Validation of non-invasive brain temperature estimation models during swine therapeutic hypothermia

Abstract

Objectives: This paper introduces a mathematical model that can estimate deep brain temperature during therapeutic hypothermia (TH) based on a double sensor method (DSM). Although the cerebral temperature is more important than the non-cerebral core temperature during TH, pulmonary artery (PA), rectal, and esophageal measurements (i.e. the typical core temperature measurement locations) have all been used for target temperature management. This is because there is no safe means of measuring the exact brain temperature. Approach: We applied a double sensor thermometer to the subject’s forehead to measure the cerebral temperature non-invasively. Invasive and non-invasive brain temperature readings were acquired for 11 pigs, seven of which were used to develop an optimal model using jackknife resampling and four of which were used to test the model. Main results: The logit model exhibited the best performance of 0.134 °C root mean square error and a 0.993 Lin’s concordance correlation coefficient (CCC). Each test dataset had acceptable results in that each 95% limit of agreement was within the range of clinical acceptance of [−0.5 °C, 0.5 °C]. Three of the four datasets yielded an ‘almost perfect’ score for Lin’s CCC. Significance: Only a small number of studies have compared invasively and non-invasively measured brain temperatures, while most previous studies have concentrated on comparison with the core temperature. Furthermore, the possibility of measuring the exact brain temperature safely during TH using a DSM is shown in this work.