Purpose/Significance To reveal the topic evolution trajectory of the application of deep learning technology in the field of medical imaging, and to provide decision-making basis for R&D resource allocation and industrial standard construction. Method/Process The “BERTopic+dynamic similarity” dual layer topic analysis framework is adopted to conduct deep semantic mining on Web of Science paper data and Derwent patent data from 2010 to 2024. Combined with the cosine similarity algorithm under the time sliding window mechanism, the fission and integration paths of technical topics are quantitatively analyzed. Result/Conclusion The technical topics show a three-stage evolution and form a “theory-engineering” dual-drive ecology, providing a new perspective for understanding the evolution of technical topics in the field of medical imaging.