PhD Scientific Days 2026
Budapest, 16-18 June 2026
Cardiovascular Medicine and Research 1.
Deep learning-enabled echocardiographic assessment of biventricular ejection fractions
Előadó neve
Szijártó, Ádám, MSc
Neptune code
C4442F
Előadó munkahelye
Semmelweis Egyetem
Előadó telefonszáma
+36203666935
Előadó e-mail címe
sz.adam1996@gmail.com
Az előadás címe
Deep learning-enabled echocardiographic assessment of biventricular ejection fractions
Szerző(k) neve és munkahelye
MSc Ádám Szijártó1
1: Semmelweis Egyetem
Bemutatás módja
Szóbeli
Szekció
Cardiovascular Medicine and Research 1.
Language of the presentation
English
Preferred session
Cardiovascular Medicine and Research
Összefoglaló szövege
Introduction
Accurate assessment of left ventricular (LV) and right ventricular (RV) systolic function via ejection fractions (LVEF and RVEF) is crucial in cardiology, guiding diagnosis, therapy, and prognosis. While 2D echocardiography (2DE) is widely used for its accessibility, it requires multiple views for LVEF and cannot reliably quantify RVEF. 3D echocardiography (3DE) provides superior biventricular evaluation but is underutilized due to training demands, time constraints, limited probes/software, and poor acoustic windows.
Aims
This study develops QUEST-EF (QUantification of Echocardiographic STudies—Ejection Fraction), a dual-task deep learning (DL) model to predict 3DE-derived LVEF and RVEF from a single apical four-chamber (A4C) 2DE video. It aims to validate performance across diverse cardiac diseases, geographies, and populations.
Methods
QUEST-EF was trained in two steps. First pre-trained in a self-supervised manner on 29,876 unlabeled A4C videos, supervised for LVEF on EchoNet-Dynamic (10,030 videos) and a dual-center 3DE dataset (5,341 videos) and for RVEF on the latter. Beyond testing QUEST-EF internally on 20% of the dual-centre dataset, its performance was also externally validated in patients with acquired and congenital cardiac diseases from four international centers and healthy adults from six continents enrolled in the WASE study.
Results
Internally, QUEST-EF achieved LVEF MAE 4.56% (95% CI 4.11–5.09%) and RVEF MAE 4.82% (4.33–5.41%), with AUCs 0.94 and 0.88 for dysfunction. Externally, LVEF MAE 4.60% (4.41–4.80%), RVEF MAE 5.42% (5.18–5.65%). Predictions associated with composite heart failure/death (n=187; LVEF aHR 0.945, P=0.002; RVEF aHR 0.927, P=0.006), independent of age and sex, and 10-year mortality (n=1,166; LVEF aHR 0.947, P=0.001; RVEF aHR 0.877, P=0.001), independent of the Framingham Risk Score and E/e′ ratio.
Conclusion
QUEST-EF enables rapid, accurate, vendor-independent biventricular EF prediction from routine A4C views, robust across pathologies and regions.
Funding
Semmelweis Egyetem Predoktori Ösztöndíj
University
Semmelweis University
Supervisor
Dr. Kovács Attila, Dr. Tokodi Márton
Publication of my abstract
I do not give consent to the publication of my abstract on the website of the congress.
phd.section.field
after finishing doctoral studies with absolutorium (PhD)
Kind
Szabad
Status
elfogadva
Accepted presentation method
szóbeli
Előadás fájl jóváhagyás
jóváhagyta
Előadó
9729
Start
16:15
End
16:25