We have developed a double-matching method and an artificial visual neural network technique for lung nodule detection. This neural network technique is generally applicable to the recognition of medical image pattern in gray scale imaging. The structure of the artificial neural net is a simplified network structure of human vision. The fundamental operation of the artificial neural network is local two-dimensional convolution rather than full connection with weighted multiplication. Weighting coefficients of the convolution kernels are formed by the neural network through backpropagated training. In addition, we modeled radiologists' reading procedures in order to instruct the artificial neural network to recognize the image patterns predefined and those of interest to experts in radiology. We have tested this method for lung nodule detection. The performance studies have shown the potential use of this technique in a clinical setting. This program first performed an initial nodule search with high sensitivity in detecting round objects using a sphere template double-matching technique. The artificial convolution neural network acted as a final classifier to determine whether the suspected image block contains a lung nodule. The total processing time for the automatic detection of lung nodules using both prescan and convolution neural network evaluation was about 15 seconds in a DEC Alpha workstation.

Artificial convolution neural network techniques and applications for lung nodule detection

The generalized back propagation algorithm is extended to multi-layer higher-order neural networks (HONNs). The performance of HONNs is presented. Two basic structures, the standard form and the polynomial form, are discussed. The performance of these two structures is compared using the classical TC test case, the geometric rotation problem. Simulation results show that both types of constructing strategies can recognize noisy data under rotation up to 70% and noisy irrational data up to 94%. The effect of the number of hidden neurons is discussed. >

Polynomial and standard higher order neural network

Higher Order Neural Network Group-based Adaptive Tolerance Trees

New open box and nonlinear model of Ultra High Frequency Polynomial and Sine Artificial Higher Order Neural Network (UPS-HONN) is presented in this paper. A new learning algorithm for UPS-HONN is also developed from this study. A control signal generating system, UPS-HONN Simulator, is built based on the UPS-HONN models. Test results show that, to generate any nonlinear control signal, average error of UPS-HONN models is under 1e-6.

Ultra high frequency polynomial and sine artificial higher order neural networks for control signal generator

A selection scheme to determine the optimal number of input nodes for neural networks is presented. The proposed scheme selects only the features contributing to correct classification. Experimental results indicates that the proposed selection scheme improves the speed of neural networks.

Optimization of input size for neural networks

A proposal by M. B. Reid et al. (1989) to improve the efficiency of higher-order neural networks was built into a pattern recognition system that autonomously learns to categorize and recognize patterns independently of their position in an input image. It does this by combining higher-order with first-order networks and the mechanisms known from ART. Its recognition is based on a 16*16 pixel input which contains a section of the image found by a separate centering mechanism. With this system position invariant recognition can be implemented efficiently, while combining all the advantages of the subsystems. >

G. Linhart

Papers

A self-learning visual pattern explorer and recognizer using a higher order neural network