FIGURE 1.

Schematic Illustration of the deep learning (DL) framework and deep neural network (DNN) training process used to identify mesenchymal stem cell differentiation. (A) Illustration of overall deep learning framework. Mesenchymal stem cells were acquired from four different donors. Bright field images of hMSCs with different treatments were obtained for classification. (B) Illustration of the process of deep neural network (DNN) training. The raw image data were initially obtained and divided into different datasets: training, validation, and testing sets on a ratio of 60:20:20. To increase the datasets, the images were cropped to increase the total number of datasets. Finally, the datasets were trained, tested, and validated using transfer learning.