A hybrid chromaticity-morphological machine learning model to overcome the limit of detecting newcastle disease in experimentally infected chicken within 36 h

Abstract

The nexus between animal and human health is crucial in upholding global health. Humans are at risk of food security due to fatal infections associated with the Newcastle disease virus (NDV), resulting in severe disease outbreaks. This work reports on the early detection of experimentally NDV-infected chickens to prevent such catastrophic events. Image processing techniques were employed to extract the chromaticity and morphological features of the chicken comb and standing posture. The changes in these features across different stages of symptom severity, indicated by the post-infection period in hours, were examined through statistical and Spearman coefficient correlation analysis. Various hybrid chromaticity-morphology machine learning (HCMML) classifier models, including Logistic Regression, Support Vector Machine (SVM) with different kernels, K-Nearest Neighbour (KNN), Decision Tree, and Artificial Neural Network (ANN), were trained using selected feature variables and different variation of datasets to detect infected chickens. The statistical analysis on individual features demonstrates the necessity of HCMML models to predict infected chicken with a reasonably high accuracy. Based on the coefficient correlation analysis, the chromaticity features demonstrate a higher correlation to the chickens with NDV infection than the morphological features. These findings highlight the importance of extracting chromaticity features in predicting infected chicken, especially at the early phase of infection. Based on the HCMML models result, SVM with Polynomial kernel achieved a test accuracy of 82·39 % with 79·00 % validation accuracy at 36 h post-infection after feature optimization and > 95·00 % test accuracy after 96 h post-infection. This work demonstrates a promising methodology in developing machine learning algorithm using hybrid chromaticity-morphological features for early detection of virus-infected chickens, contributing to the goal of a sustainable and healthier planet.