Towards a deeper understanding of pain: How machine learning and deep learning algorithms are needed to provide the next generation of pain medicine for use in the clinic.

As our definition of pain evolves, the factors implicit in defining and predicting pain status grow. These factors each have unique data characteristics and their outcomes each have unique target attributes. The clinical characterization of pain does not, as defined in the most recent IASP definition, require any tissue pathology, suggesting that the experience of pain can be uniquely psychological in nature. Predicting a persons pain status may be optimized through integration of multiple independent observations; however, how they are integrated has direct relevance towards predicting chronic pain development, clinical application, and research investigation. The current challenge is to find clinically-mindful ways of integrating clinical pain rating scales with neuroimaging of the peripheral and central nervous system with the biopsychocial environment and improving our capacity for diagnostic flexibility and knowledge translation through data modeling. This commentary addresses how our current knowledge of pain phenotypes and risk factors interacts with statistical models and how we can proceed forward in a clinically responsible way.