Application of Kohonen neural network for evaluation of the contamination of Brazilian breast milk with polychlorinated biphenyls

Artificial intelligence applied to the study of human milk and breastfeeding: a scoping review

BACKGROUND

Breastfeeding rates remain below the globally recommended levels, a situation associated with higher infant and neonatal mortality rates. The implementation of artificial intelligence (AI) could help improve and increase breastfeeding rates. This study aimed to identify and synthesize the current information on the use of AI in the analysis of human milk and breastfeeding.

METHODS

A scoping review was conducted according to the PRISMA Extension for Scoping Reviews guidelines. The literature search, performed in December 2023, used predetermined keywords from the PubMed, Scopus, LILACS, and WoS databases. Observational and qualitative studies evaluating AI in the analysis of breastfeeding patterns and human milk composition have been conducted. A thematic analysis was employed to categorize and synthesize the data.

RESULTS

Nineteen studies were included. The primary AI approaches were machine learning, neural networks, and chatbot development. The thematic analysis revealed five major categories: 1. Prediction of exclusive breastfeeding patterns: AI models, such as decision trees and machine learning algorithms, identify factors influencing breastfeeding practices, including maternal experience, hospital policies, and social determinants, highlighting actionable predictors for intervention. 2. Analysis of macronutrients in human milk: AI predicted fat, protein, and nutrient content with high accuracy, improving the operational efficiency of milk banks and nutritional assessments. 3. Education and support for breastfeeding mothers: AI-driven chatbots address breastfeeding concerns, debunked myths, and connect mothers to milk donation programs, demonstrating high engagement and satisfaction rates. 4. Detection and transmission of drugs in breast milk: AI techniques, including neural networks and predictive models, identified drug transfer rates and assessed pharmacological risks during lactation. 5. Identification of environmental contaminants in milk: AI models predict exposure to contaminants, such as polychlorinated biphenyls, based on maternal and environmental factors, aiding in risk assessment.

CONCLUSION

AI-based models have shown the potential to increase breastfeeding rates by identifying high-risk populations and providing tailored support. Additionally, AI has enabled a more precise analysis of human milk composition, drug transfer, and contaminant detection, offering significant insights into lactation science and maternal-infant health. These findings suggest that AI can promote breastfeeding, improve milk safety, and enhance infant nutrition.

Neural network for evaluation of the impact of the UK COVID-19 national lockdown on atmospheric concentrations of PAHs and PBDEs

The impact of measures to restrict population mobility during the COVID-19 pandemic on atmospheric concentrations of polycyclic aromatic hydrocarbons (PAH) and brominated flame retardants (BFRs) is poorly understood. This study analyses the effects of meteorological parameters and mobility restrictions during the COVID-19 pandemic on concentrations of PAH and BFRs at the University of Birmingham in the UK utilising a neural network (self-organising maps, SOM). Air sampling was performed using Polyurethane Foam (PUF) disk passive samplers between October 2019 and January 2021. Data on concentrations of PAH and BFRs were analysed using SOM and Spearman's rank correlation. Data on meteorological parameters (air temperature, wind, and relative humidity) and mobility restrictions during the pandemic were included in the analysis. Decabromodiphenyl ether (BDE-209) was the most abundant polybrominated diphenyl ether (PBDE) (23-91% Σ7PBDEs) but was detected at lower absolute concentrations (4.2-35.0 pg m-3) than in previous investigations in Birmingham. Air samples were clustered in five groups based on SOM analysis and the effects of meteorology and pandemic-related restrictions on population mobility could be visualised. Concentrations of most PAH decreased during the early stages of the pandemic when mobility was most restricted. SOM analysis also helped to identify the important influence of wind speed on contaminant concentrations, contributing to reduce the concentration of all analysed pollutants. In contrast, concentrations of most PBDEs remained similar or increased during the first COVID-19 lockdown which was attributed to their primarily indoor sources that were either unaffected or increased during lockdown.

Distinct weather conditions and human mobility impacts on the SARS-CoV-2 outbreak in Colombia: Application of an artificial neural network approach

The coronavirus disease 2019 (COVID-19) is still spreading fast in several tropical countries after more than one year of pandemic. In this scenario, the effects of weather conditions that can influence the spread of the virus are not clearly understood. This study aimed to analyse the influence of meteorological (temperature, wind speed, humidity and specific enthalpy) and human mobility variables in six cities (Barranquilla, Bogota, Cali, Cartagena, Leticia and Medellin) from different biomes in Colombia on the coronavirus dissemination from March 25, 2020, to January 15, 2021. Rank correlation tests and a neural network named self-organising map (SOM) were used to investigate similarities in the dynamics of the disease in the cities and check possible relationships among the variables. Two periods were analysed (quarantine and post-quarantine) for all cities together and individually. The data were classified in seven groups based on city, date and biome using SOM. The virus transmission was most affected by mobility variables, especially in the post-quarantine. The meteorological variables presented different behaviours on the virus transmission in different biogeographical regions. The wind speed was one of the factors connected with the highest contamination rate recorded in Leticia. The highest new daily cases were recorded in Bogota where cold/dry conditions (average temperature <14 °C and absolute humidity >9 g/m³) favoured the contagions. In contrast, Barranquilla, Cartagena and Leticia presented an opposite trend, especially with the absolute humidity >22 g/m³. The results support the implementation of better local control measures based on the particularities of tropical regions.

