No abstract

We introduce an iterative thresholding algorithm for the segmentation of cells from noisy cell images. The thresholding image, which is initially a constant, changes iteratively with both the previous segmentation and image local activity. Experimental results for both synthesized and real cell images are provided to demonstrate the performance of the algorithm.

Background. Ovarian dysplasia has been defined by histologic1,2 and morphometric studies3,4 focusing on architectural and nuclear profile changes. A new technique is used to enhance the accuracy of this diagnosis by a quantitative evaluation of the nuclear texture that represents the nuclear chromatin pattern on which conventional diagnoses of malignancy are usually made. Methods. Histologic sections from 35 ovaries classified as malignant (12), dysplastic (12), and normal (11) were evaluated by tracing boundaries of nuclear profiles and measuring “textons” (texture primitives) with a histogram analysis of three zones of gray level densities (called for simplification white, gray, and dark). The average combined area was tabulated for specimens with the same diagnosis, and linear regression plots compared the texton area with total nuclear area. Results. The dimensions of textons originally hidden inside the chromatin and revealed by histograms were found to be closely clustered in normal epithelium, and increasingly dissociated from the containing nucleus as the lesion progressed from dysplastic to malignant. The statistical multivariate analysis including nine parameters correctly classified the three diagnostic categories as normal, dysplastic, and malignant. Conclusions. Computerized image analysis of nuclear texture adds accuracy to the recently elaborated morphometric methods to define ovarian dysplasia, a potential precursor of ovarian carcinoma.

SUMMARYAccurate edge detection is a fundamental problem in the areas of image processing and pattern recognition/classification. The lack of effective edge detection methods has slowed the application of image processing to many areas, in particular diagnostic cytology, and is a major factor in the lack of acceptance of image processing in service orientated pathology. In this paper, we present a two‐step procedure which detects edges. Since most images are corrupted by noise and often contain artefacts, the first step is to clean up the image. Our approach is to use a median filter to reduce noise and background artefacts. The second operation is to locate image pixels which are ‘information rich’ by using local statistics. This step locates the regions of the image most likely to contain edges. The application of a threshold can then pin‐point those pixels forming the edge of structures of interest. The procedure has been tested on routine cytologic specimens.

Health disparities are preventable differences in the incidence, prevalence and burden of disease among communities targeted by gender, geographic location, ethnicity and/or socio-economic status. While biomedical research has identified partial origin(s) of divergent burden and impact of disease, the innovation needed to eradicate health disparities in the United States requires unique engagement from biomedical engineers. Increasing awareness of the prevalence and consequences of health disparities is particularly attractive to today’s undergraduates, who have undauntedly challenged paradigms believed to foster inequality. Here, the Department of Biomedical Engineering at The City College of New York (CCNY) has leveraged its historical mission of access-and-excellence to integrate the study of health disparities into undergraduate BME curricula. This article describes our novel approach in a multiyear study that: (i) Integrated health disparities modules at all levels of the required undergraduate BME curriculum; (ii) Developed opportunities to include impacts of health disparities into undergraduate BME research projects and mentored High School summer STEM training; and (iii) Established health disparities-based challenges as BME capstone design and/or independent entrepreneurship projects. Results illustrate the rising awareness of health disparities among the youngest BMEs-to-be, as well as abundant undergraduate desire to integrate health disparities within BME education and training.

Summary A measure of texture, the nuclear diffuseness, was formulated for use in biological classification, and specifically to characterize quantitatively chromatin texture. Nuclear diffuseness corresponds to the amount of local intensity variation in the digitized image of a nuclear profile. As a setting in which to test the efficacy of nuclear diffuseness as a diagnostic tool, the identification of parathyroid adenoma and carcinoma was considered. Digitized images of sections of parathyroid chief cell nuclei were obtained from 16 biopsies, and the nuclear diffuseness, as well as other morphometric descriptors, were computed. With just the average nuclear diffuseness and average nuclear profile area, jackknife (leave‐one‐out) classification using an artificial neural network was able to diagnose correctly and unambiguously the condition (normal, parathyroid adenoma, or parathyroid carcinoma) in 15 of 16 cases. In one case, the neural network assigned a higher weight to the correct diagnosis, but was unable to distinguish between normal and adenoma conclusively.

No abstract

Summary An interactive semi‐automatic procedure for extraction of cell contours from light microscope images is presented. The user is required to specify four contour points and the algorithm determines the rest of the contour automatically. The algorithm exploits the fact that cell contours have lower grey level than their immediate surrounding and are usually very similar in shape to some piece‐wise ellipses. A cost function is defined to detect the cell contours incorporating both the elliptical shape and local image intensities. The procedure is fast, reliable and well suited for routine interactive applications.