Ensuring diagnostic consistency in typical medical cases helps physicians select appropriate treatment plans based on patient records, avoiding discrepancies in treatment strategies caused by differing interpretations among doctors. This consistency ultimately ensures uniformity in medical services provided to patients. This paper proposes a case diagnosis method in the evaluation task of the 10th China Health Information Processing (CHIP 2024). This method fine-tunes and trains three models through the LoRA framework and performs voting. Furthermore, we incorporated a closed-source LLM to refine the results through an error-correction process. Our approach also features post-processing techniques to avoid the need for text alignment. Experimental results demonstrate that the proposed method achieves an accuracy of 0.96, underscoring its effectiveness and reliability.