Stanley M. Dunn

Measuring the area and volume of the human body with structured light

An inexpensive computer imaging system capable of accurately recovering the 3D surface of the human body is described. This system uses biologically safe structured white light. A uniform square grid pattern is projected onto the body and an image of this pattern is recorded using a single solid-state camera. By locating the intersections of the image grid stripes and matching them correctly to the projected grid pattern, the 3D positions of points on the body can be determined by triangulation. The authors summarize the processing, discuss the geometry of the imaging system and show how 3D information can be recovered. They describe the actual processing steps and algorithms used to locate the data used to reconstruct the patch. Experiments are presented. These preliminary results demonstrate the feasibility of the imaging system. Future developments are discussed. >

Biosignal pattern recognition and interpretation systems

A general framework is given to describe pattern recognition and interpretation. Pattern analysis stages are described, with consideration of difficulties in implementation and uncertainties present at each level. The main forms of pattern analysis-statistical, syntactic, and artificial intelligence (connectionist and symbolic) methods-have different strengths and weaknesses, depending on the stage of pattern analysis at which they are used. In general, statistical, syntactic, and connectionist techniques are used for pattern recognition, and statistical and symbolic techniques are used for pattern interpretation. Largely, pattern interpretation involves reasoning with uncertainty. Multichannel recordings increase the information available about specific physiologic events, at the expense of processing complexity.<<ETX>>

Learning shape classes

This paper is a summary of a new approach to learning shape concepts. In this system, a shape is represented by conjunctions of local shape properties. Conjunctions of local properties are consistent and unique for distinct shapes and are robust enough to represent shape in the presence of occlusion. A new learning method, called property based learning, is developed and used to learn conjunctions of local properties. Unlike other classification methods based on distances or similarities, classification performance does not degrade linearly as the number of classes increases and classification can be done correctly with only partial information of instances. Property based learning is an incremental learning method that selects properties crucial for classification. Two experiments are reported. In the first experiment with tool shapes, this shape learning system is used to classify shapes in the presence of view point changes, local movements such as moving handles of pliers, and occlusion. In the second experiment with hand gestures, the system can classify different gestures regardless of the movement in joints, fingers, and palms. >

The performance of projective standardization for digital subtraction radiography.

OBJECTIVE We sought to test the performance and robustness of projective standardization in preserving invariant properties of subtraction images in the presence of irreversible projection errors. Study design Twenty bone chips (1-10 mg each) were placed on dentate dry mandibles. Follow-up images were obtained without the bone chips, and irreversible projection errors of up to 6 degrees were introduced. Digitized image intensities were normalized, and follow-up images were geometrically reconstructed by 2 operators using anatomical and fiduciary landmarks. Subtraction images were analyzed by 3 observers. RESULTS Regression analysis revealed a linear relationship between radiographic estimates of mineral loss and actual mineral loss (R(2) = 0.99; P <.05). The effect of projection error was not significant (general linear model [GLM]: P >.05). There was no difference between the radiographic estimates from images standardized with anatomical landmarks and those standardized with fiduciary landmarks (Wilcoxon signed rank test: P >.05). Operator variability was low for image analysis alone (R(2) = 0.99; P <.05), as well as for the entire procedure (R(2) = 0.98; P <.05). The predicted detection limit was smaller than 1 mg. CONCLUSIONS Subtraction images registered by projective standardization yield estimates of osseous change that are invariant to irreversible projection errors of up to 6 degrees. Within these limits, operator precision is high and anatomical landmarks can be used to establish correspondence.

Texture classification using noncausal hidden Markov models

This paper addresses the problem of using noncausal hidden Markov models (HMMs) for texture classification. In noncausal models, the state of each pixel may be dependent on its neighbors in all directions. New algorithms are given to learn the parameters of a noncausal HMM of a texture and to classify it into one of several learned categories. Texture classification results using these algorithms are provided. >

The perception of breast cancers-a spatial frequency analysis of what differentiates missed from reported cancers

The primary detector of breast cancer is the human eye. Radiologists read mammograms by mapping exogenous and endogenous factors, which are based on the image and observer, respectively, into observer-based decisions. These decisions rely on an internal schema that contains a representation of possible malignant and benign findings. Thus, to understand the hits and misses made by the radiologists, it is important to model the interactions between the measurable image-based elements contained in the mammogram and the decisions made. The image-based elements can be of two types, i.e., areas that attracted the visual attention of the radiologist, but did not yield a report, and areas where the radiologist indicated the presence of an abnormal finding. In this way, overt and covert decisions are made when reading a mammogram. In order to model this decision-making process, we use a system that is based upon the processing done by the human visual system, which decomposes the areas under scrutiny in elements of different sizes and orientations. In our system, this decomposition is done using wavelet packets (WPs). Nonlinear features are then extracted from the WP coefficients, and an artificial neural network is trained to recognize the patterns of decisions made by each radiologist. Afterwards, the system is used to predict how the radiologist will respond to visually selected areas in new mammogram cases.

Biosignal pattern recognition and interpretation systems. 3. Methods of classification

The following topics are discussed: Bayes/minimum distance classifiers; maximum likelihood classification estimation; k-nearest neighbor classification; entropy criteria; syntactic techniques; string matching; the Cocke-Younger-Kasami parsing algorithm; syntactic learning; finite-state automata; neural network classification techniques; learning vector quantization; cluster swapping; hierarchial clustering procedures.<<ETX>>

Biosignal pattern recognition and interpretation systems. 2. Methods for feature extraction and selection

Some feature extraction methods used in biomedical signal pattern recognition are presented. Particular attention is given to nontransformed signal characteristics, transformed signal characteristics, structural descriptors, graph descriptors, and feature selection methods. It is noted that the wide variety of techniques used for feature extraction presents two problems: which techniques should be used and how to select from among the features that each extraction technique generates. Selected features are best only by some standard; therefore, techniques for generation of features tend not to be very portable from one pattern processing problem to another. Production of salient features is the connecting link between prototypical and symbolic representations of a class. Often, thresholds govern the selection of features. Many techniques do not generate independent features; therefore, there is redundancy in the data, which potentially affects both efficiency and accuracy in pattern recognition.<<ETX>>

Using an architectural knowledge base to generate code for parallel computers

The authors present a reconfigurable compiler for distributed memory parallel computers that performs automatic program partitioning, mapping, and communication code generation under the guidance of directives supplied by the programmer.

An analysis of perceptual errors in reading mammograms using quasi-local spatial frequency spectra.

In this pilot study the authors examined areas on a mammogram that attracted the visual attention of experienced mammographers and mammography fellows, as well as areas that were reported to contain a malignant lesion, and, based on their spatial frequency spectrum, they characterized these areas by the type of decision outcome that they yielded: true-positives (TP), false-positives (FP), true-negatives (TN), and false-negatives (FN). Five 2-view (craniocaudal and medial-lateral oblique) mammogram cases were examined by 8 experienced observers, and the eye position of the observers was tracked. The observers were asked to report the location and nature of any malignant lesions present in the case. The authors analyzed each area in which either the observer made a decision or in which the observer had prolonged (>1,000 ms) visual dwell using wavelet packets, and characterized these areas in terms of the energy contents of each spatial frequency band. It was shown that each decision outcome is characterized by a specific profile in the spatial frequency domain, and that these profiles are significantly different from one another. As a consequence of these differences, the profiles can be used to determine which type of decision a given observer will make when examining the area. Computer-assisted perception correctly predicted up to 64% of the TPs made by the observers, 77% of the FPs, and 70% of the TNs.