Lung nodule early classification plays a significant role in lung cancer prediction. A nodule in the lung is very prone to cancer, early detection of malignant nodules can increase the patient’s recovery rate. Manual process and statistical machine learning approaches detect the malignant nodule based on the size and color of a nodule, but not in the early stages. It will take months and years to confirm the malignant. Early detection of malignant nodules at the early stage of cancer will increase the recovery of human life. A novel approach is proposed with advanced deep learning on CT images to detect whether a lung nodule is malignant, and when the malignant is small, and the nodules are overlapped; vision transformers is implemented to segment the image and identify the nodule pattern deeply. The proposed approach can identify malignant nodules irrespective of size and color and achieves optimal results compared to other prescribed models with an accuracy of 87.2. The analysis also presented true positive and false negative results.