Related Experiment Video
Updated: Dec 12, 2025
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
16.0K
A Network-Based "Phenomics" Approach for Discovering Patient Subtypes From High-Throughput Cardiac Imaging Data
Jung Sun Cho1, Sirish Shrestha2, Nobuyuki Kagiyama2
1West Virginia University Heart & Vascular Institute, Morgantown, West Virginia; Division of Cardiology, Daejeon St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea.
JACC. Cardiovascular Imaging
|August 9, 2020
Summary
This study introduces a computational method to identify patient subgroups using echocardiography data. The approach effectively distinguishes cardiac phenogroups, aiding in personalized cardiovascular care and outcome prediction.
Area of Science:
- Cardiology
- Computational Biology
- Medical Imaging
Background:
- Multiomics data integration offers new avenues for classifying cardiovascular states and tailoring therapies.
- Heterogeneous data sources present challenges in comprehensive patient classification.
Purpose of the Study:
- To develop and validate a computational method for patient phenotyping using echocardiography.
- To identify distinct patient subgroups based on multiple echocardiographic parameters.
- To predict clinical characteristics and outcomes of identified patient subgroups.
Main Methods:
- Utilized 42 echocardiography features from 297 patients, including 2D, Doppler, speckle-tracking, and vector flow mapping.
- Developed a similarity network to delineate patient phenotypes.
- Trained neural network models to discriminate between phenotypic presentations.
Main Results:
- Identified 4 distinct patient clusters (phenogroups) with varying clinical presentations and outcomes.
- Cluster IV showed significantly higher rates of advanced heart failure and major adverse cardiac events.
- Neural network models achieved high accuracy (AUC 0.82-0.99) in predicting patient clusters.
Conclusions:
- Automated computational phenotyping effectively integrates multidimensional echocardiographic data.
- This method can identify distinct cardiac phenogroups based on clinical features, function, and outcomes.
- The approach facilitates personalized medicine by stratifying patients for targeted therapies.
