Detecting Incipient Heart Failure in Asymptomatic Patients with Normal Ejection Fraction and comparisons with patients with heart failure and preserved ejection fraction using TimeSformer for classifying Echocardiography videos.
Adhyapak, S. M.; Menon, P.
Show abstract
BackgroundRecently deep learning models have helped differentiate echocardiography images of patients with heart failure with preserved ejection fraction (HFpEF) from normal controls. Our aim was to develop a model capable of detecting early signs of heart failure in asymptomatic patients with a normal ejection fraction. MethodsWe employed the TimeSformer, a video transformer model that classifies video data using a novel attentionbased mechanism. This self-attention mechanism that diverges from traditional convolutional neural networks (CNNs). It focuses on relevant parts of the video across both space and time, split into spatial attention, which processes each frame individually, and temporal attention, which integrates information across different frames. The training and validation of the TimeSformer model were conducted on the same dataset of echocardiography videos from 50 normal controls and 80 patients diagnosed with HFpEF, employing 5-fold cross-validation to ensure robust performance evaluation. ResultsThe TimeSformer model effectively identified HFpEF in patients, as all diagnosed with HFpEF were flagged as abnormal. The echocardiography assessment, along with NT pro BNP levels, supported the diagnoses, with patients showing NT pro BNP levels of 1016{+/-}32 pg/ml. Conversely, 26 out of 50 normal controls were correctly identified. While 24 normal controls were identified as abnormal. Their NT pro BNP levels were 52{+/-}12 pg/ml. At 2 years 8 (33.3%) controls identified as abnormal developed symptoms of heart failure with NT pro BNP levels of 800{+/-}41 pg/ml. ConclusionsThe TimeSformer model demonstrated capability in identifying subtle deviations indicative of incipient heart failure in videos of normal controls, despite normal NT pro BNP levels. A significant number of these controls developed heart failure with elevated NT pro BNP levels at 2 years.
Matching journals
The top 8 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Predictors and outcomes of Cardiac Dyssynchrony among patients with heart failure attending Benjamin Mkapa Hospital in Dodoma, central Tanzania: A protocol of prospective-longitudinal study 97%
- Assessing Red Blood Cell Distribution Width in Vietnamese Heart Failure Patients: A Cross-Sectional Study 96%
- Predicting 30-Day and 1-Year Mortality in Heart Failure with Preserved Ejection Fraction (HFpEF) 95%
Similar papers in this journal
- Screening for Right Ventricular Dysfunction in the Emergency Department Using a Smartphone ECG Analysis Application: An External Validation Study with Acute Pulmonary Embolism Patients 95%
- Myocardial performance improvement after iron replacement in heart failure patients: The IRON-PATH II echo-substudy 95%
- Centhaquine Increases Stroke Volume and Cardiac Output in Patients with Hypovolemic Shock 93%
Similar papers in this journal
- Detecting Heart Failure using novel bio-signals and a Knowledge Enhanced Neural Network 98%
- Detection and removal of pacing artifacts prior to automated analysis of 12-lead ECG 95%
- Hemodynamics Indicates Differences Between Patients With And Without A Stroke Outcome After Left Ventricular Assist Device Implantation 93%
Similar papers in this journal
- Automated Echocardiographic Detection of Mitral Valve Prolapse and Mitral Regurgitation with Video-based Artificial Intelligence Algorithms 95%
- Using ECG Machine Learning for Detection of Cardiovascular Disease in African American Men and Women: the Jackson Heart Study 94%
- Deep Learning-Based Multi-View Echocardiographic Framework for Comprehensive Diagnosis of Pericardial Disease 93%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.