Pascal Boilot

Electronic noses inter-comparison, data fusion and sensor selection in discrimination of standard fruit solutions

Intensive research and fast developments in electronic nose (EN) technologies provide the users with a wide spectrum of sensors and systems for their applications. This paper presents some of the results obtained with four different ENs on a series of collaborative tests carried out on six standard fruit samples, pure liquids and mixtures. These experiments, part of the EU ASTEQ concerted action, were designed for inter-comparison of the system’s performances. Various feature extraction techniques are considered along with inter-comparison of the individual results obtained with radial basis function (RBF) and probabilistic neural networks (PNN). A low-level data fusion technique is used to merge the various datasets together, considering all extracted parameters in order to increase the amount of information available for classification. We achieve 86.7% correct classification with the fusion system, which outperforms the results obtained with individual ENs. With this fusion array, a problem of dimensionality occurs and it is possible to find an optimal array configuration of reduced dimensionality considering a subset of parameters. We report on various parameter selection methods: principal component analysis (PCA) as a mathematical transformation and two types of genetic algorithms (GAs) optimisation as search methods. Various subsets of parameters are selected and all techniques return improved classification rates, 80% with PCA, 96.7% with 6-integer gene GAs and 93.3% with 72-binary gene GAs. In order to overcome cost and technology limitations, optimisation techniques can be used to create application specific arrays selecting the best sensors or the correct parameters. # 2002 Elsevier Science B.V. All rights reserved.

Classification of bacteria responsible for ENT and eye infections using the Cyranose system

The Cyranose 320 (Cyrano Sciences Inc., USA), comprising an array of 32 polymer carbon black composite sensors, has been used to identify species of bacteria commonly associated with medical conditions. Results from two experiments are presented: one on bacteria causing eye infections and one on a new series of tests on bacteria responsible for some ear, nose, and throat (ENT) diseases. For the eye bacteria tests, pure lab cultures were used and the electronic nose (EN) was used to sample the headspace of sterile glass vials containing a fixed volume of bacteria in suspension. For the ENT bacteria, the system was taken a step closer toward medical application, as readings were taken from the headspace of the same blood agar plates used to culture real samples collected from patients. After preprocessing, principal component analysis (PCA) was used as an exploratory technique to investigate the clustering of vectors in multi-sensor space. Artificial neural networks (ANNs) were then used as predictors, and a multilayer perceptron (MLP) trained with back-propagation (BP) and with Levenberg-Marquardt was used to identify the different bacteria. The optimal MLP was found to correctly classify 97.3% of the six eye bacteria of interest and 97.6% of the four ENT bacteria including two sub-species. A radial basis function (RBF) network was able to discriminate between the six eye bacteria species, even in the lowest state of concentration, with 92.8% accuracy. These results show the potential application of the Cyranose together with neural network-based predictors, for rapid screening and early detection of bacteria associated with these medical conditions, and the possible development of this EN system as a near-patient tool in primary medical healthcare.

Comparison of neural network predictors in the classification of tracheal-bronchial breath sounds by respiratory auscultation

Despite extensive research in the area of identification and discrimination of tracheal-bronchial breath sounds by computer analysis, the process of identifying auscultated sounds is still subject to high estimation uncertainties. Here we assess the performance of the relatively new constructive probabilistic neural network (CPNN) against the more common classifiers, namely the multilayer perceptron (MLP) and radial basis function network (RBFN), in classifying a broad range of tracheal-bronchial breath sounds. We present our data as signal estimation models of the tracheal-bronchial frequency spectra. We have examined the trained structure of the CPNN with respect to the other architectures and conclude that this architecture offers an attractive means with which to analyse this type of data. This is based partly on the classification accuracies attained by the CPNN, MLP and RBFN which were 97.8, 77.8 and 96.2%, respectively. We concluded that CPNN and RBFN networks are capable of working successfully with this data, with these architectures being acceptable in terms of topological size and computational overhead requirements. We further believe that the CPNN is an attractive classification mechanism for auscultated data analysis due to its optimal data model generation properties and computationally lightweight architecture.

Classifying coronary dysfunction using neural networks through cardiovascular auscultation

The paper applies artificial neural networks (ANNs) to the analysis of heart sound abnormalities through auscultation. Audio auscultation samples of 16 different coronary abnormalities were collected. Data pre-processing included down-sampling of the auscultated data and use of the fast Fourier transform (FFT) and the Levinson-Durbin autoregression algorithms for feature extraction and efficient data encoding. These data were used in the training of a multi-layer perceptron (MLP) and radial basis function (RBF) neural network to develop a classification mechanism capable of distinguishing between different heart sound abnormalities. The MLP and RBF networks attained classification accuracies of 84% and 88%, respectively. The application of ANNs to the analysis of respiratory auscultation and consequently the development of a combined cardio-respiratory analysis system using auscultated data could lead to faster and more efficient treatment.