Lung cancer is one of the major causes of cancer-related mortality worldwide, making early identification and reliable assessment of suspicious lung abnormalities essential. Conventional computed tomography image interpretation requires specialized expertise and may be affected by variations in lesion appearance and image quality. This paper proposes an intelligent healthcare decision-support framework for explainable lung cancer screening and risk assessment using deep learning techniques. The proposed framework employs a convolutional neural network, such as ResNet50 or EfficientNet, to extract meaningful spatial and textural features from thoracic CT images. A multiscale feature fusion mechanism is incorporated to capture lesion characteristics at different spatial resolutions, improving the representation of suspicious abnormalities. Furthermore, Grad-CAM-based visual explanations identify image regions that influence model predictions, while uncertainty estimation provides additional information regarding prediction reliability. The framework will be evaluated using an appropriately annotated CT imaging dataset and compared with conventional CNN-based baseline models. Performance assessment will consider classification metrics, including accuracy, precision, recall, F1-score, and calibration measures. The proposed approach aims to improve model transparency and provide interpretable decision-support information for healthcare professionals, subject to further clinical validation before practical medical deployment.

Keywords: Lung Cancer Screening, Computed Tomography, Deep Learning, Multiscale Feature Fusion, Grad-CAM, Explainable Artificial Intelligence .