AI algorithm shows promise for early detection of PH: Study
Tool uses ECG data to find disease at early stages
Anumana‘s artificial intelligence (AI) algorithm shows promise for the early detection of pulmonary hypertension (PH), a study suggested.
The PH Early Detection Algorithm, which analyzes data collected from a routine electrocardiogram (ECG) heart test, was developed by scientists at Anumana, Janssen Research and Development, the Mayo Clinic, and the Vanderbilt University Medical Center.
“The promising data from our study suggest that an AI algorithm has the potential to non-invasively detect PH at an early stage using standard ECGs,” Hilary DuBrock, MD, a pulmonologist at Mayo Clinic and co-first author of the study, said in a company press release. “This finding marks a significant step forward in the care and management of PH patients who often have a long diagnostic journey,”
The study, “An electrocardiogram-based AI algorithm for early detection of pulmonary hypertension,” was published in the European Respiratory Journal.
In PH, abnormally high pressure in the lungs’ blood vessels creates a strain on the heart. The fact that PH symptoms, such as fatigue or breathlessness, are often seen in other conditions makes diagnosis difficult and invasive tests necessary. And diagnostic delays — PH patients are often diagnosed more than two years after symptom onset — increase the risk of worse outcomes.
ECG is a standard, noninvasive test of the heart’s electrical activity and structure. It is common in many primary care and emergency room settings.
Algorithm uses reference data along with ECG recordings
Anumana’s PH-EDA is designed to enhance the predictive power of ECG recordings. The algorithm uses data from a reference platform that contains information from more than 6 million de-identified patient records and more than 8 million ECG recordings.
The researchers used data from ECG, right heart catheterization — a procedure that inserts a catheter into a vein in the neck or groin to measure the pressure in the heart and lungs — and ultrasounds from 39,823 patients likely to have PH and 219,404 patients without PH, who served as controls, from Mayo Clinic. The findings were then confirmed in a second group of 6,045 PH-likely patients and 24,256 controls from Vanderbilt.
The ability to identify PH was measured by the area under the receiver operating characteristic curve (AUC). This measure tells how well a given parameter can differentiate between two groups (in this case, those with PH and those without). AUC values range from 0.5 to 1, with higher numbers indicating a better ability to differentiate.
The PH-EDA algorithm showed promise for diagnosing PH, with an AUC of 0.92 using the data collected at Mayo Clinic and 0.88 in the Vanderbilt dataset. The algorithm also fared well when using ECGs taken six to 18 months before a PH diagnosis, and also up to five years prior to diagnosis.
“These new data underscore the potential of AI algorithms to empower clinicians to uncover diseases earlier, improve patient outcomes and bring us one step closer to our mission to transform cardiac care,” said Maulik Nanavaty, PhD, CEO of Anumana. “We’re continuing to work closely with our partners to further clinically validate this much-needed algorithm, which can help clinicians worldwide get PH patients into treatment sooner to address symptoms and prolong life.”
The U.S. Food and Drug Administration (FDA) designated PH-EDA a breakthrough device in May 2022. The designation is designed to expedite the development and regulatory review of medical devices deemed to have potential to treat or help diagnose life-threatening or irreversibly debilitating diseases.
“The PH-EDA can detect PH at diagnosis and 6–18 months prior, demonstrating the potential to accelerate diagnosis and management of this debilitating disease,” the researchers wrote.
About the Author
