Earl E. Gose

A Comparison of Seven Techniques for Choosing Subsets of Pattern Recognition Properties

The only guaranteed technique for choosing the best subset of N properties from a set of M is to try all (MN) possible combinations. This is computationally impractical for sets of even moderate size, so heuristic techniques are required. This paper presents seven techniques for choosing good subsets of properties and compares their performance on a nine-class vectorcardiogram classification problem.

Classification of benign and malignant breast tumors on the basis of 36 radiographic properties

Using a semiquantitative question sheet, a radiologist estimated 36 radiographic properties upon each of 102 pathologically proven cases of benign and malignant disease of the breast. From these properties, a probability of malignancy was assigned to each case using an automatic clustering algorithm. The algorithm examined the properties of half the cases and evolved parameters with which to assign probabilities of malignancy to unseen cases. Then the validity of these parameters was tested using the other half of cases. Every mammogram used in this study had a pathologic diagnosis indicating that the surgeon was suspicious enough of malignant disease to operate. Using the properties and decision method described in the paper, the false‐positive rate was lowered 45% while keeping a false‐negative rate of zero in the same set of mammograms.

Dermatomal somatosensory evoked potential demonstration of nerve root decompression after VAX-D therapy

Abstract Reductions in low back pain and referred leg pain associated with a diagnosis of herniated disc, degenerative disc disease or facet syndrome have previously been reported after treatment with a VAX-D table, which intermittently distracts the spine. The object of this study was to use dermatomal somatosensory evoked potentials (DSSEPs) to demonstrate lumbar root decompression following VAX-D therapy. Seven consecutive patients with a diagnosis of low back pain and unilateral or bilateral L5 or S1 radiculopathy were studied at our center. Disc herniation at the L5-S1 level was documented by MRI or CT in all patients. All patients were studied bilaterally by DSSEPs at L5 and S1 before and after VAX-D therapy. All patients had at least 50% improvement in radicular symptoms and low back pain and three of them experienced complete resolution of all symptoms. The average pain reduction was 77%. The number of treatment sessions varied from 12 to 35. DSSEPs were considered to show improvement if triphasic characteristics returned or a 50% or greater increase in the P1-P2 amplitude was seen. All patients showed improvement in DSSEPs after VAX-D therapy either ipsilateral or contralateral to the symptomatic leg. Two patients showed deterioration in DSSEPs in the symptomatic leg despite clinically significant improvement in pain and radicular symptoms. Overall, 28 nerve roots were studied before and after VAX-D therapy. Seventeen nerve root responses were improved, eight remained unchanged and three deteriorated. The significance of DSSEP improvement contralateral to the symptomatic leg is emphasized. Direct compression of a nerve root by a disc herniation is probably not the sole explanation for referred leg pain. [Neurol Res 2001; 23: 706-714]

An Evolutionary Pattern Recognition Network

A pattern recognition network with two types of adaptation has been investigated. The network output is a weighted sum of the outputs of elements which compute real functions of the discrete network inputs. The first type of adaptation involves the adjustment of the weights while the second type involves the periodic replacement of the least valuable network elements with new ones. The expected error of the network in realizing arbitrary input-output functions has been found by Monte-Carlo simulation for simple weight adaptation and for the case where the population of network elements is allowed to evolve. Three heuristics for determining which elements are to be replaced in each generation have been evaluated and compared. These were based on the size of the weight associated with each element after training, a normalized weight size, and the cross correlation between the elemental function and the desired network function. All three selection criteria resulted in improvements of the network performance over the nonevolutionary case. The normalized weight size criterion was most effective while the cross-correlation criterion was least effective.

INTRODUCTION TO BIOLOGICAL AND MECHANICAL PATTERN RECOGNITION

Publisher Summary Pattern recognition or classification is a process by which groups input signals or stimuli (patterns) into categories according to their properties. The ability of animals to recognize patterns is one aspect of their intelligence, and the study of mechanical pattern recognition is a subdivision of the field of artificial intelligence. All these fields fall into the general category of cybernetics, defined as the sciences of communication and control in animals and machines. The study of pattern recognition in animals is difficult because of their behavioral and neurophysiological complexity. Behaviorally, there would always be uncertainty regarding the stimuli an organism responds to. Direct investigation of the inner workings of the human nervous system is not permissible because of the injurious nature of the present measurement techniques, except in rare cases where data may be obtained without additional risk to the patient during brain operations. This chapter highlights only two of the most important senses involved in pattern recognition, namely, vision and hearing. Of these, vision is probably the most studied, and, informationally, it is the largest input channel. However important vision may be, the loss of hearing is often considered to be more of a handicap than the loss of vision, for it is through this sense that most of the day-to-day communication is carried out.

Evolutionary Pattern Recognition in Incomplete Nonlinear Multithreshold Networks

A pattern recognition network which computes a weighted sum of nonlinear functions of its inputs is considered. An algorithm for training this multithreshold network is presented. The multithreshold device is used for classifying patterns into more than two categories. Experimental results on the recognition of hand-printed characters are shown. In this case, the nonlinear functions consisted of an orthogonal set of property detectors. The network changed its structure by an evolutionary technique which consisted of periodic replacement of the least useful elements by new ones, randomly chosen.

Time--domain analysis in gated cardiac blood pool studies.

A technique is presented for producing functional images derived from equilibrium gated blood pool studies as a means of diagnosing cardiac disease. These functional images are based on characteristics associated with the time variation of the count rate (the time domain) at each point of the image matrix rather than on the Fourier transform of the time-activity curve (the frequency domain) which has gained recent attention. As examples of this method, we present images which display the statistical variance of the time-activity curve at each pixel, corrected for the expected contribution due to random statistical fluctuation, and images which display the time at which each pixel reaches its minimum count value. Variance and time-to-minimum images are comparable to Fourier amplitude and phase images, respectively, and have been found to be useful in facilitating the diagnosis of wall motion abnormalities. A major advantage of time-domain analysis is the wide variety of features of potential clinical significance which may be investigated.

A Synthesis Technique for Networks Consisting of Logical Functions Feeding a Linear Summation Element

Compiler programs are used to convert engineering test specifications to a digital code which can be interpreted by the automatic test equipment system. A programming language based upon decision-table techniques allows the test engineer to write his test statements in an extremely convenient fashion and permits him to program any test specification with only a minimum of knowledge about the specific test equipment system and particular programming techniques.