Abstract— Labor is the physiological process during which
the fetus is expelled from the uterus and is normally clinically
diagnosed. However, the process leading to labor, typically
involving an increase in contractions frequency and intensity
can be monitored non-invasively using electrohysterography
(EHG) and maternal heart rate (HR). Despite recent
technological improvements, diagnosing or detecting labor
outside of the hospital environment remains challenging due to
much inter-personal variability and lack of data collected in
free-living settings. In this abstract we present preliminary
results of the deployment of the Bloomlife pregnancy tracker, a
unique sensor able to capture EHG, HR, and accelerometer
data. A sample of 51 women wore the sensor between
pregnancy week 26 and 40 (or a subset of such period),
including 12 labor recordings. We show that labor probability
models developed in the lab can effectively provide information
on the odds of delivering and could potentially be used in
real-life situations to improve decision making as delivery
I. INTRODUCTION AND RELATED WORKS
One of the most promising non-invasive markers of labor
and preterm labor is the electrical activity of the uterus .
Wearable sensors able to acquire physiological data
non-invasively, together with recent advances in signal
processing and machine learning techniques, can finally
provide a way to passively and safely investigate changes in
the electrohysterogram (EHG) and heart rate (HR)
throughout pregnancy and potentially provides a new tool
able to predict or detect labor outside of hospital settings.
In our previous work we have shown how physiological
changes in uterine and cardiac activity could be used to
discriminate between labor and pregnancy recordings with
high accuracy (87%, reported in , compared to 68% when
using only information related to gestational age, GA).
However, all labor recordings were collected under
supervised laboratory conditions. On the other hand, in this
work we analyze data collected longitudinally in free-living
conditions, without any researcher’s supervision, and show
how our labor probability models capture differences
between labor and pregnancy recordings on longitudinal data
acquired from week 26 of pregnancy onwards.
II. DATA ANALYSIS
A total of 679 recordings from 51 pregnant women were
collected at different time points during pregnancy (GA mean
36w5, standard deviation 2w8). The actual delivery date was
M. Altini and J. Penders are with Bloomlife, San Francisco, USA.
Corresponding author email: firstname.lastname@example.org
M.J. Rooijakkers and E. Rossetti are with Bloom Technologies, Genk,
reported following an interview after delivery. All women
agreed to their data being used for research as part of
Measurements were obtained with the commercial version of
the Bloomlife wearable device, attached to the skin using a
medical grade adhesive patch. Recordings were split in
20-minutes chunks and time and frequency domain features
were extracted from EHG and HR data. Generalized linear
models to predict the probability of being in labor were
developed using a previously collected dataset including
laboratory data and reference TOCO measurements. Such
models were used to determine the probability of being in
labor for each individual recording collected in free-living,
and showed how labor recordings were assigned much higher
probability with respect to other recordings despite some of
the labor recordings being quite early in pregnancy (range:
week 34 to week 41). In particular, mean probability across
labor recordings was 67% while mean probability across all
non-labor recordings was 25% (see Figure 1).
Figure 1. Probability of being in labor with respect to time to delivery,
clustered by actual labor (in blue) and pregnancy recordings (in red). It can
be seen that actual labor recordings are assigned higher probability.
In this abstract we show preliminary data highlighting
how labor probability models can effectively provide
information on the odds of delivering and could potentially
be used in real-life situations to improve decision making.
 S. Rihana et al. Mathematical modeling of electrical activity of uterine
muscle cells, Medical & biological engineering & computing, vol. 47,
no. 6, pp. 665–675, 2009.
 M. Altini, et al. Combining electrohysterography and heart rate data to
detect labour. In Biomedical & Health Informatics (BHI), 2017 IEEE
EMBS International Conference on (pp. 149-152). IEEE.
Towards Personalized and Non-Invasive Labor Detection Using
Bloomlife Pregnancy Tracker
M. Altini, M.J. Rooijakkers, E. Rossetti and J. Penders