Online Sensor Drift Compensation for E-Nose Systems Using Domain Adaptation and Extreme Learning Machine

Volatile Organic Compound Assessment as a Screening Tool for Early Detection of Gastrointestinal Diseases

Gastrointestinal (GI) diseases have a high prevalence throughout the United States. Screening and diagnostic modalities are often expensive and invasive, and therefore, people do not utilize them effectively. Lack of proper screening and diagnostic assessment may lead to delays in diagnosis, more advanced disease at the time of diagnosis, and higher morbidity and mortality rates. Research on the intestinal microbiome has demonstrated that dysbiosis, or unfavorable alteration of organismal composition, precedes the onset of clinical symptoms for various GI diseases. GI disease diagnostic research has led to a shift towards non-invasive methods for GI screening, including chemical-detection tests that measure changes in volatile organic compounds (VOCs), which are the byproducts of bacterial metabolism that result in the distinct smell of stool. Many of these tools are expensive, immobile benchtop instruments that require highly trained individuals to interpret the results. These attributes make them difficult to implement in clinical settings. Alternatively, electronic noses (E-noses) are relatively cheaper, handheld devices that utilize multi-sensor arrays and pattern recognition technology to analyze VOCs. The purpose of this review is to (1) highlight how dysbiosis impacts intestinal diseases and how VOC metabolites can be utilized to detect alterations in the microbiome, (2) summarize the available VOC analytical platforms that can be used to detect aberrancies in intestinal health, (3) define the current technological advancements and limitations of E-nose technology, and finally, (4) review the literature surrounding several intestinal diseases in which headspace VOCs can be used to detect or predict disease.

Gas Recognition in E-Nose System: A Review

Gas recognition is essential in an electronic nose (E-nose) system, which is responsible for recognizing multivariate responses obtained by gas sensors in various applications. Over the past decades, classical gas recognition approaches such as principal component analysis (PCA) have been widely applied in E-nose systems. In recent years, artificial neural network (ANN) has revolutionized the field of E-nose, especially spiking neural network (SNN). In this paper, we investigate recent gas recognition methods for E-nose, and compare and analyze them in terms of algorithms and hardware implementations. We find each classical gas recognition method has a relatively fixed framework and a few parameters, which makes it easy to be designed and perform well with limited gas samples, but weak in multi-gas recognition under noise. While ANN-based methods obtain better recognition accuracy with flexible architectures and lots of parameters. However, some ANNs are too complex to be implemented in portable E-nose systems, such as deep convolutional neural networks (CNNs). In contrast, SNN-based gas recognition methods achieve satisfying accuracy and recognize more types of gases, and could be implemented with energy-efficient hardware, which makes them a promising candidate in multi-gas identification.

Online Inertial Machine Learning for Sensor Array Long-Term Drift Compensation

The sensor drift problem is objective and inevitable, and drift compensation has essential research significance. For long-term drift, we propose a data preprocessing method, which is different from conventional research methods, and a machine learning framework that supports online self-training and data analysis without additional sensor production costs. The data preprocessing method proposed can effectively solve the problems of sign error, decimal point error, and outliers in data samples. The framework, which we call inertial machine learning, takes advantage of the recent inertia of high classification accuracy to extend the reliability of sensors. We establish a reasonable memory and forgetting mechanism for the framework, and the choice of base classifier is not limited. In this paper, we use a support vector machine as the base classifier and use the gas sensor array drift dataset in the UCI machine learning repository for experiments. By analyzing the experimental results, the classification accuracy is greatly improved, the effective time of the sensor array is extended by 4–10 months, and the time of single response and model adjustment is less than 300 ms, which is well in line with the actual application scenarios. The research ideas and results in this paper have a certain reference value for the research in related fields.

Recent Progress in Smart Electronic Nose Technologies Enabled with Machine Learning Methods

Machine learning methods enable the electronic nose (E-Nose) for precise odor identification with both qualitative and quantitative analysis. Advanced machine learning methods are crucial for the E-Nose to gain high performance and strengthen its capability in many applications, including robotics, food engineering, environment monitoring, and medical diagnosis. Recently, many machine learning techniques have been studied, developed, and integrated into feature extraction, modeling, and gas sensor drift compensation. The purpose of feature extraction is to keep robust pattern information in raw signals while removing redundancy and noise. With the extracted feature, a proper modeling method can effectively use the information for prediction. In addition, drift compensation is adopted to relieve the model accuracy degradation due to the gas sensor drifting. These recent advances have significantly promoted the prediction accuracy and stability of the E-Nose. This review is engaged to provide a summary of recent progress in advanced machine learning methods in E-Nose technologies and give an insight into new research directions in feature extraction, modeling, and sensor drift compensation.

