Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Machine-learned analysis of quantitative sensory testing responses to noxious cold stimulation in healthy subjects

 
: Weyer-Menkhoff, I.; Thrun, M.C.; Lötsch, J.

:

European journal of pain : EJP 22 (2018), No.5, pp.862-874
ISSN: 1090-3801 (print)
ISSN: 1532-2149 (online)
English
Journal Article
Fraunhofer IME ()

Abstract
Background
Pain in response to noxious cold has a complex molecular background probably involving several types of sensors. A recent observation has been the multimodal distribution of human cold pain thresholds. This study aimed at analysing reproducibility and stability of this observation and further exploration of data patterns supporting a complex background.
Method
Pain thresholds to noxious cold stimuli (range 32–0°C, tonic: temperature decrease −1 °C/s, phasic: temperature decrease −8 °C/s) were acquired in 148 healthy volunteers. The probability density distribution was analysed using machine‐learning derived methods implemented as Gaussian mixture modeling (GMM), emergent self‐organizing maps and self‐organizing swarms of data agents.
Results
The probability density function of pain responses was trimodal (mean thresholds at 25.9, 18.4 and 8.0°C for tonic and 24.5, 18.1 and 7.5°C for phasic stimuli). Subjects' association with Gaussian modes was consistent between both types of stimuli (weighted Cohen's κ = 0.91). Patterns emerging in self‐organizing neuronal maps and swarms could be associated with different trends towards decreasing cold pain sensitivity in different Gaussian modes. On self‐organizing maps, the third Gaussian mode emerged as particularly distinct.
Conclusion
Thresholds at, roughly, 25 and 18 °C agree with known working temperatures of TRPM8 and TRPA1 ion channels, respectively, and hint at relative local dominance of either channel in respective subjects. Data patterns suggest involvement of further distinct mechanisms in cold pain perception at lower temperatures. Findings support data science approaches to identify biologically plausible hints at complex molecular mechanisms underlying human pain phenotypes.
Significance
Sensitivity to pain is heterogeneous. Data‐driven computational research approaches allow the identification of subgroups of subjects with a distinct pattern of sensitivity to cold stimuli. The subgroups are reproducible with different types of noxious cold stimuli. Subgroups show pattern that hints at distinct and inter‐individually different types of the underlying molecular background.

: http://publica.fraunhofer.de/documents/N-524715.html