ABSTRACT
Medical imaging techniques aim to detect and track abnormalities in the human body. These strategies are essential to maintain viable evaluating, diagnosing, and treating of chest ailments. Chest radiography is considered the most affordable medical imaging technology to diagnose patients. Unfortunately, the lack of competent radiologists has severely restricted the technique's usage. Advanced technologies, such as artificial intelligence and deep learning, must be employed to detect abnormal chest X-rays to improve performance and diagnosis accuracy. Deep learning algorithms based on convolutional neural networks (CNNs) have come a long way. CNN performance in image class prediction has prompted academics to look at its feasibility as a diagnostic approach or tool in the identification and recognition of lung malignancy. In this chapter, we propose a very robust pipeline for detecting eight lung legions with an area under the curve (AUC) value between 0.835 and 0.963 and a sensitivity above 0.8 for 15 epochs of training and a weighted cross-entropy loss that minimize a significant load of the unbalanced combination of three datasets.
