Convolutional Neural Networks (CNNs) is a classic and powerful model widely used in computer vision. This paper presents a novel CNN architecture inspired by the information processing mechanisms of biological neurons, applied to skin cancer detection. Specifically, the new model is inspired by a creative analogy: applying the concepts of “convergence” and “divergence” from neural systems to CNNs. We observed that convolutional layers exhibit both convergent and divergent properties, while pooling layers mainly reflect convergence and lack a mechanism for divergence. Based on this, we propose an adaptive multi-pooling CNNs to enhance feature diversity. This study was grounded in the task of automatic classification of skin lesions, which aimed to provide a practical application for validating the effectiveness of the proposed model. Experimental results show that it achieves modest performance improvements over traditional CNNs. More importantly, this study aims to highlight that the value of scientific research lies not only in performance gains but also in its ability to inspire new ideas and meaningful innovation. This work displays a journey from AI users to learners and to creators, demonstrating the value of interdisciplinary thinking in AI research and education.