Quantitative image analysis of microplastics in bottled water using artificial intelligence

The ubiquitous occurrence of microplastics (MPs) in the environment and the use of plastics in packaging materials result in the presence of MPs in the food chain and exposure of consumers. Yet, no fully validated analytical method is available for microplastic (MP) quantification, thereby preventing the reliable estimation of the level of exposure and, ultimately, the assessment of the food safety risks associated with MP contamination. In this study, a novel approach is presented that exploits interactive artificial intelligence tools to enable automation of MP analysis. An integrated method for the analysis of MPs in bottled water based on Nile Red staining and fluorescent microscopy was developed and validated, featuring a partial interrogation of the filter and a fully automated image processing workflow based on a Random Forest classifier, thereby boosting the analysis speed. The image analysis provided particle count, size and size distribution of the MPs. From these data, a rough estimation of the mass of the individual MPs, and consequently of the MP mass concentration in the sample, could be obtained as well. Critical materials, method performance characteristics, and final applicability were studied in detail. The method showed to be highly sensitive in sizing MPs down to 10 μm, with a particle count limit of detection and quantification of 28 and 85 items/500 mL, respectively. Linearity of mass concentration determined between 10 ppb and 1.5 ppm showed a regression coefficient (R2) of 0.99. Method precision was demonstrated by a repeatability of 9-16% RSD (n = 7) and within-laboratory reproducibility of 15-27% RSD (n = 21). Accuracy based on recovery was 92 ± 15% and 98 ± 23% at a level of 0.1 and 1.0 ppm, respectively. The quantitative performance characteristics thus obtained complied with regulatory requirements. Finally, the method was successfully applied to the analysis of twenty commercial samples of bottled water, with and without gas and flavor additives, yielding results ranging from values below the limit of detection to 7237 (95% CI [6456, 8088]) items/500 mL.

Human scent characterization: A review

Human scent has long been cited as a probable parameter that can be exploited as a biometric measure. Identifying the scent of individual persons using specially trained canines is a well-known forensic method which is frequently used in criminal investigations. To date there has been limited research on the chemical components present in human scent and their usefulness in distinguishing between people. This review delivers insight into studies which have dealt with human scent in forensics. Sample collection methods, sample preparation, instrumental analysis, compounds identified in human scent and data analysis techniques are discussed. Methods for sample collection and preparation are presented, but to date, there is no available validated method. Instrumental methods are presented and from the overview it is clear that gas chromatography combined with mass spectrometry is the method of choice. New developments such as two-dimensional gas chromatography offer exiting possibilities to collect more information. Given the amount and complexity of data, data processing is used to extract the relevant information to discriminate people. Finally, sensors offer new opportunities for the characterization of human scent.

Evaluation of a Commercial Electronic Nose Based on Carbon Nanotube Chemiresistors

Recently a hand-held, carbon-nanotube-based electronic nose became available on the market. Such an electronic nose could be interesting for applications in the food industry, health monitoring, environmental monitoring, and security services. However, not much is known about the performance of such an electronic nose. In a series of measurements, the instrument was exposed to low ppm vapor concentrations of four volatile organic compounds with different scent profiles and polarities. Detection limits, linearity of response, repeatability, reproducibility, and scent patterns were determined. The results indicate detection limits in the range of 0.1-0.5 ppm and a linear signal response in the range of 0.5-8.0 ppm. The repeatability of the scent patterns at compound concentrations of 2 ppm allowed the identification of the tested volatiles based on their scent pattern. However, the reproducibility was not sufficient, since different scent profiles were produced on different measurement days. In addition, it was noted that the response of the instrument diminished over time (over several months) possibly by sensor poisoning. The latter two aspects limit the use of the current instrument and make future improvements necessary.

Experimental Flight Patterns Evaluation for a UAV-Based Air Pollutant Sensor

The use of drones in combination with remote sensors have displayed increasing interest over the last years due to its potential to automate monitoring processes. In this study, a novel approach of a small flying e-nose is proposed by assembling a set of AlphaSense electrochemical-sensors to a DJI Matrix 100 unmanned aerial vehicle (UAV). The system was tested on an outdoor field with a source of NO₂. Field tests were conducted in a 100 m² area on two dates with different wind speed levels varying from low (0.0–2.9m/s) to high (2.1–5.3m/s), two flight patterns zigzag and spiral and at three altitudes (3, 6 and 9 m). The objective of this study is to evaluate the sensors responsiveness and performance when subject to distinct flying conditions. A Wilcoxon rank-sum test showed significant difference between flight patterns only under High Wind conditions, with Spiral flights being slightly superior than Zigzag. With the aim of contributing to other studies in the same field, the data used in this analysis will be shared with the scientific community.