A Sparsity-Driven Backpropagation-Less Learning Framework Using Populations of Spiking Growth Transform Neurons

Growth-transform (GT) neurons and their population models allow for independent control over the spiking statistics and the transient population dynamics while optimizing a physically plausible distributed energy functional involving continuous-valued neural variables. In this paper we describe a backpropagation-less learning approach to train a network of spiking GT neurons by enforcing sparsity constraints on the overall network spiking activity. The key features of the model and the proposed learning framework are: (a) spike responses are generated as a result of constraint violation and hence can be viewed as Lagrangian parameters; (b) the optimal parameters for a given task can be learned using neurally relevant local learning rules and in an online manner; (c) the network optimizes itself to encode the solution with as few spikes as possible (sparsity); (d) the network optimizes itself to operate at a solution with the maximum dynamic range and away from saturation; and (e) the framework is flexible enough to incorporate additional structural and connectivity constraints on the network. As a result, the proposed formulation is attractive for designing neuromorphic tinyML systems that are constrained in energy, resources, and network structure. In this paper, we show how the approach could be used for unsupervised and supervised learning such that minimizing a training error is equivalent to minimizing the overall spiking activity across the network. We then build on this framework to implement three different multi-layer spiking network architectures with progressively increasing flexibility in training and consequently, sparsity. We demonstrate the applicability of the proposed algorithm for resource-efficient learning using a publicly available machine olfaction dataset with unique challenges like sensor drift and a wide range of stimulus concentrations. In all of these case studies we show that a GT network trained using the proposed learning approach is able to minimize the network-level spiking activity while producing classification accuracy that are comparable to standard approaches on the same dataset.

Sensor Failure Tolerable Machine Learning-Based Food Quality Prediction Model

For the agricultural food production sector, the control and assessment of food quality is an essential issue, which has a direct impact on both human health and the economic value of the product. One of the fundamental properties from which the quality of the food can be derived is the smell of the product. A significant trend in this context is machine olfaction or the automated simulation of the sense of smell using a so-called electronic nose or e-nose. Hereby, many sensors are used to detect compounds, which define the odors and herewith the quality of the product. The proper assessment of the food quality is based on the correct functioning of the adopted sensors. Unfortunately, sensors may fail to provide the correct measures due to, for example, physical aging or environmental factors. To tolerate this problem, various approaches have been applied, often focusing on correcting the input data from the failed sensor. In this study, we adopt an alternative approach and propose machine learning-based failure tolerance that ignores failed sensors. To tolerate for the failed sensor and to keep the overall prediction accuracy acceptable, a Single Plurality Voting System (SPVS) classification approach is used. Hereby, single classifiers are trained by each feature and based on the outcome of these classifiers, and a composed classifier is built. To build our SPVS-based technique, K-Nearest Neighbor (kNN), Decision Tree, and Linear Discriminant Analysis (LDA) classifiers are applied as the base classifiers. Our proposed approach has a clear advantage over traditional machine learning models since it can tolerate the sensor failure or other types of failures by ignoring and thus enhance the assessment of food quality. To illustrate our approach, we use the case study of beef cut quality assessment. The experiments showed promising results for beef cut quality prediction in particular, and food quality assessment in general.

Sensor Drift Compensation Based on the Improved LSTM and SVM Multi-Class Ensemble Learning Models

Drift is an important issue that impairs the reliability of sensors, especially in gas sensors. The conventional method usually adopts the reference gas to compensate for the drift. However, its classification accuracy is not high. We propose a supervised learning algorithm that is based on multi-classifier integration for drift compensation in this paper, which incorporates drift compensation into the classification process, motivated by the fact that the goal of drift compensation is to improve the classification performance. In our method, with the obtained characteristics of sensors and the advantage of Support Vector Machine (SVM) in few-shot classification, the improved Long Shot Term Memory (LSTM) is integrated to build the multi-class classifier model. We tested the proposed approach on the publicly available time series dataset that was collected over three years by the metal-oxide gas sensors. The results clearly indicate the superiority of multiple classifier approach, which achieves higher classification accuracy as compared with different approaches during testing period with an ensemble of classifiers in the presence of sensor drift over time.

A Spike Time-Dependent Online Learning Algorithm Derived From Biological Olfaction

We have developed a spiking neural network (SNN) algorithm for signal restoration and identification based on principles extracted from the mammalian olfactory system and broadly applicable to input from arbitrary sensor arrays. For interpretability and development purposes, we here examine the properties of its initial feedforward projection. Like the full algorithm, this feedforward component is fully spike timing-based, and utilizes online learning based on local synaptic rules such as spike timing-dependent plasticity (STDP). Using an intermediate metric to assess the properties of this initial projection, the feedforward network exhibits high classification performance after few-shot learning without catastrophic forgetting, and includes a none of the above outcome to reflect classifier confidence. We demonstrate online learning performance using a publicly available machine olfaction dataset with challenges including relatively small training sets, variable stimulus concentrations, and 3 years of sensor drift.

Odor Fingerprint Analysis Using Feature Mining Method Based on Olfactory Sensory Evaluation

In this paper, we aim to use odor fingerprint analysis to identify and detect various odors. We obtained the olfactory sensory evaluation of eight different brands of Chinese liquor by a lab-developed intelligent nose. From the respective combination of the time domain and frequency domain, we extract features to reflect the samples comprehensively. However, the extracted feature combined time domain and frequency domain will bring redundant information that affects performance. Therefore, we proposed data by Principal Component Analysis (PCA) and Variable Importance Projection (VIP) to delete redundant information to construct a more precise odor fingerprint. Then, Random Forest (RF) and Probabilistic Neural Network (PNN) were built based on the above. Results showed that the VIP-based models achieved better classification performance than PCA-based models. In addition, the peak performance (92.5%) of the VIP-RF model had a higher classification rate than the VIP-PNN model (90%). In conclusion, odor fingerprint analysis using a feature mining method based on the olfactory sensory evaluation can be applied to monitor product quality in the actual process of industrialization.