IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 54, NO. 12, DECEMBER 2007 2237 Impedance-Based Ventilation Detection During Cardiopulmonary Resuscitation Martin Risdal*, Sven Ole Aase, Member, IEEE, Mette Stavland, and Trygve Eftestøl, Member, IEEE Abstract—It has been suggested to develop automated external defibrillators with the ability to monitor cardiopulmonary resus- citation (CPR) performance online and give corrective feedback in order to improve the resuscitation quality. Thoracic impedance changes are closely correlated to lung volume changes and can be used to monitor the ventilatory activity. We developed a pattern- recognition-based detection system that uses thoracic impedance to accurately detect ventilation during ongoing CPR. The detection system was developed and evaluated on recordings of real-world resuscitation efforts of cardiac arrest patients where ventilations were manually annotated by human experts. The annotated ven- tilations were detected with an overall positive predictive value of 95.5% for a sensitivity of 90.4%. During chest compressions, the detection system achieved a mean positive predictive value of 94.8% for a sensitivity of 88.7%. The results suggest that accurate ventilation detection during CPR based on the proposed approach is feasible, and that the performance is not significantly degraded in the presence of chest compressions. Index Terms—Defibrillators, impedance, neural networks, pat- tern recognition. I. INTRODUCTION R ECENT publications [1], [2] concerning cardiopulmonary resuscitation (CPR) quality have shown high incidence of divergence from the recommendations of the CPR guidelines from 2000 [3]. The main problems are long pauses in CPR and too shallow compressions [2], and high ventilation rates and low compression rates [1]. Animal studies have shown that high ven- tilation rates markedly decreases survival rates [4]. Other studies also indicate that the quality of CPR performance influences the outcome [5]–[7]. A possible way to improve CPR quality can be to use real-time automated feedback for guidance of personnel performing CPR. By online monitoring of the delivered CPR, voice prompts can, for example, instruct the rescuer on how to adjust the ventilation rate and inspiration time so that they ad- here to the recommendations of the resuscitation guidelines [8]. In [8], it is stated that several studies have shown consistent im- provement in quality of CPR or end tidal CO , or both, when feedback was provided with a variety of formats to guide CPR. Transthoracic impedance (TI) changes are closely correlated to lung volume changes [9]. During inspiration, TI increases, Manuscript received March 27, 2006; revised February 27, 2007. This work was supported by the Norwegian Research Council and the Laerdal Medical AS. Asterisk indicates corresponding author. *M. Risdal was with the Department of Electrical and Computer Engineering, University of Stavanger, Stavanger 4036, Norway. He is now with the Roxar Flow Measurements AS, PDMS, Gamle Forusvei 17, P.O. Box 112, Stavanger N-4065, Norway (e-mail: martin.risdal@roxar.com). S. O. Aase and T. Eftestøl are with the Department of Electrical and Computer Engineering, University of Stavanger, Stavanger 4036, Norway. M. Stavland is with Laerdal Medical AS, Stavanger 4002, Norway. Digital Object Identifier 10.1109/TBME.2007.908328 Fig. 1. Illustration of the onset of inspiration and the onset of expiration to be detected. while it decreases during expiration. The TI measured via the defibrillator pads using a modified automated external defibrillator (AED) can be used for ventilation assessment [10]. Through analysis of the TI signal, the AED might be able to automatically give accurate feedback on ventilation rate, inspiration time, and possibly, inspiration depth. The first step of a TI-based feedback system is ventilation detection. During CPR, this is not easy as the TI signal is very sensitive to movement of the patient, and chest compressions severely corrupt the signal. It also presents a significant base- line drift, and the ventilatory-related curves may assume a wide range of amplitudes, durations, and shapes. The relationship be- tween tidal volume and the resulting impedance change is, in ad- dition, patient dependent [9]. We, therefore, wish to explore the potential of automatically detecting ventilatory activity during CPR using the TI signal measured by an AED. In this paper, we present an impedance-based ventilation de- tection system for use during CPR. Our goal was to develop a detection system that is similar to the way human recognition is done to identify ventilations. The system first recognizes onset of inspiration (OI) and onset of expiration (OE) (see Fig. 1). A set of rules is then used to make sure that the potential ventila- tion cycle fulfils certain criteria. A pattern recognition approach using a neural network is used to identify the potential OIs and OEs. A similar approach was used in [11] for use in apnea mon- itoring, and was found to be superior to two less complex de- flection point algorithms. We have organized this paper as follows. In Section II, we present an overview of the detection system and continue by 0018-9294/$25.00 © 2007 IEEE