Deniz Aykac

Statistical characterization and segmentation of drusen in fundus images

Age related Macular Degeneration (AMD) is a disease of the retina associated with aging. AMD progression in patients is characterized by drusen, pigmentation changes, and geographic atrophy, which can be seen using fundus imagery. The level of AMD is characterized by standard scaling methods, which can be somewhat subjective in practice. In this work we propose a statistical image processing approach to segment drusen with the ultimate goal of characterizing the AMD progression in a data set of longitudinal images. The method characterizes retinal structures with a statistical model of the colors in the retina image. When comparing the segmentation results of the method between longitudinal images with known AMD progression and those without, the method detects progression in our longitudinal data set with an area under the receiver operating characteristics curve of 0.99.

Practical considerations for optic nerve location in telemedicine

The projected increase in diabetes in the United States and worldwide has created a need for broad-based, inexpensive screening for diabetic retinopathy (DR), an eye disease which can lead to vision impairment. A telemedicine network with retina cameras and automated quality control, physiological feature location, and lesion / anomaly detection is a low-cost way of achieving broad-based screening. In this work we report on the effect of quality estimation on an optic nerve (ON) detection method with a confidence metric. We report on an improvement of the method using a data set from an ophthalmologist practice then show the results of the method as a function of image quality on a set of images from an on-line telemedicine network collected in Spring 2009 and another broad-based screening program. We show that the fusion method, combined with quality estimation processing, can improve detection performance and also provide a method for utilizing a physician-in-the-loop for images that may exceed the capabilities of automated processing.

Baseline face detection, head pose estimation, and coarse direction detection for facial data in the SHRP2 naturalistic driving study

Keeping a driver focused on the road is one of the most critical steps in insuring the safe operation of a vehicle. The Strategic Highway Research Program 2 (SHRP2) has over 3,100 recorded videos of volunteer drivers during a period of 2 years. This extensive naturalistic driving study (NDS) contains over one million hours of video and associated data that could aid safety researchers in understanding where the drivers attention is focused. Manual analysis of this data is infeasible; therefore efforts are underway to develop automated feature extraction algorithms to process and characterize the data. The real-world nature, volume, and acquisition conditions are unmatched in the transportation community, but there are also challenges because the data has relatively low resolution, high compression rates, and differing illumination conditions. A smaller dataset, the head pose validation study, is available which used the same recording equipment as SHRP2 but is more easily accessible with less privacy constraints. In this work we report initial head pose accuracy using commercial and open source face pose estimation algorithms on the head pose validation data set.

Automatic detection of retina disease: Robustness to image quality and localization of anatomy structure

The automated detection of diabetic retinopathy and other eye diseases in images of the retina has great promise as a low-cost method for broad-based screening. Many systems in the literature which perform automated detection include a quality estimation step and physiological feature detection, including the vascular tree and the optic nerve / macula location. In this work, we study the robustness of an automated disease detection method with respect to the accuracy of the optic nerve location and the quality of the images obtained as judged by a quality estimation algorithm. The detection algorithm features microaneurysm and exudate detection followed by feature extraction on the detected population to describe the overall retina image. Labeled images of retinas ground-truthed to disease states are used to train a supervised learning algorithm to identify the disease state of the retina image and exam set. Under the restrictions of high confidence optic nerve detections and good quality imagery, the system achieves a sensitivity and specificity of 94.8% and 78.7% with area-under-curve of 95.3%. Analysis of the effect of constraining quality and the distinction between mild non-proliferative diabetic retinopathy, normal retina images, and more severe disease states is included.

Improvements in level set segmentation of 3D small animal imagery

In this paper, we investigate several improvements to region-based level set algorithms in the context of segmenting x-ray CT data from pre-clinical imaging of small animal models. We incorporate a recently introduced signed distance preserving term into a region-based level set model and provide formulas for a semi-implicit finite difference implementation. We illustrate some pitfalls of topology preserving level sets and introduce the concept of connectivity preservation as a potential alternative. We illustrate the benefits of these improvements on phantom and real data.

