Graduation / internship project: A Multiparametric Machine Learning Approach for the Early Detection of Clinical Alarms in Preterm Infants

Background

Patient monitoring generates a large number of alarms, the vast majority of which are false. Excessive non-actionable medical alarms leads to alarm fatigue, a well-recognized patient safety issue1. Previous neonatal research in our group has identified specific patterns in red alarms (critical alarms) that can be exploited to reduce clinically non-actionable alarms2. Exploratory research on the early detection of red alarms based on the occurrence of yellow alarms (alerts) has also shown promise for early detection but is limited by poor sensitivity and specificity.

Aim

The aim of this project is to develop an algorithm for early warning of alarms using the continuously acquired vital signs, the heart rate, SpO2 and breathing rate, for the early detection of red desaturation and red bradycardia alarms in preterm infants.

Methods

The project will be based on developing features and using a machine learning approach for classifying those yellow alarms that lead to a red alarm within a short window of time. Data for the project is available and has been acquired from the data warehouse in the NICU of Máxima Medical Center, Veldhoven.

Position

A 6 month internship contract (extension possible) with Philips Research, Eindhoven (HTC). This can be a Master thesis, an internship or a long project. The internship will however be carried out at the department of clinical physics at Máxima Medical Center, Veldhoven which is where the student will be located.

References

1. Mitka M. Joint commission warns of alarm fatigue: multitude of alarms from monitoring devices problematic. JAMA. 2013;309(22):2315-2316. doi:10.1001/jama.2013.6032.
2. Joshi R, van Pul C, Atallah L, Feijs L, Van Huffel S, Andriessen P. Pattern discovery in critical alarms originating from neonates under intensive care. Physiol Meas. 2016;37(4):564-579. doi:10.1088/0967-3334/37/4/564.

Requirements

- A master’s student in Electrical Engineering, Biomedical Engineering, Computer Science or related discipline.
- Good background in signal processing and machine learning (or willingness to learn).
- Good Matlab programming skills.
- Willingness to learn neonatal physiology and develop physiologically meaningful features.

Contact details

If interested, please send an email with your CV to the undersigned:
- Rohan Joshi, IMPULS-I PhD candidate at the Eindhoven University of Technology
- Email: rohan.joshi@philips.com

Starting date

As soon as possible. After selection it takes approximately 4 weeks and 8 weeks respectively for administrative processing of Dutch students and international students.