A new multimodal fusion model (MMF-PH) has outperformed transthoracic echocardiography (TTE) in accurately diagnosing pulmonary hypertension (PH) and identifying its subtypes, as presented at the CHEST Annual Meeting in Boston from 6-9 October. By integrating data from demographics, chest X-rays, electrocardiogram, and transthoracic echocardiography, this model reduces the need for invasive right heart catheterization.
In a study of 4,576 patients, the multimodal fusion model showed a diagnostic accuracy of 96.2% and an AUROC (AUROC is a performance measurement that tells you whether your model is able to correctly rank examples) of 0.994, surpassing transthoracic echocardiography’s performance. Prospective analysis also confirmed the multimodal fusion model’s high sensitivity (95.4%) and specificity (96.9%). Consistent accuracy across various patient subgroups and an ability to differentiate pulmonary hypertension types suggest that the multimodal fusion model could support personalized treatment, improve outcomes, and reduce healthcare costs.
Read more at this link on the CHEST web page
Citation
ZHIHUA HUANG, ZHIHONG LIU, QIN LUO, ZHIHUI ZHAO, WEI ZHAO, QING ZHAO,
REVOLUTIONIZING PULMONARY HYPERTENSION DETECTION: A MULTICENTER STUDY ON A COMPREHENSIVE MULTIMODAL DEEP LEARNING APPROACH,
CHEST, Volume 166, Issue 4, Supplement, 2024, Pages A5899-A5900, ISSN 0012-3692, https://doi.org/10.1016/j.chest.2024.06.3494.
(https://www.sciencedirect.com/science/article/pii/S001236922404296X)
