ConvMixer deep learning model for detection of pneumonia disease using chest X-ray images

Pneumonia is a common and fatal disease in children nowadays. It infects the lungs resulting in difficulties in breathing. Severe cases of it may lead to death. Therefore, early and accurate detection of pneumonia disease is essential. There are existing various methods for the detection of pneumonia disease today. Deep learning methods are considered more effective for this. We have applied a novel deep learning model i.e. the ConvMixer model for the detection of pneumonia disease. The model replaces traditional convolutional layers with a mixer of channels and spatial dimensions. By mixing channels and spatial dimensions, it reduces the number of parameters and computations required for processing each layer leading to improved efficiency. We have applied the model to the large numbers of chest X-ray images that are publicly available on Kaggle, provided by Guangzhou Women and Children’s Medical Centre in Guangzhou. The model has achieved the highest accuracy of 95.11%. It has also been evaluated for precision, recall, and f1-score parameters.