Continuous monitoring of multi-channel blood pressure during the perioperative period is crucial for predicting complications. ProtoPNet has garnered attention as powerful tools for providing interpretable support for complication prediction, enabling clinicians to better understand the impact of blood pressure trajectories on complications. However, existing prototype-based models for complication prediction tasks based on channel-wise blood pressure fail to recognize the unique characteristics of individual channels such as systolic pressure, diastolic pressure, and mean arterial pressure. Instead, they treat them as a unified entity for prediction and interpretation, resulting in performance degradation. To address this issue, we proposed the Channel-wise Prototypical Part Transformer (C-PPT). Firstly, we match the encoded data independently for each channel using a ProtoPNet, allowing for the extraction of unique features. Secondly, we enhance the prototypes by incorporating distance information between prototypes in the loss function. Finally, we optimize the influence of different channels on the results using pre-set weights. Experimental results conducted on a real dataset of perioperative blood pressure and cardiovascular adverse events classification tasks in a hospital setting demonstrate that our proposed method effectively interprets the relationship between blood pressure trajectories in different channels and the occurrence of cardiovascular adverse events, outperforming other benchmark models on relevant metrics.