Pneumonia and brain tumors diagnosis using machine learning algorithms

Early diagnosis of life-threatening diseases, such as Pneumonia and Brain Tumors, is crucial for improving patient outcomes and reducing mortality rates. Additionally, accurate classification of disease types is essential for administering appropriate treatment, and in the rapidly evolving field of medical imaging, leveraging the power of Artificial Intelligence can significantly enhance diagnostic capabilities. This work developed an automatic disease detection system focusing on Pneumonia and Brain Tumors using a mix of techniques, utilizing pre-trained Convolutional Neural Network architectures with Deep Learning methods, Support Vector Machine, and custom ensemble models. The Brain Tumor dataset includes Normal, Glioma, Meningioma, and Pituitary Tumor MRIs, while the Pneumonia dataset consists of X-rays of Normal, COVID-19, Bacterial, and Viral Pneumonia images. This work presents multiple models, including a novel two-part pneumonia detection ensemble, combining pre-trained DenseNet201 CNN and CNN-SVM to achieve 91.97% accuracy. Additionally, multiple ensemble techniques were investigated for Brain Tumor detection, demonstrating different approaches’ performance and computational efficiency, with the highest accuracy of 98.65%. The work also introduces user-friendly software that accepts any image (including DICOM), automatically predicts the radiology image type, and diagnoses the disease accordingly. Therefore, the proposed study can be a valuable tool for radiologists, enabling quicker and more accurate diagnosis of Brain Tumors and Pneumonia. This chapter shows a couple of contributions to the field. State of the art accuracy for both Tumors and Pneumonia detection and proposal of Automatic Disease Detection tool are the most important contributions.