Lung cancer, a leading contributor to cancer-related fatalities globally, necessitates prompt and precise identification to enhance survival outcomes. This study explores the use of machine learning models, specifically a custom-built Convolutional Neural Network (CNN), MobileNetV2, and ResNet50, to detect lung cancer through CT scan images from the Iraq-Oncology Teaching Hospital/National Center for Cancer Diseases (IQ-OTH/NCCD). The dataset includes cases that are normal, benign, and malignant. Our custom-built CNN was compared to the pre-trained Mo-bileNetV2 and ResNet50 models, with adjustments made to enhance their effectiveness in lung cancer detection. The results are promising: the custom CNN achieved an impressive accuracy of 98.32%, recall of 97.48% and 98.78% specificity proving highly effective in making correct predictions and identifying nearly all relevant cases. While ResNet50 also performed well, with an accuracy of 91.24%, recall of 86.86% and specificity of 94.22%, it fell short of the custom-built CNN. MobileNetV2 demonstrated the lowest performance, achieving an accuracy of 88.43%, a recall of 82.65%, and a specificity of 92.78%, underscoring its limitations in providing precise and thorough predictions. This study underscores the potential of AI, particularly CNNs, in transforming lung cancer detection, the path to earlier detection and improved patient results. By integrating image enhancement, advanced network architectures, and transfer learning, our custom-built CNN aims not only to match but to exceed the diagnostic capabilities of medical professionals.