Noninvasive diagnosis and treatment of disease, which can greatly reduce the pain of patients and improve the efficiency of diagnosis, is increasingly advocated in modern health care. Traditional endoscopy as an important clinical diagnostic tool can provide only rough information on the structure of the tissue. In order to obtain morphological information at the cellular level, biopsy is often required. However, biopsy brings suffering to the patient and is costly. In recent years, the confocal endoscopic imaging system has been developed, yet has not been widely applied owning to the small imaging field of view. In view of these problems, we designed an endoscopic microscopic imaging system based on the combination of macroscopic vision and microscopic resolution. Here, we describe the design principle of the system. We further demonstrate the feasibility of the system by performing automated cell counts in an in vivo subcutaneous cancer model.