Screening for peripartum cardiomyopathy using AI-enhanced digital stethoscope shows promise
Click Here to Manage Email Alerts
Key takeaways:
- The study utilized a digital stethoscope that records ECGs and heart sounds and then uses an AI algorithm to predict the likelihood of cardiomyopathy.
- Additional research in other geographic settings is warranted.
PHILADELPHIA — In the SPEC-AI trial, use of an artificial intelligence-enhanced digital stethoscope doubled the number of cases of peripartum cardiomyopathy identified among pregnant and postpartum women.
“The number of individuals who die during pregnancy or shortly after childbirth has remained either the same or is increasing in many parts of the world,” Demilade A. Adedinsewo, MD, MPH, assistant professor of medicine at Mayo Clinic in Jacksonville, Florida, said during a press conference at the American Heart Association Scientific Sessions.
Image: Adobe Stock
Here in the United States, pregnancy-related death is on the rise, in particular among Black women, who have a threefold higher risk of dying compared with white women.
The highest risk for maternal death is among pregnant and new mothers in sub-Saharan Africa. In Nigeria, the country where the SPEC-AI trial was conducted, the number of maternal deaths was 82,000 in 2020 and the incidence of peripartum cardiomyopathy is approximately one in 100 live births, according to Adedinsewo.
Previous research in the U.S. and United Kingdom demonstrated the effectiveness of using an AI-enhanced stethoscope to detect cardiac dysfunction. The SPEC-AI researchers conducted this study to evaluate whether AI-guided screening using this technology would impact the diagnosis of pregnancy-related cardiac dysfunction in an obstetric population in Nigeria.
“This study is the first large clinical trial evaluating an AI intervention in pregnancy,” Adedinsewo said.
Detection of cardiac dysfunction
The trial enrolled 1,195 women aged 18 to 49 years at six sites in Nigeria who were pregnant or had delivered a baby within the previous year. Nearly three-quarters of women were pregnant and 39% were in their third trimester at the start of the study. Participants were randomly assigned to usual care comprising a traditional ECG or to AI-guided screening using a digital stethoscope in addition to a traditional ECG. The digital stethoscope records an ECG and phonocardiogram (heart sounds) and then uses an AI algorithm to predict the likelihood of cardiomyopathy. Recordings were taken at three locations: the V2 position, angled and handheld.
Among 1,195 pregnant and postpartum women, cardiac dysfunction was detected in 1.8% of participants assigned usual care compared with 4.1% of participants assigned AI-guided screening with the digital stethoscope (OR = 2.3).
“This AI-guided screening ... increased the number of cases detected by double, suggesting that up to half of cases of HF cardiomyopathy are very likely missed with usual care,” Adedinsewo said.
The researchers also aimed to validate the algorithm, which was not previously tested in an African population because it was developed in the U.S., Adedinsewo said. Women in the intervention group also received an echocardiogram. In the intervention group, the digital stethoscope correctly identified 92% of women who had an ejection fraction less than 50%, and 100% of women who had an EF less than 40%.
The researchers noted several limitations including a lack of generalizability to the general obstetric population in Nigeria and beyond and the EF cutoff for cardiomyopathy detection of less than 50%.
Additional trials enrolling a diverse group of women in other locations are needed to further evaluate the impact of this technology on cardiomyopathy diagnoses and adverse maternal outcomes, according to Adedinsewo.
‘Potential to improve cardio-obstetric care’
The AI-enhanced digital stethoscope offers several advantages, Adedinsewo said. It is portable and can provide AI prediction in real time, it is battery-powered, and it can be used in different clinical settings and even nonclinic settings.
“AI-guided screening using a digital stethoscope ... is effective for detecting cardiac dysfunction among pregnant and postpartum women. This intervention has the potential to improve cardio-obstetric care by reducing delays in the diagnosis of a life-threatening but treatable condition.”
Commenting on the trial, Alexander Tarlochan Singh Sandhu, MD, MS, assistant professor of medicine (cardiovascular medicine) at Stanford University, said: “This is an example of the potential of AI tools that can actually improve our access to care and improve our quality of care in resource-limited settings.
“The authors found incredibly high sensitivity but, it’s also important for us to understand what the false-positive rate is of their screening tool,” Sandhu said. “That’s especially important when we think about the potential resource utilization of false-positives and how that would lead to potentially additional echocardiograms and confirmatory testing.
“The importance of these findings is incredibly clear and now we need to move to understand how to implement this into subsequent care. How do we take a tool like this in a resource-limited setting and actually roll it out broadly? Most importantly, after individuals are identified, how do we integrate it into subsequent care workstreams to provide care for those individuals? We’ve made a lot of advances with being able to use AI tools to identify individuals at risk, but it’s just as important as figuring out what the next steps are to subsequently improve their outcomes,” Sandhu said.
Perspective
Back to Top
Clyde W. Yancy, MD, MSc, MACC, FAHA, MACP, FHFSA
The SPEC-AI Nigeria trial draws attention to the importance of postpartum cardiomyopathy in general and especially in a population in Africa. If we extrapolate from this work that persons of African ancestry may have a similar burden, then our ongoing contemporary concerns regarding disproportionate burden of maternal morbidity and mortality, at the time of pregnancy, in women of color, some of whom have African ancestors, become illuminated. Highlighting this problem with whatever technology, in this case AI, helps us appreciate new biology and perhaps a root cause or a root source for a problem that has been impenetrable and unsolvable here in this country.
In my opinion, this technology is brilliant in simplicity and execution. Simple because it is reduced to this: an electrocardiogram and a phonocardiogram. From that, the execution of data science approximates a functional assessment of ventricular function. I think that is actually phenomenal. It’s almost a misrepresentation that it is called a stethoscope.
With that in mind, we see the capacity of the technology. This is a kind of technology where I’m OK with false positives. I’d rather have a high false-positive rate capturing/screening as many women as possible that might benefit from early therapy than to have false negatives and missed opportunities to intervene. So, there’s a lot of promise, but we need lots more data as well.
But, I am a bit anxious about costs when we think about low- and middle-income countries. Particularly if we extrapolate that technology to this country, what will be the cost here? We have a compelling need to improve our ability to manage women in this peripartum state. I hope the research continues.
Collapse
Adedinsewo DA, et al. LBS.08. Artificial intelligence at the bedside. Presented at: American Heart Association Scientific Sessions; Nov. 11-13, 2023; Philadelphia.
Click Here to Manage Email Alerts