Unveiling Diagnostic Precision: Evaluating Machine Learning and Deep Learning Approaches for Pneumonia Recognition of COVID-19 Patients Using Chest X-Rays

Machine Learning (ML) and Deep Learning (DL) have become indispensable for medical image analysis due to their capacity to extract meaningful information from large and complex medical image datasets. Currently, medical image datasets are increasing rapidly in size and complexity. Additionally, these algorithms are capable of processing and analyzing enormous amounts of data much more quickly and precisely than manual methods. However, it is challenging to determine which of these two approaches yields more accurate results. Therefore, this research attempt proposes a complete performance comparison of ML and DL methodologies for recognizing pneumonia from thorax X-ray or radiograph pictures of COVID-affected patients. For this research, a dataset consisting of 575 chest radiographs of different age groups of people affected by COVID-19 is used. To analyze the efficacy of the ML strategy, seven distinct machine learning algorithms, including K-Nearest Neighbor (KNN), Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), Naive Bayes (NB), RandomForest, and AdaBoost, are employed, followed by the best classifier being chosen using GridSearchCV. On the other hand, Convolutional Neural Network (CNN) dependent DL frameworks are designed for running an analysis of the deep learning approach. After a rigorous analysis of both approaches, it was found that the ML approach obtained 93% accuracy, whereas the DL approach acquired 99% accuracy. The observed elevation in accuracy by 6% exhibits a significant influence on the field of medical image analysis. Finally, the thorough performance study indicates that the DL approach exhibits excellent reliability and accuracy compared to the machine learning approach.