Deep Learning-based Comorbidity Detection in Children: A Prophylactic Tool for Viral Diseases

Viral diseases continue to have a devastating effect on the health and the economy of the globe. During the recent pandemic, a substantial percentage of young people suffered from a serious infection. An effective way to reduce serious infections and fatalities among the younger population is to screen for deficiency diseases that have been observed as comorbidities, especially in developing economies. We present an approach to comorbidity detection taking advantage of wide availability of mobile X-ray units. These units facilitate massive screening and when combined with contemporary artificial intelligence-based methods, help analyse and detect the diseases quickly. We propose deep learning-based strategies for detecting comorbidities from X-ray images. In particular, we present custom convolutional neural network-based designs for the detection of two deficiency diseases, namely the Plummer-Vinson Syndrome (PVS) and Rickets. Implementation of the designs show that a Rickets knee is classified with 98% accuracy while a detection accuracy of 95% is achieved for PVS. The proposed networks outperform generic benchmarks in terms of accuracy. Further, the proposed networks process the X-ray images within a second. The results are valuable for quick and reliable identification of children for prioritised vaccination and enhanced nutritional support.