Recent studies in oncology have addressed the importance of detecting circulating tumor cell clusters because circulating tumor cell clusters might survive and metastasize more easily than single circulating tumor cells. Signals with larger peak widths detected by in vivo°ow cytometer (IVFC) have been used to identify cell clusters in previous studies. However, the accuracy of this criterion might be greatly degraded by variance in blood°ow and the rolling behaviors of circulating tumor cells. Here, we propose a criterion and algorithm to distinguish cell clusters from single cells. In this work, we¯rst used area-based and volume-based models for single°uorescent cells. Simulating each model, we analyzed the corresponding morphology of IVFC signals from cell clusters. According to the Rayleigh criterion, the valley between two adjacent peak signals from two distinguishable cells should be lower than 73.5% of the peak values. A novel signal processing algorithm for IVFC was developed based on this criterion. The results showed that cell clusters can be reliably identi¯ed using our proposed algorithm.