An Efficient Deep Learning based Approach for Pneumonia Detection

Pneumonia is a major global cause of death, hence early detection and precise diagnosis are essential for successful management. Conventional techniques for detecting pneumonia, like radiological interpretation and clinical assessment, are frequently laborious and prone to variation amongst observers. Deep learning models have transformed medical imaging by offering automated, reliable, and accurate diagnostic instruments. The goal of this paper is to use data (image) pre-processing techniques to increase the accuracy of the deep learning model, EfficientNetB0 for pneumonia identification. The dataset utilized in this paper consists of annotated chest X-ray images from the publically accessible source “Chest X-Ray Images (Pneumonia)”; Paul Mooney submitted 5,863 photos in 2 categories on Kaggle. To enhance the model's resilience and ability to generalize, the images undergo preprocessing using techniques such as normalization, gray scaling, median filtering, CLAHE, and data augmentation. A standardized evaluation framework is used to evaluate each model's performance, and measures like Fl score and confusion matrix analysis are used. The results of this paper demonstrate how pre-processing the data can increase the accuracy of the deep learning model and can enhance medical imaging diagnostic accuracy and dependability.