Unleashing the Power of Deep Convolutional Neural Networks in a Stacked Ensemble Model for Precise Paediatric Pneumonia Diagnosis from Chest Radiographs

Pneumonia is a global health concern caused by various pathogens and its most prevalent impact is in Sub-Saharan Asia and South Asia. The Reports indicate that highest mortality is under the age of 5 years. The mortality specially in developing countries is attributed to delayed diagnosis and cost involved in conducting the tests. Chest X-ray has become the preferable diagnosis tool for pneumonia detection. But the doctor patient ratio in developing nations is also of concern. Automated diagnosis to detect the disease will aid in addressing these challenges. Deep learning based automated techniquesto detect pneumonia from Chest X-ray images have become an important advent in the field. However, the accuracy of these techniques is still in the exploration phase. We have proposed a stacked ensemble learning classifier using transfer learning to detect pneumonia. Our designed classified is based on DenseNet201, MobileNetV2 and InceptionResNetV2 pretrained classifiers. The accuracy of the pretrained individual model is 91%, 90% and 90% respectively but our designed stacked classifier is with 94%. The performance gain suggests its promising application in the field.