Engineering topics
Publications and source records attributed to White, B. S..
Technique rapidly diagnoses diabetes mellitus. Photographs of "whites" of patients' eyes scanned by computerized image analyzer programmed to quantify density of small blood vessels in conjuctiva. Comparison with data base of known normal and diabetic patients facilitates rapid diagnosis.
This paper is concerned with the propagation of laser light through a slab of a randomly varying medium. A theoretical analysis is presented which relates the spectrum of the recorded-intensity field some distance downstream of the medium to the spectrum of the index-of-refraction field. For a homogeneous and isotropic random field, the 3-D spectrum of the medium is obtained from the 2-D spectrum of the photograph by dividing each component of the spectrum by the frequency raised to the fourth power. Free-space propagation outside the random medium is accounted for by a scaling factor. Experimental results are presented which support the theoretical analysis. The nonintrusive diagnostic technique presented here is applicable to photographs which contain partially developed caustic networks.
A stochastic method using geometrical acoustics is employed to investigate the growth of a large fluctuation in the amplitude of high-frequency waves or shocks propagating through turbulent, inhomogeneous media. Nonlinear terms are retained in the analysis to correctly model focusing and growth of singular fluctuations in the amplitude. A two-dimensional analysis reveals that fluctuations in the ray-tube area grow exponentially and every ray displays caustics. Probability densities for the appearance of caustics are provided, and moments of the ray-tube area distribution and amplitude-related statistics are formulated for distances far into the region of caustic formation. Finally, a relationship is defined between theoretical predictions and measurements on an image plane.
An important step in the diagnosis of a cervical cytology specimen is estimating the proportions of the various cell types present. This is usually done with a cell classifier, the error rates of which can be expressed as a confusion matrix. We show how to use the confusion matrix to obtain an unbiased estimate of the desired proportions. We show that the mean square error of this estimate depends on a 'befuddlement matrix' derived from the confusion matrix, and how this, in turn, leads to a figure of merit for cell classifiers. Finally, we work out the two-class problem in detail and present examples to illustrate the theory.
An analysis is presented of the results obtained from a cell classifier which is confronted with an abnormal/normal cell ratio which is different from the ratio assumed in the calibration of the classifier. False negative and false positive error rates are determined in advance for classifier operation, along with the necessary sample size in order to validate the predicted distributions. Changes are demonstrated to happen only regarding the false negative rate, where reductions in the abnormal cell rate below the expected rates would cause totally unreliable data. Substantial overproduction of abnormal cells would be quickly noticeable, while production rates beyond, but close to, the expected rates would only require more extensive sampling. Classifier systems for 10% proportions of abnormal cells are concluded to be possible, but difficulties are present with much lower rates