<p>Emerging topics (ETs) detection plays a significant role in various research fields in the era of rapidly evolving innovation. This study proposes a four-stage detection framework to detect ETs. First, time-sliced term citation networks are constructed based on paper citation relationships and paper-term relationships. Second, we apply the concept of motifs and use the discovery algorithms along with two filtration procedures to extract significant motif types for subsequent analysis. Third, we extract critical knowledge units through motif analysis, use novel term pairs for fine-grained topic representation, and propose an evaluation mechanism to detect ETs. Finally, we applied feature analysis and term network analysis to analyze ETs within scientific fields. Empirical studies based on two datasets, brain neoplasms and cardiovascular abnormalities, containing 96,365 and 109,884 publications, were used to evaluate our framework. The proposed framework involves complex higher-order structures and dynamic relationships among topics, offering deeper insights into ETs and the evolution of scientific knowledge. Overall, this study enhances the theoretical framework for ETs detection, offering valuable insights into the evolution of disciplinary fields and innovation management.</p>