PULMONARY hypertension (PH) is a progressive and potentially fatal condition that can go undetected until it reaches advanced stages. Current diagnostic methods, such as transthoracic echocardiography (TTE), lack both sensitivity and specificity and may require invasive right heart catheterization to confirm a diagnosis. However, a groundbreaking study presented at the CHEST Annual Meeting 2024 in Boston, held from October 6 to 9, introduces a multimodal fusion model (MMF-PH) that could transform the detection of PH.
This innovative deep learning algorithm integrates data from electronic health records, including chest radiographs, ECG, demographics, and TTE findings, to deliver more accurate results than traditional TTE screening methods alone. The study found that MMF-PH demonstrated superior diagnostic performance across various patient groups, showing great promise for clinical use.
The research analyzed a dataset of 4,576 patients with unconfirmed but suspected PH at Beijing Fuwai Hospital, China, between 2019 and 2021. Of these, 2,451 cases were confirmed using right heart catheterization. The algorithm was rigorously trained and validated using both retrospective and prospective data, including external datasets from other medical institutions.
Dr. Zhihua Huang, from Chinese Academy of Medical Sciences and Fuwai Hospital said of the algorithm: “This multimodal deep learning algorithm enables early, accurate identification of pulmonary hypertension and its subtypes, potentially reducing the need for invasive procedures. It has the potential to significantly improve patient management and outcomes while reducing healthcare costs.”
With its potential to improve early diagnosis and guide personalized treatments, the MMF-PH algorithm could revolutionize the clinical management of pulmonary hypertension, offering a less invasive, AI-driven solution.
Reference: Huang Z et al. Revolutionizing pulmonary hypertension detection: a multicenter study on a comprehensive multimodal deep learning approach. Abstract 4029. CHEST Annual Meeting. October 6-9, 2024.
Anaya Malik | AMJ
