Medical Artificial Intelligence (AI) is a next-generation medical technology that presents a diagnosis based on EBM regardless of the experience of clinical laboratory technologists. The technology is characterized by learning large amounts of patient data diagnosed by experts based on years of experience. In this study, we examined the clinical usefulness of screening technology with AI for peripheral leukocyte classification. The subjects were 57 healthy person's peripheral blood smears performed MG staining. The first convolutional neural network (CNN) model performed transfer learning with background trimmed training images of mature leukocyte cells that show typical morphology, and parameter tuning for optimization was performed. Then, we performed additional learning and fine-tuning on first CNN model with leukocyte images which include background cells, and the second CNN model was created. As a result of clinical data evaluation, the accuracy of five classification showed 0.990 and six classification showed 0.822 respectively in the first CNN model with background-less images. Contrast, the accuracy of five classification showed 0.992 and six classification showed 0.879 respectively in the second CNN model with leukocyte images which include background cells. It was cleared that the mature leukocyte cell morphology screening with CNN was highly accurate and useful. However, it is necessary to examine further the cutoff value and the judgement pending condition for the boundary area cells in the clinical application.