<p>Respiratory cancer, the second prevalent contributor to cancer-related fatalities, necessitates timely diagnosis for improved patient outcomes. Nevertheless, there are drawbacks to the current approaches, such as computational complexity, limited robustness, data loss, and problems with data availability. To mitigate the drawbacks of the previous approaches, this research presents the Distributed Attention-Based Pyramid Convolutional Neural Network (DisAPCN) model to achieve effective lung cancer detection. The proposed approach incorporates the Zero attention-based U-Net (ZU-Net) for segmenting the affected regions of the lungs. The ZU-Net effectively upgrades the ability of the model to attain precision outcomes with the reduction of numerous parameters that lead to minimizing the resources and time. Further, the deepened network of DisAPCN enables the model to diminish its computation at each layer and its size results in a pyramidical shape. Specifically, the DisAPCN model effectively captures the features of varied scales with progressive downsampling which enhances the training capability. Furthermore, the edge features along with the pre-trained models mitigate the vanishing gradient problems as well as enhance the framework’s detection performance. The evaluation findings reveal the dominance of DisAPCN model methods achieving a 95.52% F1 score, 95.40% precision, and 95.64% recall which is improved over the other traditional methods.</p> Graphical Abstract <p></p>