Evaluating the Accuracy of Optic Nerve Detections in Retina Imaging Using Complementary Methods

The detection of the optic nerve in retina images is a key element of detecting the anatomical structure of the retina. In this work we report on the fusion of two complementary optic nerve detection methods. The methods are complementary in the sense that they use different fundamental queues for locating the optic nerve. By fusing the methods through applying simple distance-based threshold, a confidence level may be assigned to the quality of the optic nerve detection. In a practical screening system this metric can be used to determine if the images should be immediately reviewed by a physician or if further automatic processing is feasible. We report on the results for two different data sets and show that the use of the threshold improves detection quality.

New developments in image-based characterization of coated particle nuclear fuel

We describe in this paper new developments in the characterization of coated particle nuclear fuel using optical microscopy and digital imaging. As in our previous work, we acquire optical imagery of the fuel pellets in two distinct manners that we refer to as shadow imaging and cross-sectional imaging. In shadow imaging, particles are collected in a single layer on an optically transparent dish and imaged using collimated back-lighting to measure outer surface characteristics only. In cross-sectional imaging, particles are mounted in acrylic epoxy and polished to near-center to reveal the inner coating layers for measurement. For shadow imaging, we describe a curvaturebased metric that is computed from the particle boundary points in the FFT domain using a low-frequency parametric representation. We also describe how missing boundary points are approximated using band-limited interpolation so that the FFT can be applied. For cross-section imaging, we describe a new Bayesian-motivated segmentation scheme as well as a new technique to correct layer measurements for the fact that we cannot observe the true mid-plane of the approximately spherical particles.

A 3D Level Sets Method for Segmenting the Mouse Spleen and Follicles in Volumetric microCT Images

We present a semi-automatic, 3D approach for segmenting the mouse spleen, and its interior follicles, in volumetric microCT imagery. Based upon previous 2D level sets work, we develop a fully 3D implementation and provide the corresponding finite difference formulas. We incorporate statistical and proximity weighting schemes to improve segmentation performance. We also note an issue with the original algorithm and propose a solution that proves beneficial in our experiments. Experimental results are provided for artificial and real data

Combining image and non-image data for automatic detection of retina disease in a telemedicine network

A telemedicine network with retina cameras and automated quality control, physiological feature location, and lesion / anomaly detection is a low-cost way of achieving broad-based screening for diabetic retinopathy (DR) and other eye diseases. In the process of a routine eye-screening examination, other non-image data is often available which may be useful in automated diagnosis of disease. In this work, we report on the results of combining this non-image data with image data, using the protocol and processing steps of a prototype system for automated disease diagnosis of retina examinations from a telemedicine network. The system includes quality assessments, automated physiology detection, and automated lesion detection to create an archive of known cases. Non-image data such as diabetes onset date and hemoglobin A1c (HgA1c) for each patient examination are included as well, and the system is used to create a content-based image retrieval engine capable of automated diagnosis of disease into “normal” and “abnormal” categories. The system achieves a sensitivity and specificity of 91.2% and 71.6% using hold-one-out validation testing.

Quantitative Comparison of Mitotic Spindles by Confocal Image Analysis

The mitotic spindle is a subcellular protein structure that facilitates chromosome segregation and is crucial to cell division. We describe an image processing approach to quantitatively characterize and compare mitotic spindles that have been imaged three dimensionally using confocal microscopy with fixed-cell preparations. The proposed approach is based on a set of features that are computed from each image stack representing a spindle. We compare several spindle datasets of varying biological (genotype) and/or environmental (drug treatment) conditions. The goal of this effort is to aid biologists in detecting differences between spindles that may not be apparent under subjective visual inspection, and furthermore, to eventually automate such analysis in high-throughput scenarios (thousands of images) where manual inspection would be unreasonable. Experimental results on positive- and negative-control data indicate that the proposed approach is indeed effective. Differences are detected when it is known they do exist (positive control) and no differences are detected when there are none (negative control). In two other experimental comparisons, results indicate structural spindle differences that biologists had not observed previously.