Abstract Human embryonic stem cells (hESCs) and pluripotent stem cells (iPSCs)-based disease modelings are potential platforms for cancer research and development of new cancer therapeutics. Differentiation of stem cells is an essential step for those disease models. For tissue-specific differentiation, hESCs or iPSCs are cultured in specific receipts of induction and differentiation media for developing different types of tissue cells such as muscle, skin, and liver providing further study or clinical applications. During lineage differentiation, researcher needs to closely monitor stem cell differentiation to be on track via checking cell morphological changes under microscope since this procedure has high probability of failed differentiation results (e.g., no differentiation or differentiating unwanted tissue types). However, monitoring via microscopy is labor-intensive and time-consuming, and also has high inter-observer variation issues. Therefore, it significantly impedes the progression of stem cell research and clinical applications nowadays. In recent years, machine learning has shown promising results in many applications of artificial intelligence (AI) in different fields, especially computer vision and image analysis. AI-based computational tool will bring benefits like high-throughput, high accuracy, and reproductivity in many medical applications. In stem cell culture and differentiation, we believe that applying this new technology will help researcher detect abnormal stem cell differentiation at the early stage via microscopy to save time, labor, and cost for the study and aggregate reproducible data along the process. To this end, we developed a machine learning-based AI model to assist in monitoring morphological changes of hESCs culture in bright-field microscopy images obtained from different differentiation stages to mature hepatocytes. We conducted a pilot study to train an AI model estimating efficiency of stem cell differentiation at Hepatic Progenitor Cell (HPC) stage, which is a critical checkpoint for hepatocyte differentiation. To prepare datasets for training, experienced researchers annotated the morphology of HPC in hundreds of microscope images and determined a differentiation result (success/fail) for every image. During the model training, the initial model was first trained by a training dataset consisting of 341 success and 366 fail HPC results. Subsequently, a smaller separate dataset comprising of 86 success and 51 fail HPC results was then used for cross-validation. Finally, the test set containing 64 success and 29 fail HPC results was used to evaluate the AI model performance. In result, the AI model presented an excellent performance (accuracy= 0.978 and F1 score= 0.975). Our study suggests a potential application of AI-assisted monitoring model for stem cell culture and differentiation in the future. Citation Format: Wei-Lei Yang, Zijun Huo, ShihYu Chen, Dandan Zhu, Tien-Jen Liu, Dung-Fang Lee. Monitoring of stem cell differentiation to mature hepatocytes with a machine learning-based AI model [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 185.