OR WAIT null SECS
AI-Powered ECG and Echo Could Shrink 2.5-Year Diagnostic Delay in PH
Mohammed Chowdhury, MD, a Temple University cardiologist, says the field is 2 to 5 years from routine clinical integration — but validation and insurance coverage remain the critical hurdles.
Pulmonary hypertension (PH) carries an average diagnostic delay of approximately 2.5 years, driven by the subtlety of early symptoms, low index of suspicion across primary care and subspecialty settings, and a referral chain that often routes patients through multiple providers before the correct diagnosis is reached.1 Approximately 60% of patients arrive at tertiary PH centers already at an advanced stage — a pattern that has direct implications for right ventricular outcomes and survival, given that early treatment initiation is associated with better long-term prognosis.1 Artificial intelligence offers a potential path to shortening that window, and according to Mohammed Andaleeb Chowdhury, MD, Assistant Professor of Medicine at the Lewis Katz School of Medicine at Temple University, Philadelphia, and a member of the Temple Pulmonary Hypertension, Right Heart Failure, and CTEPH Program, the tools to do so are closer to clinical deployment than many providers realize.
Chowdhury discussed AI's role in PH detection at a recent session on the topic, and spoke with HCPLive about how these models work, where they currently stand, and what it will take to get them into routine use. The core mechanism of image-based AI in this context is pattern recognition at a scale and granularity that exceeds human perception: models are trained on large labeled datasets — such as the Mayo Clinic's PH Early Detection Algorithm (PH-EDA), developed using more than 250,000 de-identified ECG records and confirmed against right heart catheterization — and learn to identify signal in the brightness and contrast gradients of waveform data rather than discrete morphological criteria.2 The PH-EDA, developed by Mayo Clinic in collaboration with nference, recently received FDA clearance, and in a multicenter real-world analysis detected more than 85% of patients with pulmonary arterial hypertension and 78% with chronic thromboembolic PH.2 A parallel echocardiography-based AI approach applies the same pattern-recognition principle to imaging data, potentially flagging subtle right heart remodeling before conventional criteria are met.
The challenge, Chowdhury emphasized, is generalizability. Models that perform well on the datasets they were trained on — often from single institutions with their own demographic and clinical profile — frequently encounter performance degradation when exposed to unseen data from diverse populations. Rigorous multicenter validation is therefore the critical next step before any of these tools can be recommended as standard of care. He estimated that, contingent on that validation going smoothly and downstream logistics being resolved, clinical integration could arrive within 2 to 5 years. The most pragmatic near-term application, he suggested, would be an integrated EMR alert system that synthesizes ECG findings, echocardiographic data, and clinical risk factors from the patient record to generate a probability estimate for PH and trigger a referral — particularly valuable for patients in rural or underserved settings where access to PH specialists is limited.
He flags 1 particular point of concern: how will AI-generated diagnoses affect insurance coverage? If an algorithm flags a patient as high-risk for PH and prompts a referral and right heart catheterization, will payers reimburse the downstream workup? The question highlights a dimension of AI implementation — healthcare policy and reimbursement infrastructure — that the clinical and research communities developing these tools often overlook. "That was something I never thought of," Chowdhury said, "…every time there's a new therapy or technology, the insurance doesn't cover it."
Chowdhury has no relevant disclosures.
References
Galie N, Humbert M, Vachiery JL, et al. 2015 ESC/ERS guidelines for the diagnosis and treatment of pulmonary hypertension. Eur Heart J. 2016;37(1):67–119. doi:10.1093/eurheartj/ehv317
DuBrock HM, Wagner TE, Carlson K, et al. An electrocardiogram-based AI algorithm for early detection of pulmonary hypertension. Eur Respir J. 2024;64(1):2400192. doi:10.1183/13993003.00192-2024
