Progress in non-invasive respiratory monitoring using uncalibrated breathing movement components

Estimation of respiratory volume from thoracoabdominal breathing distances: comparison of two models of machine learning

PURPOSE

The purposes of this study were to both improve the accuracy of respiratory volume (V) estimates using the respiratory magnetometer plethysmography (RMP) technique and facilitate the use of this technique.

METHOD

We compared two models of machine learning (ML) for estimating [Formula: see text]: a linear model (multiple linear regression-MLR) and a nonlinear model (artificial neural network-ANN), and we used cross-validation to validate these models. Fourteen healthy adults, aged [Formula: see text] years participated in the present study. The protocol was conducted in a laboratory test room. The anteroposterior displacements of the rib cage and abdomen, and the axial displacements of the chest wall and spine were measured using two pairs of magnetometers. [Formula: see text] was estimated from these four signals, and the respiratory volume was simultaneously measured using a spirometer ([Formula: see text]) under lying, sitting and standing conditions as well as various exercise conditions (working on computer, treadmill walking at 4 and 6 km[Formula: see text], treadmill running at 9 and 12  km [Formula: see text] and ergometer cycling at 90 and 110 W).

RESULTS

The results from the ANN model fitted the spirometer volume significantly better than those obtained through MLR. Considering all activities, the difference between [Formula: see text] and [Formula: see text] (bias) was higher for the MLR model ([Formula: see text] L) than for the ANN model ([Formula: see text] L).

CONCLUSION

Our results demonstrate that this new processing approach for RMP seems to be a valid tool for estimating V with sufficient accuracy during lying, sitting and standing and under various exercise conditions.

Adaptive motion artefact reduction in respiration and ECG signals for wearable healthcare monitoring systems

Wearable healthcare monitoring systems (WHMSs) have received significant interest from both academia and industry with the advantage of non-intrusive and ambulatory monitoring. The aim of this paper is to investigate the use of an adaptive filter to reduce motion artefact (MA) in physiological signals acquired by WHMSs. In our study, a WHMS is used to acquire ECG, respiration and triaxial accelerometer (ACC) signals during incremental treadmill and cycle ergometry exercises. With these signals, performances of adaptive MA cancellation are evaluated in both respiration and ECG signals. To achieve effective and robust MA cancellation, three axial outputs of the ACC are employed to estimate the MA by a bank of gradient adaptive Laguerre lattice (GALL) filter, and the outputs of the GALL filters are further combined with time-varying weights determined by a Kalman filter. The results show that for the respiratory signals, MA component can be reduced and signal quality can be improved effectively (the power ratio between the MA-corrupted respiratory signal and the adaptive filtered signal was 1.31 in running condition, and the corresponding signal quality was improved from 0.77 to 0.96). Combination of the GALL and Kalman filters can achieve robust MA cancellation without supervised selection of the reference axis from the ACC. For ECG, the MA component can also be reduced by adaptive filtering. The signal quality, however, could not be improved substantially just by the adaptive filter with the ACC outputs as the reference signals.

Nonlinear model for estimating respiratory volume based on thoracoabdominal breathing movements

BACKGROUND AND OBJECTIVE

Respiratory inductive plethysmography is a non-invasive technique for measuring respiratory function. However, there are challenges associated with using linear methods for calibration of respiratory inductive plethysmography. In this study, we developed two nonlinear models, artificial neural network and adaptive neuro-fuzzy inference system, to estimate respiratory volume based on thoracoabdominal movements, and compared these models with routine linear approaches, including qualitative diagnostic calibration and multiple linear regression.

METHODS

Recordings of spirometry volume and respiratory inductive plethysmography were obtained for 10 normal subjects and 10 asthmatic patients, during asynchronous breathing for 7 min. The first 5 min of recording were used to develop the models; the remaining data were used for subsequent validation of the results.

