Agreement between automated ThermoHuman® and manual thermographic analyses for assessing lower-limb skin temperatures before and after running

Although different studies are using ThermoHuman software to delimitate regions of interest (ROI) of lower limbs, no study has validated this automatic analysis. This work evaluated the level of agreement and time efficiency of the automated ThermoHuman software compared with manual determination of the regions of interest in assessing lower-limb skin temperature before and after exercise. Seventy-six thermographic images from 19 recreational runners (10 males, 9 females) were analyzed before and after a 10-km outdoor run. Automated analysis using ThermoHuman was compared with Manual analysis (ROI delineation based on general template) and Manual-TH (manual replication of the same ROI delimitations generated by ThermoHuman). Agreement was assessed using concordance correlation coefficients (CCC), mean absolute error (MAE), Bland-Altman analysis, and linear mixed models. High agreement was observed between methods, with CCC values ranging from 0.95 to 0.98 and MAE between 0.3 °C and 0.5 °C, and bias of 0.0 °C for all comparisons. Mixed-model analysis revealed no significant main effect of method. Additionally, the ROI selection criterion (Manual-TH vs. Manual) did not result in significant differences. Compared with Manual (13.9 h) and Manual-TH (28.7 h), ThermoHuman required substantially less time for image analysis (2.0 h), representing time savings of 85.6 % and 93 %, respectively. In conclusion, automated analysis using ThermoHuman showed high agreement with both manual approaches, including a higher time efficiency.