S-Space College of Engineering/Engineering Practice School (공과대학/대학원) Program in Bioengineering (협동과정-바이오엔지니어링전공) Theses (Master's Degree_협동과정-바이오엔지니어링전공)
Development of a Kicking Detection Algorithm for Extracorporeal Membrane Oxygenation
- 공과대학 협동과정 바이오엔지니어링전공
- Issue Date
- 서울대학교 대학원
- Cardiopulmonary support system; Extracorporeal membrane oxygenation (ECMO); Suction detection; Machine learning; Mock circulation system
- 학위논문 (석사)-- 서울대학교 대학원 : 공과대학 협동과정 바이오엔지니어링전공, 2018. 2. 이정찬.
- Extracorporeal membrane oxygenation (ECMO) is an extracorporeal technique to provide both cardiac and respiratory support to person whose heart and lungs are unable to supply an adequate amount of gas exchange to sustain life. Kicking phenomenon which is the blockage of the drainage cannula creates instant vacuum in the pump which causes cavitation, hemolysis and drop in pumping efficiency. Studies of suction detection on ECMO system seems to be insufficient since previous studies of suction detection algorithms have mostly focused on implantable blood pumps such as total artificial heart (TAH) and ventricular assist devices (VAD).
The purpose of this research was to develop an algorithm which detect kicking phenomenon
such as inlet pressure, flow rate, rotating speed and current consumption of rotating motor of the pump were selected as candidates for kicking detection indicators. The data of the candidate parameters were collected for 24 hours from veno-arterial ECMO operation to a female pig. The correlation between acquired data from ECMO device and the kicking data earned with acceleration data of drainage circuit were analyzed by evaluating confusion matrix of models attained by machine learning algorithm.
Motor current consumption data has outperformed other parameters over than 50% in terms of sensitivity and precision. Various algorithms were developed using the current consumption data. The algorithm using standard deviation was selected as the final kicking detection algorithm by showing better detection ability than the other detection algorithms in various in vitro experimental conditions. The suggested algorithm showed 97% accuracy when applied to the actual kicking data.
Motor current consumption data which the suggested algorithm adopted, has as several advantages other than accuracy of the algorithm itself. Other parameters such as drainage pressure and flow rate require additional sensors on the ECMO circuit, while it is unnecessary to equip any additional sensors utilizing motor current data. Also, motor current data acquisition is never disrupted unless the motor turns off.
Although there remain some limitations in the present study, the results light up the potential of further studies on suction detection system in terms of ECMO operations.