The ECG-based stethoscope tracking technology described in this abstract aims to improve the process of obtaining abnormal cardiac auscultation findings when using a standardized patient to train medical and health profession students in the acquisition of that skill. Since SPs are typically healthy individuals with normal findings, the number and range of conditions that students can be exposed to is limited without the use of one of many technology applications meant to enhance cardiac auscultation in clinical simulation.1 Real time tracking of a stethoscope head using ECG signals would allow for the development of a teaching stethoscope that provides the standardized patient and the learner a more realistic, reproducible and flexible experience compared to currently available technology.
In a previous study, a modified stethoscope head with two electrodes was used to pick up ECG signals at the four cardiac auscultation sites (and at various angle orientations from vertical position) on two male subjects in the seated position. The method correctly identified the cardiac auscultation sites with 95 percent accuracy, irrespective of angle variation. This study was conducted to extend the validity of the ECG-based stethoscope tracking technology to a larger subject pool, assess classification accuracy of the technology on subjects in a supine vs. seated position and determine if subject characteristics (gender and body mass index) influence classification accuracy. Data collection for this study was achieved using the Welch-Allyn Meditron stethoscope to record ECG signals from the four cardiac auscultation areas of 35 individual participants. Five 10-second runs were recorded with the subject supine and seated up-right. Electrode gel was used to decrease motion artifacts from the signals.2 The signals that were obtained were then preprocessed and filtered. Noises and artifacts from breathing, body movements and power line interference were filtered using low pass and high pass filters. After the collected ECG signals had been processed and filtered a Pan-Tompkins algorithm3 was used to identify characteristics of QRS complexes. Amplitude and time interval features (e.g. RQamp and RQ) were then extracted and the Results from the algorithm predicted cardiac auscultation sites of the collected signals.
The algorithm classified the cardiac auscultation areas with the stethoscope head placed at the four target sites. Based on features extracted from each QRS wave, a 10-fold cross validation of a Random Forest classifier was conducted. The classifier accuracy for subjects in the supine position was 89.3 ± 6.80, in the seated upright position was 89.12 ± 5.07, and combined was 85.73 ± 6.18 at the 95 percent confidence level. Paired and two sample t-tests revealed no significant difference in auscultation area classification from ECG signals collected from supine and seated, among males and females, and among normal and overweight/obese (Body Mass Index> 25) subjects.
The result of 86% accuracy for the combined data set was obtained in classifying the four different auscultation areas. The statistical comparative tests showed that the ECG-based stethoscope tracking technology can be extended to a larger subject pool and function reliably irrespective of subject’s body position, gender or body mass index. The classification Results can be further improved by performing an online classification that makes predictions based on a sequence of QRS waves from the incoming ECG signal increasing cardiac auscultation area classification accuracy. Advances in this technology hope to produce a versatile stethoscope that allows for a more realistic experience for learners when acquiring the skills necessary for cardiac auscultation using standardized patients.
1. Ward J, Wattier B: Technology for Enhancing Chest Auscultation in Clinical Simulation. Respir Care 2011;56(6):834-845.
2. Cömert A, Honkala M, Hyttinen J: Effect of pressure and padding on motion artifact of textile electrodes. Biomed Eng Online 2013;12(1):26.
3. Pan J, Tompkins W: A real-time QRS detection algorithm. IEEE Trans Biomed Eng 1985;32(3):230-236.
Research could result in a future product that would be marketed by Cardionics. Inc. with which this author has a patent licensure and royalty agreement.
© 2013 by Lippincott Williams & Wilkins, Inc.