Early cancer identification is essential for patients’ prognosis and survival rates to be improved. Due to their capacity to record precise anatomical details, computed tomography (CT) scans are often utilized for cancer screening and diagnosis. However, detecting subtle anomalies indicative of early-stage cancer in CT scans can be challenging for radiologists, often leading to delayed diagnoses. This study suggests a Deep Learning-Based Anomaly diagnosis method for CT scan-based early cancer diagnosis. We use deep convolutional neural networks’ strength to automatically learn and extract beneficial characteristics from CT scans, enabling the detection of slight abnormalities that might not be visible to human observers. The proposed model is trained on a carefully curated dataset extracted from the Cancer Imaging Archive, comprising middle slices from 475 CT series obtained from 69 patients. We assess the deep learning model's performance by utilizing various metrics, such as sensitivity, specificity, and area under the receiver operating characteristic curve (AUC-ROC). Our findings show that the suggested approach outperforms conventional computer-aided diagnosis methods in detecting early cancer abnormalities with high accuracy and sensitivity. The results of this study have the potential to substantially influence clinical practice by assisting radiologists in quickly identifying cases of early-stage cancer and enabling timely and focused therapies, eventually improving patient outcomes and survival rates.