Enhancing Medical Image Analysis: Leveraging Transfer Learning in Convolutional Neural Networks for Tuberculosis Detection

Advancements in medical imaging have paved the way for image-based computer-aided diagnosis (CADx) systems, revolutionizing the detection of lung abnormalities, including lung nodules and diffuse lung diseases. This paper explores the potential of convolutional neural networks (CNNs) in image-based CADx without the reliance on image-feature extraction, presenting a powerful alternative to feature-based CADx methods. Additionally, we introduce an image-based computer-aided detection (CADe) algorithm employing regions with CNN features (R-CNN) for precise identification of lung abnormalities. We conduct comprehensive evaluations of our image-based CADx utilizing CNN and CADe employing R-CNN, demonstrating their effectiveness across various lung abnormalities.