Feasibility of energy expenditure estimation during outdoor walking using multimodal in-ear sensors: a preliminary validation study against indirect calorimetry.

A growing number of portable devices has become available for daily physical activity and energy expenditure (EE) monitoring. Although several studies have examined EE estimation using heart rate (HR) and accelerometry (ACC), no study to date has explored EE prediction based on HR and ACC data measured simultaneously in the ear canal by a single commercial wearable device. Therefore, the aim of this study was to assess the feasibility of estimating EE using the cosinuss[Formula: see text] in-ear sensors [Formula: see text]One and [Formula: see text]Two. Twenty-four healthy adults (12 female, 12 male, age 26.91 ± 2.91 years, body mass 65.43 ± 8.01 kg) completed two test sessions: a laboratory treadmill test (Modified Bruce Protocol) and an outdoor walking test consisting of three phases (uphill at comfortable pace, downhill at comfortable pace, and uphill as fast as possible). EE was measured by the portable indirect calorimetry device MetaMax 3B (criterion measure), while HR and ACC were recorded continuously by the in-ear sensors [Formula: see text]One and [Formula: see text]Two. Linear mixed-effects models were developed from laboratory data and applied to outdoor walking data from the same participants (participant-calibrated prediction). Prediction accuracy was assessed using root mean square error (RMSE), mean absolute percentage error (MAPE), and Bland-Altman analysis. Both in-ear sensors showed preliminary potential for EE prediction during outdoor walking. The integration of HR, ACC, and biological sex produced the best-fitting calibration models, with marginal [Formula: see text] values of 0.79 and 0.80 for [Formula: see text]One and [Formula: see text]Two respectively. In the outdoor walking test, no significant systematic bias was detected across conditions ([Formula: see text]). However, prediction accuracy was variable across conditions and subgroups, with MAPE ranging from 22% to 37% for [Formula: see text]One and 22% to 38% for [Formula: see text]Two (following exclusion of one participant with implausible reference data in the Down phase). Pearson correlation coefficients were stronger for female participants, while male subgroup correlations were weak or negative in several conditions, limiting conclusions about individual-level prediction accuracy for this subgroup. The cosinuss[Formula: see text] in-ear sensors demonstrated preliminary feasibility for group-level EE estimation during outdoor walking, with accuracy broadly comparable to several commercial wrist-worn devices. Given the variable prediction accuracy, particularly for male participants and during downhill walking, further validation with larger and more diverse samples is needed before clinical or sports application.