Evidence-based clinical engineering: Machine learning algorithms for prediction of defibrillator performance

• Published in: Biomedical signal processing and control Vol. 54; p. 101629
• Main Authors: Badnjević, Almir, Gurbeta Pokvić, Lejla, Hasičić, Mehrija, Bandić, Lejla, Mašetić, Zerina, Kovačević, Živorad, Kevrić, Jasmin, Pecchia, Leandro
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.09.2019
• Subjects: Automated system; Evidence-based; Inspection; Machine learning; Maintenance; Management; Medical device; Prediction; Automated system; Machine learning; Prediction; Inspection; Medical device; Maintenance; Management; Evidence-based; Performance
• ISSN: 1746-8094; 1746-8108
• DOI: 10.1016/j.bspc.2019.101629

•Paper structure was adjusted.•Part with explination of different machine learning algorithms was deleted.•Abstract was adjusted.•Overall english literacy check was performed.•Discussion was corrected in a more comprihensive matter. Poorly regulated and insufficiently supervised medical devices (MDs) carry high risk of performance accuracy and safety deviations effecting the clinical accuracy and efficiency of patient diagnosis and treatments. Even with the increase of technological sophistication of devices, incidents involving defibrillator malfunction are unfortunately not rare. To address this, we have developed an automated system based on machine learning algorithms that can predict performance of defibrillators and possible performance failures of the device which can affect performance. To develop an automated system, with high accuracy, overall dataset containing safety and performance measurements data was acquired from periodical safety and performance inspections of 1221 defibrillator. These inspections were carried out in period 2015–2017 in private and public healthcare institutions in Bosnia and Herzegovina by ISO 17,020 accredited laboratory. Out of overall number of samples, 974 of them were used during system development and 247 samples were used for subsequent validation of system performance. During system development, 5 different machine learning algorithms were used, and resulting systems were compared by obtained performance. The results of this study demonstrate that clinical engineering and health technology management benefit from application of machine learning in terms of cost optimization and medical device management. Automated systems, based on machine learning algorithms, can predict defibrillator performance with high accuracy. Systems based on Random Forest classifier with Genetic Algorithm feature selection yielded highest accuracy among other machine learning systems. Adoption of such systems will help in overcoming challenges of adapting maintenance and medical device supervision mechanism protocols to rapid technological development of these devices. Due to increased complexity of healthcare institution environment and increased technological complexity of medical devices, performing maintenance strategies in traditional manner is causing a lot of difficulties.