RESULTS

The results from the nonlinear models fitted the spirometry volume curve significantly better than those obtained by linear methods, particularly during asynchrony (P < 0.05). On a breath-by-breath analysis, estimates of tidal volume, total cycle time and sigh values using the artificial neural network model were accurate by comparison with qualitative diagnostic calibration. In contrast to the artificial neural network model, there was a significant correlation between values for thoracoabdominal asynchrony and increased error of qualitative diagnostic calibration (P < 0.05).

CONCLUSIONS

These results indicate that the nonlinear methods can be adapted to closely simulate variable conditions and used to study the patterns of volume changes during normal and asynchronous breathing.

Laser 3-D measuring system and real-time visual feedback for teaching and correcting breathing

We present a novel method for real-time 3-D body-shape measurement during breathing based on the laser multiple-line triangulation principle. The laser projector illuminates the measured surface with a pattern of 33 equally inclined light planes. Simultaneously, the camera records the distorted light pattern from a different viewpoint. The acquired images are transferred to a personal computer, where the 3-D surface reconstruction, shape analysis, and display are performed in real time. The measured surface displacements are displayed with a color palette, which enables visual feedback to the patient while breathing is being taught. The measuring range is approximately 400×600×500 mm in width, height, and depth, respectively, and the accuracy of the calibrated apparatus is ±0.7 mm. The system was evaluated by means of its capability to distinguish between different breathing patterns. The accuracy of the measured volumes of chest-wall deformation during breathing was verified using standard methods of volume measurements. The results show that the presented 3-D measuring system with visual feedback has great potential as a diagnostic and training assistance tool when monitoring and evaluating the breathing pattern, because it offers a simple and effective method of graphical communication with the patient.

Scope of linear estimators of tidal and occluded volumes using thoracoabdominal indications of breathing movement coordination

The basic theory for respiratory inductive plethysmography (RIP) applications was re-examined, refined and tested. A realistic model of the RIP interpretation of respiratory mechanics related tidal volumes (VT) to a linear combination of ribcage and abdomen movements. Lissajous plots of asynchronous thoracoabdominal movements revealed their net effect equivalent to the superposition of synchronous and antipathetic respiration modes at right angles, along the principal axes specific to the combined motion. Predictors of relative changes in VT, degree of asynchrony and volume thus being occluded were developed via least squares estimation theory, with an optional validation facility. The approach enabled clinically adequate analysis of 452 h of RIP data from 29 postoperative patients. Correct identification of only seven complete apnoeas in 111 incidences of obstruction during periodic, variable, asynchronous or paradoxical natural breathing was substantiated via non-invasive airflow monitoring. The modelling helped clarify RIP limitations--the possibility of misleading indications from obese or abnormal physiques or movement artefacts degrading its otherwise nearly optimal performance. Nevertheless, our uncalibrated predictors had better theoretical basis, improved reliability and more convenient practical utility than the traditional approach of calibrating RIP by spirometry prior to non-invasive monitoring and identifying and classifying apnoeas.

Key to better qualitative diagnostic calibrations in respiratory inductive plethysmography

Least-squares estimates for coefficients of linear models that predict tidal volume (VT) via respiratory inductive plethysmography (RIP) are given. The qualitative diagnostic calibration sum formula M(RC + KAbd) arises for idealized thoracoabdominal co-ordination within this model-fitting framework. For a normal synchronous breath K is then optimally determined from the ratio of its associated ribcage (RC) and abdomen (Abd) movement standard deviations, not from a ratio that applied to a previously measured breath. M merely rescales relative changes in (RC + KAbd) to absolute changes in VT for correct proportioning. RC and Abd move in complete antipathy during an obstructive apnoea, so use of optimal K ensures (RC + KAbd) tends to zero for such unproductive breathing efforts. The interpretation is extended to more general breathing patterns by using a complementary difference expression M(RC-KAbd) to help identify any antagonistic respiratory actions. The two new constructs are equivalent to the principal components of the combined ribcage and abdomen movements. Together they demonstrate versatile capability in uncalibrated RIP applications for obstructive apnoea detection and tracking relative changes in VT during paradoxical or variable natural breathing. Calibration is appropriate for model-fitting quality assessment but otherwise usually too patient demanding, unnecessary or detrimental to prediction monitoring efficacy.