TL;DR:
- Machine learning, wearables, and pregnancy: A promising intersection.
- Stanford University researchers apply deep learning to wearable data from pregnant individuals.
- Distinctive data profiles hint at preterm birth risk.
- Preterm birth remains a global health challenge, disproportionately affecting Black women.
- Landmark research in npj Digital Medicine highlights the size and diversity of participant groups.
- Wearable data complements medical records, enabling accurate gestational age estimation.
- Women with disrupted activity and sleep patterns face a higher risk of preterm birth.
- Research opens avenues for early interventions and exploring racial disparities.
- Wearable technology’s potential in maternal and child health has gained traction.
- Prospective analyses and machine learning are essential in navigating complex wearable data.
Main AI News:
In the realm of healthcare, wearables have encountered formidable obstacles in their quest to make a lasting impact. Nonetheless, the volatile journey of pregnancy presents an opportune arena to explore their latent potential. Recent research highlights the role of machine learning in scrutinizing data streams from wearable devices to unravel the enigma surrounding premature birth.
At Stanford University, machine learning experts harnessed the power of a deep learning model to dissect data derived from wearable devices, focusing on the activity and sleep patterns of pregnant individuals. Notably, as pregnancy unfolded, sleep quality deteriorated, and physical activity dwindled—a pattern to be expected. However, intriguingly, certain participants displayed data profiles incongruent with their pregnancy stage. It is precisely these pregnancies that the researchers discovered to be more susceptible to preterm birth.
Preterm birth is a global scourge, claiming the lives of numerous children under the age of five, and in the United States, it accounts for approximately 11% of all live births— a statistic that has been on an unsettling ascent over the last decade. Black women, in particular, face an elevated risk, being 1.5 times more likely than their white counterparts to experience premature delivery.
Distinguishing itself from previous studies, the recently published research in npj Digital Medicine stands out due to the sheer scale and diversity of its participant pool. Over 1,000 women from the St. Louis region, with over half of them being Black, were meticulously tracked throughout their pregnancies by a team of dedicated researchers from Washington University in St. Louis.
Nima Aghaeepour, a machine learning researcher at Stanford and the senior author of the paper, applauded the monumental effort, emphasizing their commitment to collecting data as early as possible once pregnancy was confirmed. Participants were entrusted with motion- and light-sensing watches, to be worn for at least a week during each trimester, with some opting to wear them throughout the entire nine months of pregnancy.
The trove of raw, continuous data was then fed into an intricate deep learning pipeline and juxtaposed with the participants’ medical records, including their gestational progress. Remarkably, Aghaeepour stated, “It turns out that we can look at the wearable device and tell how pregnant somebody is, give or take a few weeks.” This determination was based on discerning patterns in physical activity and sleep that are characteristic of different pregnancy stages.
However, the true breakthrough occurred when the researchers identified a subset of women who appeared more pregnant than their gestational age would suggest, as per the deep learning algorithm. These women exhibited disrupted activity and sleep patterns compared to others at a similar pregnancy stage, making them approximately 44% more likely to undergo premature birth.
Notably, this result doesn’t conclusively imply that a lack of activity or poor sleep directly causes preterm births. Erik Herzog, a circadian biologist at Washington University in St. Louis and a co-author of the study, cautioned against this assumption. Instead, it offers a hypothesis for further investigation into the potential role of activity and sleep in premature births.
Benjamin Smarr, a professor at the University of California San Diego with expertise in using temperature data from wearables to detect pregnancy, suggested future avenues of research. He proposed exploring the possibility of generating alerts when wearables indicate a woman appears closer to delivery than her actual gestational age. This could lead to early interventions that might benefit both mother and child.
Furthermore, such prospective analyses could shed light on the persistent racial disparities in preterm birth outcomes. Environmental and societal factors that contribute to poorer outcomes for Black women have remained elusive, and research in this area has been challenging. Smarr emphasized that the root causes extend beyond access to education and socioeconomic status.
Jessica Walter, a reproductive endocrinologist at Northwestern specializing in wearables’ applications in women’s health, noted that pregnancy offers a unique window for such research. Pregnant individuals are often highly motivated to monitor their health and well-being during this dynamic period of physiological change.
Several research groups are now endeavoring to leverage wearable devices to uncover pregnancy-related signals. One noteworthy endeavor is the “Better Understanding the Metamorphosis of Pregnancy” study, led by the nonprofit 4YouandMe, co-founded by former Apple Health researcher Stephen Friend. This ambitious study aims to collect extensive data from about 1,000 pregnant participants using commercial devices, seeking to elucidate the variability in their values.
However, these research endeavors confront a formidable challenge—working with the continuous data generated by wearables. The output is inherently variable, both among individuals and in daily fluctuations, making it challenging to extract actionable insights from simple measures like average activity or daily sleep duration. Machine learning, as highlighted by Jessica Walter, becomes an indispensable tool to navigate vast and intricate datasets, uncovering subtle patterns that carry significant implications for health.
As Herzog aptly puts it, “This machine learning approach said, we’re just going to teach the model that these data are associated with these gestational ages, and it learned features in the data that allowed it to better estimate gestational age.” Nevertheless, trust in the accuracy of machine learning models must be tempered with caution, especially in the realm of healthcare, where transparency and accountability are paramount.
Ultimately, if wearables can indeed be harnessed to identify individuals at risk of preterm birth, rigorous real-world testing will be imperative. Balancing the potential benefits with potential stressors and resource implications is an ongoing challenge in the quest to harness technology for the betterment of maternal and child health.
Conclusion:
The fusion of wearables and machine learning in predicting preterm birth risks presents a promising avenue for improving maternal and child health. With the potential to offer early interventions and address racial disparities, this research underscores the growing significance of technology-driven healthcare solutions in the market. Healthcare providers and wearable technology companies should consider investing in further research and development in this area to bring about meaningful improvements in pregnancy outcomes.
