Machine-learning--based metabolic imaging: A bridge between $in vitro$ models and clinical applications.

Thanks to recent advances in the optical microscopy field, new methods for fine metabolic characterization of living systems have been developed. Moreover, the combination of molecular biology and $in vivo$ imaging gives rise to metabolic functional imaging, a discipline that enables the real-time fine monitoring of molecular changes and supra-molecular properties essential to cell survival, thus allowing the enhancement of the available biological informative content. However, some limits and uncertainties mainly regarding the management of large amounts of data as well as the lack of specificity, remain. In this perspective, we have developed machine-learning--based techniques that allow assessing different metabolic cellular processes, including, among others, autophagy, and lipid turnover. Here, we propose the application of ML-based metabolic imaging to evaluate and characterize cellular metabolic reprogramming occurring in several pathological conditions, focusing on diabetes-related complications and tumors.