Katharina Jäger

  • Job title: Master's Thesis
  • Working group: Detecting Labor in Pregnant Women with a Wearable Pregnancy Tracker using Machine Learning

Robert Richer, M. Sc., Michiel Rooijakkers, PhD, Quentin Noirhomme, PhD, Violaine Emonard, MD (Bloomlife, Inc.), Prof. Dr. Björn Eskofier

11/2019 – 05/2020


More than 10% of all births in the world are premature, i.e. they occur before the end of 37 weeks of gestation [1]. Preterm birth complications are the leading cause of death among children under 5 years [2]. According to the WHO many survivors face a lifetime of handicap, including learning disabilities, visual and hearing problems [3]. Most preterm deliveries happen spontaneously due to an early onset of labor caused by either multiple pregnancies or chronic conditions, but often no specific cause can be identified [3]. Even though the total number of incidences slightly decreased over the last years, the main decrease was among late preterm deliveries [4].

To successfully treat and prevent premature birth detecting labor as soon as possible is important [5]. Labor is the physiological process of delivering a baby. It is defined as an increase in duration and regularity of uterine contractions resulting in effacement and dilation of the cervix [6]. The intrauterine pressure catheter (IUPc) is the gold standard for monitoring uterine activity. The main problem of the IUPc is its invasiveness, which requires a rupture of the membranes and can therefore only be used during delivery, at the risk of damaging maternal or fetal tissues and infection [7]. The non-invasive tocodynamometer (TOCO) is standard in everyday clinical practice but less reliable due to the indirect way of measurement. Recently, uterine electromyography (EMG) has shown to be a more accurate way of monitoring uterine contractility compared to the TOCO [8]. Nevertheless, the TOCO is still the de facto standard in clinical practice. None of these medical devices currently distinguishes between actual labor contractions and pregnancy contractions or performs decision making, which is only done in academic studies so far.

Academic studies differ mostly in the type and number of electrodes and in the applied features and methods. Very promising results have been achieved, e.g. using artificial neural networks processing on uterine EMG burst features as an automatic categorization [9]. However, the promising results found in literature are limited to data collected in controlled laboratory conditions and using a single recording per subject, and they haven’t been shown to extend to free living conditions. Lack of technology suitable for high-quality recordings in free-living situations has limited research in this field, but recent innovations in the field of wearable technology could allow overcoming previous limitations. The Bloomlife sensor is a wearable device consisting of an electronic sensor module and a disposable adhesive patch with 3 sensing electrodes, which measures electrophysiological signals detected from the pregnant women’s abdomen. Its ease of use and convenience make it an ideal tool to continuously acquire data throughout pregnancy and labor in a natural environment.

The goal of this master thesis is therefore to develop a labor detection approach using machine learning techniques and a subset of data acquired with the Bloomlife sensor. The data used in this project was collected in free-living conditions by more than 10,000 women (Bloomlife customers) during the last weeks of their pregnancy. In addition, characteristics that evolve during pregnancy and might indicate the preparation of the maternal body for delivery are to be investigated.



  1. S. Chawanpaiboon, J. P. Vogel, A. B. Moller, P. Lumbiganon, M. Petzold and D. Hogan, “Global, regional, and national estimates of levels of preterm birth in 2014: a systematic review and modelling analysis,” The Lancet Global Health, vol. 7, no. 1, pp. e37-e46, 2019.
  2.  L. Liu, S. Oza, D. Hogan, Y. Chu, J. Perin, J. Zhu, J. E. Lawn, S. Cousens, C. Mathers and R. E. Black, “Global, regional, and national causes of under-5 mortality in 2000–15: an updated systematic analysis with implications for the Sustainable Development Goals,” The Lancet, vol. 388, pp. 3027-3035, 2016.
  3. “Born Too Soon: The Global Action Report on Preterm Birth,” World Health Organization, 2012.
  4.  B. E. Hamilton, J. A. Martin and S. J. Ventura, “National Vital Statistics Reports, Volume 58, Number 16 – Provisional (April 2010),” vol. 58, 2010.
  5.  M. Altini, E. Rossetti, M. J. Rooijakkers and J. Penders, “Combining electrohysterography and heart rate data to detect labour,” 2017 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI 2017), pp. 149-152, 2017.
  6.  E. R. Norwith, J. N. Robinson and J. R. G. Challis, “The control of labor,” The New England Journal of Medicine Review, vol. 341, pp. 660-666, 1999.
  7. T. Y. Euliano, M. T. Nguyen, S. Darmanjian, S. P. McGorray, N. Euliano, A. Onkala and A. R. Gregg, “Monitoring uterine activity during labor: A comparison of 3 methods,” American Journal of Obstetrics and Gynecology, vol. 208,
  8. R. E. Garfield, M. Lucovnik and R. J. Kuon, “Diagnosis and Effective Management of Preterm Labor,” MGM Journal of Medical Sciences, vol. 1, pp. 22-37, 2014.
  9. W. L. Maner and R. E. Garfield, “Identification of human term and preterm labor using artificial neural networks on uterine electromyography data,” Annals of Biomedical Engineering, vol. 35, pp. 465-473, 2007.