Statistically extracted features of cellular morphologies clearly indicated that their information content can satisfactorily train computational models, not only to quantitatively evaluate current cellular status, but also to quantitatively forecast their future status, i.e., their potentials. The greatest advantage of our proposed morphology-based cell quality assessment is its non-invasiveness. As a result of this feature, our method has benefits that cannot be achieved by conventional techniques for producing cells for clinical regenerative medicine: elimination of risk factors, e.g., contamination and mishandling by the operator; synchronic and flexible scheduling of culture and clinical operations, for the best timing of cellular activity; and repeated assessment of the same sample, by multiple criteria and at multiple times, yielding data that better reflects the complex and dynamic features of the Vismodegib samples. Such intelligent control of culture processes is also a key technology for process automation. In this work, we expanded our previous SB203580 customer reviews efforts to predict singlelineage differentiation potentials by pursuing five important aims: Confirmation of the robustness of our method for adapting to the practical cellular variation. In our earlier work, 
it was not clear whether our original methodology was applicable to wider ranges of cellular variations. To investigate this issue, our data were expanded to cover eight continuous passages, ranging from very recently derived cells to those that had completely lost their doubling potential. Since a computational modeling solution for adapting to cellular variations resulting from patient diversity was already proposed in our previous work, our experimental design in this work was focused on cellular variations affected by culture processes, because these are the most difficult aspect of stem cells to evaluate daily. Investigation of the possibility of shifting the prediction timing to the very early stage. Our previous prediction required 2 weeks of image acquisition after the differentiation process began. In this study, however, we investigated whether much earlier and shorter periods were possible. In this work, only four images, obtained from the same sample repeatedly with a 24-hour interval during the first 4 days of expansion before differentiation culture, were used in the predictions. Multiplication of the variations of in silico predictions. Compared to the previous prediction scheme, which could predict osteogenic differentiation potential from the same image, in this study we attempted to predict four types of potentials ) from the same image. Such simultaneous prediction of multiple potentials for the same cells can be achieved by processing the same image data, although the predictions are performed by four types of differently trained prediction models running in parallel. Thus, this is a trial of ”overlapping” computational evaluation that can compensate for multiple immunohistochemical staining. Establishment of new conversion schemes of morphological feature usage that can achieve high predictive performance. Morphological features are the essential information generated from imaging data, and use of this information is critical in imaging-based applications. To date, however, there have been few comprehensive studies that compare the effects of different conversions of morphological features.