Intelligent question answering systems based on knowledge graphs exhibit robust question understanding, high accuracy, and scalability. This paper explores the complexities of multi-hop queries in the context of knowledge graph question answering, with cardiovascular disease serving as a case study to examine reasoning methodologies for these intricacies. We propose a knowledge reasoning model, designated as Convolutional Neural Network Link Prediction (LPCNN), which is designed to address these challenges. LPCNN employs link prediction methods to achieve knowledge graph reasoning tasks. The core word and the question are mapped to the head entity and relation, then encode it. A convolutional neural network-based knowledge graph embedding model is employed for knowledge graph representation and candidate triple scoring. The highest-scoring tail entity is selected as the inference chain, thereby obtaining answers for corresponding hop counts. Experiments conducted on medical cardiovascular knowledge graph and public domain knowledge graph demonstrate that LPCNN can perform intelligent multi-hop reasoning for question answering. LPCNN exhibits certain advantages over other methods in reasoning for complex questions in medical cardiovascular knowledge graphs.