Lung cancer is one of the leading causes of cancer worldwide. Early diagnosis and treatment are critical to improving patient survival. However, traditional diagnostic methods, such as histopathological evaluation of tissue biopsies, can be misleading and time-consuming. In recent years, convolutional neural networks (CNN) have become an effective method for diagnosing lung cancer. CNNs are deep learning models that learn to recognize and classify patterns in images. In this study, we used transfer learning to develop a pre-trained Google InceptionV3 CNN model to classify lung cancer from CT scan images. We reviewed and tested two different versions of the InceptionV3 model: a base model without weights and a model with some weights. The model with updated weights achieved the best performance with 97.81% training accuracy and 96.00% precision. This demonstrates that adaptive learning can be used to improve the accuracy and reliability of CNN-based lung cancer diagnosis with small datasets. The results of this study show that CNN-based lung cancer diagnosis can improve the early detection and treatment of cancer, thereby improving patients’ outcomes. However, more research is needed to confirm these findings on larger data sets and to create better and more comprehensive models.