Prevalence, associated risk factors and satellite imagery analysis in predicting soil-transmitted helminth infection in Nakhon Si Thammarat Province, Thailand

Abstract

Soil-transmitted helminth (STH) infections remain a significant public health concern in rural areas, often leading to nutritional and physical impairment, particularly in children. This study aimed to assess the prevalence and associated factors of STH infections among schoolchildren in Thasala District, Nakhon Si Thammarat Province, Thailand, and to develop a predictive model for identifying high-risk areas using satellite imagery data. A cross-sectional study was conducted with 319 primary schoolchildren from six sub-districts in Thasala District. Stool samples were analyzed for STH infections using the formalin ethyl acetate concentration technique (FECT) and agar plate culture (APC), while behavioral data were collected through questionnaires to identify key risk factors. We developed an innovative predictive model by integrating convolutional neural networks (CNNs) for land-use classification of satellite imagery with artificial neural networks (ANNs) following dimensionality reduction through principal component analysis (PCA). The STH infections were detected in 31 samples (9.72%), with higher prevalence in males (11.38%) than females (8.67%). Mono-infections predominated, with Trichuris trichiura (5.02%) and hookworm (3.49%) being the most frequent. Mixed infections accounted for 1.25%, primarily co-infections of hookworm with T. trichiura (0.94%) or Strongyloides stercoralis (0.31%). Not cutting nails was identified as a significant behavioral factor associated with STH infections (p = 0.047), while other behavioral factors showed no statistical significance. From the satellite imagery analysis, specific environmental features, particularly higher proportions of agricultural land and closer proximity to water bodies, were positively associated with elevated STH prevalence. The modelling approach generated spatial risk maps for STH infections, providing a cost-effective tool for identifying high-risk transmission zones. These findings highlight that STH infections persist among rural Thai schoolchildren, with poor hygiene practices as a contributing factor. Strengthening hygiene education, improving sanitation, and implementing targeted environmental interventions are essential for effective control.

Keywords: Predictive modeling; Prevalence; Risk factor; Satellite imagery analysis; Soil-transmitted helminth; Thailand.

Fig. 1

The study area, which includes six sub-districts in Thasala District, Nakhon Si Thammarat Province: (1) Taling Chan, (2) Klai, (3) Thai Buri, (4) Thasala, (5) Moklan, and (6) Don Tako. Map was modified from Wikipedia Commons: https://en.wikipedia.org/wiki/Tha_Sala_district and ArcGIS software by Esri. Sources: Esri, TomTom, Garmin, FAO, NOAA, USGS, OpenStreetMap contributors, and the GIS User Community. For more information about Esri software, please visit http://www.esri.com. All other layers were produced by the authors and are copyright-free.

Fig. 2

Four-stage satellite imagery analysis for STH transmission risk mapping. Our methodology includes: (a) CNN model development using UC Merced Land Use Dataset with image preprocessing; (b) Application of trained model to classify land-use patterns surrounding six schools in Thasala District; (c) Construction of a predictive model integrating parasitological diagnosis results with school environmental parameters; (d) Development of a geographic risk prediction system that generates spatial risk distribution visualizations based on coordinates.

Fig. 3

Representative satellite imagery of six school campuses across different sub-districts in Thasala District used for predictive STH infection risk modeling: (a) Thai Buri, (b) Thasala, (c) Taling Chan, (d) Klai, (e) Moklan, and (f) Don Tako sub-districts. The satellite imagery was obtained and visualized using Folium version 0.19.5 (https://python-visualization.github.io/folium/) in the Python programming environment.

Fig. 4

Dimensionality reduction and model training performance for STH infection risk prediction. (a) Cumulative explained variance ratio against number of principal components, showing that 4 components capture 95% of total variance. (b) Training loss curve of the ANN model demonstrating rapid convergence within 20 epochs.

Fig. 5

Comparative heat maps of STH prevalence with white circles indicating geographic locations on latitude-longitude coordinates. (a) Observed STH prevalence based on field data from six sub-districts in Thasala District: (1) Taling Chan, (2) Klai, (3) Thai Buri, (4) Thasala, (5) Moklan, and (6) Don Tako. (b) Predicted STH prevalence generated by the CNN-PCA-ANN model incorporating environmental and spatial features. The color gradient ranges from blue (low prevalence) to red (high prevalence), with the color bar showing prevalence percentages for direct comparison between observed and predicted patterns.