Lung Cancer Detection: The Role of Transfer Learning in Medical Imaging

Lung cancer is a major global health issue, and improved patient results and treatment efficacy rely on early diagnosis. For many human diseases, including lung cancer, computer-aided detection (CAD) systems have greatly improved clinical analysis and decision-making. This study applies pre-trained deep learning techniques with a focus on transfer learning approaches and utilizes the available computed tomography (CT) scan image dataset (IQ-OTH/NCCD) to investigate the automation of lung cancer detection through CT scans. It concentrates on the training and evaluation of VGG16, ResNet50V2, and InceptionV3 architecture using this dataset. The aim is to automate lung cancer identification, ultimately enhancing diagnostic precision and outcome for the patient. It concentrates on the training and evaluation of VGG16, ResNet50V2, and InceptionV3 architecture using this dataset. The aim is to automate lung cancer identification, ultimately improving the accuracy of diagnosis and prognosis of the patient. The findings revealed ResNet50V2 is superior with impressive 98% accuracy, high sensitivity (95<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) and excellent specification (99<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>). This investigation highlights how CAD systems can greatly improve the diagnostic process by providing radiologists with reliable tools to identify cancerous nodules, thereby reducing the risk of missed diagnosis.