LUNG CANCER CLASSIFICATION BY USING TRANSFER LEARNING

Abstract

Lung cancer poses a formidable global health challenge, necessitating advancements in diagnostic methodologies. This research addresses the exigency for efficient and accurate classification of lung cancer histopathological images  by assessing the computationally streamlined ConvNeXtTiny model. The study meticulously examines the model's aptitude in differentiating amongst benign, squamous cell carcinoma, and adenocarcinoma, thereby obviating the constraints of computational intensity that encumber many contemporary deep learning architectures. Transfer learning is employed, leveraging the pre-trained ConvNeXtTiny model and refining it for lung cancer classification. The methodology encompasses a rigorous evaluation utilizing a curated dataset of histopathological images, with a focus on metrics pertinent to clinical applicability. The results evince the model's capacity to achieve high classification accuracy with reduced computational demands, thus underscoring its potential to facilitate the pragmatic deployment of AI in diverse medical facility settings, including those with inadequate resources. This study contributes to the persistent efforts to increase diagnostic precision and efficiency in lung cancer, offering a computationally viable solution that maintains clinical reliability. The research highlights the prospective of the ConvNeXtTiny model to serve as a valuable tool in assisting medical professionals in the accurate and timely diagnosis of lung cancer subtypes.