Modeling and identification of the nociceptive system using psychophysical measurements

Efficient treatment for pain disorders could benefit from improved diagnostic tools by identifying defective nociceptive processes. To develop such a tool, this thesis follows a system approach by building plausible computational models. To offer a mechanism-based interpretation, we perform mathematical analysis on these models. In particular, we use psychophysical measurements from human subjects to perform parameter estimation and identifiability analysis. We start with building neurophysiology-based models, representing both peripheral and central nociceptive subsystems for a nociceptive detection task using electrocutaneous stimulation. In addition, we incorporate psychophysical observations of detection thresholds for a qualitative validation. In turn, it provides explanatory insights into the underlying mechanisms. In particular, our analysis reveals that the observed dependence of thresholds on the interpulse interval of double-pulse stimuli could be attributed to specific properties in central nociceptive subsystems. Next, with the developed model, we investigate nociceptive neuroplasticity. In a case study using a topical capsaicin patch, simulation results agree with observations qualitatively, and suggest that capsaicin-induced central neuroplasticity can last over a month. For a more quantitative assessment of the nociceptive system, we address estimation and identifiability of system parameters with stimulus-response pairs measured from individual subjects. A necessary condition for structural identifiability on stimulus properties follows from our theoretical analysis. Our study demonstrates the applicability of profile likelihood to assess identifiability in a nonlinear physiology-based model using psychophysical measurements. Further applications of our approach might facilitate mechanism-based differential diagnosis of malfunctioning nociceptive processes. We formulate the staircase procedures utilized in the experimental detection task as Markov models. Both our theoretical analysis and simulations reveal the existence of a bias in the threshold estimate from logistic regression. This finding also suggests that in order to estimate system parameters, one should use measured stimulus-response pairs rather than intermediate threshold estimates. Lastly, with respect to computational modeling and analysis, we point out prospective research directions including modeling of other relevant nociceptive processes as well as possible integration with other psychophysical measurements. We discuss how possible usage of the developed computational tools might advance our understanding of pain disorders towards clinical diagnosis to differentiate malfunctioning subsystems.