Estimation of respiratory rate using infrared video in an inpatient population: an observational study
- PDF / 1,208,548 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 66 Downloads / 186 Views
ORIGINAL RESEARCH
Estimation of respiratory rate using infrared video in an inpatient population: an observational study Peter Chan1,4 · Gabriel Wong1 · Toan Dinh Nguyen2 · Tam Nguyen3 · John McNeil4 · Ingrid Hopper4 Received: 30 June 2019 / Accepted: 28 November 2019 © Springer Nature B.V. 2019
Abstract Respiratory rate (RR) is one of the most sensitive markers of a deteriorating patient. Despite this, there is significant interobserver discrepancy when measured by clinical staff, and modalities used in clinical practice such as ECG bioimpedance are prone to error. This study utilized infrared thermography (IRT) to measure RR in a critically ill population in the Intensive Care Unit. This study was carried out in a Single Hospital Centre. Respiratory rate in 27 extubated ICU patients was counted by two observers and compared to ECG Bioimpedance and IRT-derived RR at distances of 0.4–0.6 m and > 1 m respectively. IRT-derived RR using two separate computer vision algorithms outperformed ECG derived RR at distances of 0.4–0.6 m. Using an Autocorrelation estimator, mean bias was − 0.667 breaths/min. Using a Fast Fourier Transform estimator, mean bias was − 1.000 breaths/min. At distances greater than 1 m no statistically significant signal could be obtained. Over all frequencies, there was a significant relationship between the RR estimated using IRT and via manual counting, with Pearson correlation coefficients between 0.796 and 0.943 (p 19 bpm (r = 0.562, p = 0.029). Overall agreement between IRT-derived RR at distances of 0.4–0.6 m and gold standard counting was satisfactory, and outperformed ECG derived bioimpedance. Contactless IRT derived RR may be feasible as a routine monitoring modality in wards and subacute inpatient settings. Keywords Contactless · Monitoring · Respiratory rate · Infrared · Critical care
1 Introduction/background * Peter Chan [email protected] Gabriel Wong [email protected] Toan Dinh Nguyen [email protected] Tam Nguyen [email protected] John McNeil [email protected] Ingrid Hopper [email protected] 1
Eastern Health Intensive Care Services, Eastern Health, Melbourne, Australia
2
Monash eResearch Centre, Monash University, Melbourne, Australia
3
St Vincent’s Hospital, Melbourne, Australia
4
School of Public Health and Prevention Medicine, Monash University, Melbourne, Australia
Identifying patients at risk of deterioration in the inpatient setting as quickly as possible is important to minimize injury and promote faster recovery. It has been shown that earlier intervention results in better outcomes [1]. Equally, reduced monitoring has been demonstrated to result in in-hospital cardiac arrests, resulting in increased morbidity to patient [2]. Continuous monitoring of patients has been shown to rapidly identify deteriorating patients and improve outcomes through a more prompt response from a Medical Emergency (MET) or Rapid Response (RRT) teams, increased admission to the Intensive Care Unit, and reduced incidence of I
