Cancer stands at the top position in crucial problems that causes many problems to all, By implementing various methods with the latest technology we can accurately identify this cancer easily. This study presents a novel dual-model approach that integrates convolutional autoencoders to extract features and uses deep architecture to classify images based on extracted features of lung cancer detection. The proposed method begins with a convolutional autoencoder, which effectively compresses and reconstructs input data, thereby extracting relevant features from medical imaging datasets. These images which are extracted are given to DNN to train the model whether to tell about cancerous and non-cancerous cases. The LC25000 dataset was evaluated with an accuracy rating from the proposed model at 98%. Also, the proposed methodology is assessed on the basis of accuracy, precision, recall, and F1 score. The results obtained do show that lung cancer detection by the proposed dual model is much better than the convolutional methods providing an effective way of detection at an early stage. This research underscores the unique approach of integrating feature extraction and classification techniques in medical diagnostics, which leads to new technologies and new algorithms to enhance the process of finding lunch cancer and other related applications.