This thesis explores the application of deep learning, specifically Convolutional Neural Networks (CNNs), for medical image processing, focusing on chest X-ray classification using the NIH Chest X-ray dataset. The study employs a DenseNet121 architecture, pre-trained on ImageNet, to classify 14 thoracic pathologies. Results indicate that the model struggles with class imbalance and achieves only moderate performance, with AUC scores ranging from 0.50 to 0.62. Training for 5 epochs prioritized sensitivity, while 11 epochs improved specificity but at the cost of recall. The findings highlight challenges in accurately diagnosing rare conditions and underscore the need for techniques like advanced augmentation or ensemble methods to enhance performance.