Figure 1

Respiratory rate (RR) estimation in spontaneously breathing humans. (A) RR estimates, in breaths per minute (bpm), during 3 levels of exercise in one subject. Algorithm-estimated RRs (estimated, blue) are compared with RR measured from the subject using the respiratory inductive plethysmography based Hexoskin monitor (expected, red) while performing three consecutive tasks: (1) resting, standing upright on a treadmill (Int 1); (2) walking on the treadmill at a moderate speed (1.2 m/s) (int 2); and (3) walking on the treadmill with 15% track inclination at the moderate speed (Int 3). (B) Summary results of algorithm-estimated and reference RRs (blue and red, respectively) during each subject-task interval. The data are from seven subjects, each performing either or all the three levels of exercise described above (subject-tasks), and presented in order of increasing average expected RR values. (C) The absolute errors (black) and relative errors (gray) of the algorithmic RR estimations across the subject-task intervals described above. Equivalence testing revealed that the expected and estimated RRs were the same (p < 0.0001) for all subject-task intervals. (D) A comparison of the ECG cycle-to-cycle estimated and expected RR for all subjects and tasks with the indicated R² value (0.9092) and low root mean square error (RMSE, 2.2bpm) support a close linear relationship between the values. (E) Absolute error (bpm) and (F) relative error (%) distributions across all subjects.