not commonly addressed by current approaches to automatic screening. In this paper we present an automatic eye screening system based on multiscale Amplitude Modulation-Frequency Modulation (AM-FM) decompositions that addresses not only the early stages, but also advanced stages of retinal and optic nerve disease. Ten different experiments were performed in which abnormal features such as neovascularization, drusen, exudates, pigmentation abnormalities, geographic atrophy (GA), and glaucoma were classified. The algorithm achieved an accuracy detection range of [0.77 to 0.98] area under the ROC curve for a set of 810 images. When set to a specificity value of 0.60, the sensitivity of the algorithm to the detection of abnormal features ranged between 0.88 and 1.00. Our system demonstrates that, given an appropriate training set, it is possible to use a unique algorithm to detect a broad range of eye diseases.

Quantitative and qualitative image quality analysis of super resolution images from a low cost scanning laser ophthalmoscope

The lurking epidemic of eye diseases caused by diabetes and aging will put more than 130 million Americans at risk of blindness by 2020. Screening has been touted as a means to prevent blindness by identifying those individuals at risk. However, the cost of most of todays commercial retinal imaging devices makes their use economically impractical for mass screening. Thus, low cost devices are needed. With these devices, low cost often comes at the expense of image quality with high levels of noise and distortion hindering the clinical evaluation of those retinas. A software-based super resolution (SR) reconstruction methodology that produces images with improved resolution and quality from multiple low resolution (LR) observations is introduced. The LR images are taken with a low-cost Scanning Laser Ophthalmoscope (SLO). The non-redundant information of these LR images is combined to produce a single image in an implementation that also removes noise and imaging distortions while preserving fine blood vessels and small lesions. The feasibility of using the resulting SR images for screening of eye diseases was tested using quantitative and qualitative assessments. Qualitatively, expert image readers evaluated their ability of detecting clinically significant features on the SR images and compared their findings with those obtained from matching images of the same eyes taken with commercially available high-end cameras. Quantitatively, measures of image quality were calculated from SR images and compared to subject-matched images from a commercial fundus imager. Our results show that the SR images have indeed enough quality and spatial detail for screening purposes.

Global optimization for motion estimation with applications to ultrasound videos of carotid artery plaques

Motion estimation from digital video is an ill-posed problem that requires a regularization approach. Regularization introduces a smoothness constraint that can reduce the resolution of the velocity estimates. The problem is further complicated for ultrasound videos (US), where speckle noise levels can be significant. Motion estimation using optical flow models requires the modification of several parameters to satisfy the optical flow constraint as well as the level of imposed smoothness. Furthermore, except in simulations or mostly unrealistic cases, there is no ground truth to use for validating the velocity estimates. This problem is present in all real video sequences that are used as input to motion estimation algorithms. It is also an open problem in biomedical applications like motion analysis of US of carotid artery (CA) plaques. In this paper, we study the problem of obtaining reliable ultrasound video motion estimates for atherosclerotic plaques for use in clinical diagnosis. A global optimization framework for motion parameter optimization is presented. This framework uses actual carotid artery motions to provide optimal parameter values for a variety of motions and is tested on ten different US videos using two different motion estimation techniques.