Anaemia detection in infants of 6–59 months using medical images: a multi-architecture convolutional neural network-based approach

Abstract Anaemia has become one of the biggest burdens to public health globally. It affects billions of people, particularly in underprivileged communities in developing countries, due to inadequate diets, such as low intake of iron or vitamins. This deficiency affects the cognitive development and psychological growth of children. This study aims to detect anaemia in children aged 6–59 months using images of the conjunctiva of the eye, employing single convolutional neural network (CNN) architectures, including AlexNet, DenseNet, and VGGNet. These CNN architectures were combined to create a multi-architecture model by using the feature concatenation technique, which outperformed the individual architectures, achieving an accuracy of 98.79%. In comparison, AlexNet, VGGNet, and DenseNet achieved an accuracy of 96.76%, 93.45%, and 94.21%, respectively. The results of this study demonstrate that integrating an advanced CNN multiarchitecture for anaemia detection represents a significant advancement in the healthcare sector. It enhances the accuracy, efficiency, and accessibility of medical disease detection while being cost-effective and scalable, providing quicker results in healthcare screening compared to traditional methods.