CNN Models Comparison for Lung Cancer Classification using CT and PET scans
K V Suma, C. S. Sonali, B S Chinmayi, John Kiran B, Muhammad Easa
Abstract
Lung cancer is the second most frequent cancer in the world, according to the World Health Organization (WHO). In 2020, there were about 2.21 million new cases of lung cancer. The ability to categorize lung cancer into distinct categories in its early stages may aid in the treatment and perhaps save millions of lives each year. PET and CT scans are used to diagnose lung cancer as well as to provide a thorough image of tumors in the body and to follow their progression. This research aims to give a more precise and prompt categorization of lung cancer. The most prevalent kind of lung cancer is adenocarcinoma, which is followed by squamous cell carcinoma. This study uses CT and PET scans to assess whether 3D CNN models (AlexNet, CNN-T5, Custom 1, Custom 2, LeNet, VGGNet) deliver greater accuracy. Experimenting with hyperparameters such as image sizes, batch sizes, and splits to increase the model’s efficiency is also required.