Comparison of Chemometric Problems in Food Analysis Using Non-Linear Methods

Food analysis is a challenging analytical problem, often addressed using sophisticated laboratory methods that produce large data sets. Linear and non-linear multivariate methods can be used to process these types of datasets and to answer questions such as whether product origin is accurately labeled or whether a product is safe to eat. In this review, we present the application of non-linear methods such as artificial neural networks, support vector machines, self-organizing maps, and multi-layer artificial neural networks in the field of chemometrics related to food analysis. We discuss criteria to determine when non-linear methods are better suited for use instead of traditional methods. The principles of algorithms are described, and examples are presented for solving the problems of exploratory analysis, classification, and prediction.

Effects of Fe(III) and quality of humic substances on As(V) distribution in freshwater: Use of ultrafiltration and Kohonen neural network

Humic substances (HS) are ubiquitous organic compounds able to affect mobility and availability of arsenic (As) in aquatic systems. Although it is known that associations between HS and As occur mainly via iron (Fe)-cationic bridges, the behaviour and distribution of this metalloid in HS- and Fe-rich environments is still not fully understood. In this paper, the quality of HS from different rivers in Brazil and Germany and its influence on the behaviour of As(V) under different Fe(III) concentrations were investigated. HS were extracted from four different rivers (Cascatinha, Holtemme, Selke and Warme Bode), characterised and fractionated into different molecular weight sizes (10, 5 and 1 kDa). Complexation tests were performed using an ultrafiltration system and 1 kDa membranes. All data was analysed using the Kohonen neural network (SOM - Self organising maps). All samples, except Selke, exhibited similar results of free As (<1 kDa). The results suggested that associations between HS, Fe and As were dependent on nitrogen (N)-aromatic carbon (C), amount of sulphur (S) and the molecular size of the HS. Although all HS appeared to be similar after looking at most variables analysed, the SOM could discriminate them into three different groups. Characterisation of the HS indicated that they had terrestrial material (from C₃ plants) as precursor material. Most of the As and Fe was distributed in the fractions of higher (>10 kDa) and lower (<1 kDa) size. HS quality is an important factor to take into account when studying the behaviour of As in HS-rich environments.

Use of self-organizing maps for analyzing the behavior of canines displaced towards midline under interceptive treatment

Background

Displaced maxillary permanent canine is one of the more frequent findings in canine eruption process and it’s easy to be outlined and early diagnosed by means of x-ray images. Late diagnosis frequently needs surgery to rescue the impacted permanent canine. In many cases, interceptive treatment to redirect canine eruption is needed. However, some patients treated by interceptive means end up requiring fenestration to orthodontically guide the canine to its normal occlusal position. It would be interesting, therefore, to discover the dental characteristics of patients who will need additional surgical treatment to interceptive treatment.

Material and Methods

To study the dental characteristics associated with canine impaction, conventional statistics have traditionally been used. This approach, although serving to illustrate many features of this problem, has not provided a satisfactory response or not provided an overall idea of the characteristics of these types of patients, each one of them with their own particular set of variables. Faced with this situation, and in order to analyze the problem of impaction despite interceptive treatment, we have used an alternative method for representing the variables that have an influence on this syndrome. This method is known as Self-Organizing Maps (SOM), a method used for analyzing problems with multiple variables.

Results

We analyzed 78 patients with a PMC angulation higher than 100º. All of them were subject to interceptive treatment and in 21 cases it was necessary to undertake the above-mentioned fenestration to achieve the final eruption of the canine.

Conclusions

In this study, we describe the process of debugging variables and selecting the appropriate number of cells in SOM so as to adequately visualize the problem posed and the dental characteristics of patients with regard to a greater or lesser probability of the need for fenestration.

Key words:Interceptive orthodontic treatment, altered eruption, impacted canines, neuronal networks, self-organizing maps.

Distribution and bioavailability of arsenic in natural waters of a mining area studied by ultrafiltration and diffusive gradients in thin films

The distribution of metals and metalloids among particulate, dissolved, colloidal, free, and labile forms in natural waters is of great environmental concern since it determines their transportation behaviour and bioavailability. Organic matter can have an important role for this distribution process, since it is an important complexing agent and ubiquitous in the aquatic environment. We studied the distribution, mobility and bioavailability of Al, As and Fe in natural waters of a mining area (Quadrilátero Ferrífero, Brazil) and the influence of organic matter in these processes. Water samples were taken from 12 points during the dry and rainy seasons, filtrated at 0.45 μm and ultrafiltrated (<1 kDa) to separate the particulate, colloidal and free fractions. Diffusive gradients in thin films (DGT) were deployed at 5 sampling points to study the labile part of the elements. Total and dissolved organic carbon and the physicochemical parameters were measured along with the sampling. The results of ultrafiltration (UF) and DGT were compared. The relationship among the variables was studied through multivariate analysis (Kohonen neural network), which showed that the seasonality did not impact most of the samples. Fe and Al occurred mainly in the particulate fraction whereas As appeared more in the free fraction. Most of the dissolved Fe and Al were inert (colloidal form) while As was more labile and bioavailable. The results showed that sampling points with a higher quantity of complexed Fe (colloidal fraction) showed less labile As, which may indicate formation of ternary complexes among organic matter, As and Fe.