CMR-Analysis (including machine learning)
Subin Hwang
BSc Student
The University of Hong Kong, Hong Kong
Subin Hwang
BSc Student
The University of Hong Kong, Hong Kong
Chi Ting Kwan
Medical Student
The University of Hong Kong, Hong Kong
Pui Min Yap
Medical Student
The University of Hong Kong
Hong Kong, Hong Kong
Sau Yung Fung
Medical Student
The University of Hong Kong
Hong Kong, Hong Kong
Hok Shing Tang
Medical Students
The University of Hong Kong, Hong Kong
Wan Wai Vivian Tse
Medical Student
The University of Hong Kong, Hong Kong
Cheuk Nam Felix Kwan
Medical Student
The University of Hong Kong, Hong Kong
Yin Hay Phoebe Chow
Medical Student
The University of Hong Kong, Hong Kong
Nga Ching Yiu
Medical Student
The University of Hong Kong, Hong Kong
Yung Pok Lee
Medical Student
The University of Hong Kong, Hong Kong
Ambrose Ho Tung Fong, MD
Assistant Project Manager
The University of Hong Kong, Hong Kong
Qing-Wen Ren
PhD student
University of Hong Kong, Hong Kong
Mei-Zhen Wu
PhD
The University of Hong Kong, Hong Kong
Eric Yuk Fai Wan
Assistant Professor
The University of Hong Kong, Hong Kong
Ka Chun Kevin Lee, MD
Cardiologist
Ruttonjee Hospital, Hong Kong
Chun Yu Leung, MD
Cardiologist
Tseung Kwan O Hospital, Hong Kong
Andrew Li, MD
Radiology Resident
Queen Mary Hospital HK, Hong Kong
David Montero
Assistant Professor
The University of Hong Kong, Hong Kong
Varut Vardhanabhuti, MD, PhD
Clinical Assistant Professor
The University of Hong Kong, Hong Kong
Jojo Siu Han Hai, MD
Clinical Assistant Professor
The University of Hong Kong
Hong Kong, Hong Kong
Chung-Wah Siu
Professor
The University of Hong Kong, Hong Kong
Hung-Fat Tse, MD
Chair Professor
The University of Hong Kong
Hong Kong, Hong Kong
Dudley Pennell, FSCMR
Professor
Guy's and St Thomas' NHS Foundation Trust
London, England, United Kingdom
Raad Mohiaddin, MD, PhD
Professor/Radiologist
Royal Brompton Hospital, Imaging Centre, Guy’s and St Thomas’ NHS Foundation Trust, United Kingdom and National Heart and Lung Institute, Imperial College London, London, United Kingdom
London, England, United Kingdom
Roxy Senior, MD,DM,FRCP,FESC,FACC
MBBS, MRCP, FESC
Northwick Park Cardiac Research Charity
Harrow,London, England, United Kingdom
Kai-Hang Yiu, MD
Clinical Professor
The University of Hong Kong
Hong Kong, Hong Kong
Ming- Yen Ng, BMBS FRCR FSCMR
Clinical Assistant Professor
The University of Hong Kong
Hong Kong, Hong Kong
Heart failure with preserved ejection fraction (HFpEF) comprises almost 50% of heart failure patients. Despite its similar mortality rate with other heart failure types, HFpEF is difficult to diagnose. Left atrial (LA) strain on echocardiography has been proposed as a diagnostic parameter for HFpEF. Cardiac magnetic resonance (CMR) left atrial strain’s diagnostic accuracy for HFpEF has not been investigated.
Methods:
Patients suspected to have HFpEF were recruited in total from four different centres. CMR, echocardiography and NT-proBNP measurements were performed within 24 hours in all patients. Patients who were indeterminate for HFpEF, had catheter pressure measurements or stress echocardiography to confirm if the patients were HFpEF or non-HFpEF.
Patients are excluded if they have (1) significant coronary artery disease, (2) severe pulmonary disease, (3) cardiomyopathy, and/or (4) constrictive pericarditis from previous cardiac imaging scanning or respiratory tests. Patients are diagnosed as HFpEF following the 2021 European Society of Cardiology Heart Failure guidelines. The diagnostic criteria is (1) symptoms and signs of heart failure, (2) ≥50% left ventricular ejection fraction, (3) cardiac structural and/or functional abnormalities consistent with LV diastolic dysfunction or raised LV filling pressure and (4) raised NT-proBNP level.
Patients with stable angina and normal volunteers were also recruited. Angina patients were chosen as these could ethically undergo invasive LV pressure measurements.
Using cvi42, the patient’s left atrium were contoured on the 2 and 4 chamber cines at the ventricular end-diastolic phase (see figure 1). LA strain values were categorised as reservoir, conduit and booster phase (see figure 2).
Receiver operator characteristic (ROC) analysis of reservoir, conduit and booster strain was performed using SPSS. Area under the curves (AUC) were determined by comparing HFpEF with non-HFpEF patients' LA strain (reservoir, conduit and booster).
Results:
A total of 131 suspected HFpEF patients were prospectively recruited from 2019 to 2022. 53 patients (40.46%) were diagnosed as HFpEF (median age 78 years, interquartile range 8 years) and 38 non-HFpEF (29.01%). 23 angina patients (17.56%) and 17 normal volunteers (12.98%) were recruited.
Left atrial reservoir strain and conduit strain had the highest diagnostic accuracy with AUC of 0.749 and 0.737 respectively. Booster phase strain showed reasonable accuracy with AUC of 0.675 (see figure 3).
Conclusion:
CMR LA reservoir and conduit strain have reasonable accuracy to distinguish between HFpEF patients and non-HFpEF patients.