Voice Biomarker May Aid in Remote Monitoring for Heart Failure Risk

Inês Martins, PhD avatar

by Inês Martins, PhD |

Share this article:

Share article via email
voice biomarker

A composite assessment of voice features may be able to monitor changes in the pulmonary arteries of people with pulmonary hypertension (PH) who are at risk of heart failure, helping identify those more likely to experience poor outcomes, a study found.

While these patients usually require invasive tests to monitor their condition, this research lays the groundwork for remote monitoring, the researchers said. That likely would reduce hospital admissions and improve the quality of life in this patient population.

The study, “Non-invasive vocal biomarker is associated with pulmonary hypertension,” was published in the journal PLOS One.

PH is a disease associated with high blood pressure in the pulmonary arteries — the blood vessels that supply the lungs — which requires the heart to pump harder for blood to flow through them. This excess work puts the heart under extra strain, which may cause heart failure later in life. 

To monitor heart failure, people with PH normally undergo a procedure called cardiac catheterization. It consists of guiding a special catheter into the heart to see how well it is pumping and to measure blood pressure in the main vessels in the lung.

While the approach helps doctors determine the severity of heart failure, and to modify treatments accordingly, its invasive nature is a clear downside. Thus, there is a need for less invasive approaches to help monitor these patients.

Measuring changes in voice has emerged as a promising approach to identify people with heart diseases. Researchers at the Mayo College of Medicine and Science, in Minnesota, have previously demonstrated that a voice biomarker is associated with an increased risk of death or hospitalization due to heart failure.

Now, the same team examined if this voice biomarker also could be associated with changes in pulmonary artery pressure and resistance that are reflective of pulmonary vascular disease.

Their study included 83 patients who were referred for cardiac catheterization between January 2017 and December 2018. Among them, 35 were diagnosed with PH, 18 with left ventricle failure, and eight with no disease.

The participants’ mean age was 61.6 years, and a majority (55.4%) were women. Most of the participants also had other medical conditions, such as diabetes, high blood pressure, abnormal levels of fats, and obesity.

For the voice biomarker, participants were asked to speak aloud into a recording device for three separate voice recordings, each lasting 30 seconds. In the first recording, patients read a pre-specified text. In the second, they were asked to describe a positive emotional experience. Then, in the third recording, they were tasked with describing a negative emotional experience.

Using machine learning and artificial intelligence techniques, the researchers were able to examine 223 acoustic features from each patient’s speech, which were all combined into a single score. The team found that, for each patient, the three recordings usually had a very good agreement in their voice biomarker scores. Thus, the mean, or average scores were used for further assessments.

The researchers divided patients into two main groups, those with moderate or greater PH — defined as a pulmonary arterial pressure (PAP) of 35 mmHg or higher — and those without, who had PAP scores below 35 mmHg.

Overall, the patients with moderate or greater PH had significantly higher levels of the voice biomarker (0.74) compared with those with less severe PH (0.40).

No differences were found in the voice biomarkers when the researchers evaluated patients based on high or low values for other measures. Such measures included pulmonary vascular resistance (PVR), cardiac index (CI), and pulmonary capillary wedge pressure (PCWP), which assesses if PH is caused by heart problems or by underlying lung conditions.

Similarly, statistical analysis showed that an increase by one unit in the voice biomarker increased the odds of having moderate or greater PH. But when patients were divided based on PH related or unrelated to heart failure, the biomarker was no longer associated with moderate or greater PH.

After adjusting for age, sex, high blood pressure, diabetes, and severity of symptoms, the researchers found that the voice biomarker remained associated with moderate or greater PH. Indeed, an increase by one unit in the biomarker score increased the odds of being in this group by 2.3-fold.

This association was significant in the group of patients with PH related to heart failure, but not in those with PH without heart involvement. Similarly, there was a significant association between the biomarker and high vascular resistance in the lung only in patients with heart involvement.

Overall, the analysis showed that the non-invasive voice biomarker is associated with invasive measures of pulmonary arterial and heart function.

“The current study demonstrates an association between a non-invasively measured voice biomarker that can be obtained remotely, and invasive measurements of PH (PAP and PVR) that have a known role in predicting adverse outcomes in HF [heart failure] patients,” the researchers wrote.

“These results may have important and practical clinical implications for telemedicine and remote monitoring of patients with HF and PH,” they concluded.


A Conversation With Rare Disease Advocates