Author + information
- Jung Sun Cho, MD, PhD1,2,
- Sirish Shrestha, MS1,
- Nobuyuki Kagiyama, MD, PhD1,
- Lan Hu, RN, MPH1,
- Yasir Abdul Ghaffar, MD1,
- Grace Casaclang-Verzosa, MD1,
- Irfan Zeb, MD1 and
- Partho P. Sengupta, MD, DM, FACC, FASE1,∗ ()
- 1West Virginia University Heart & Vascular Institute, Morgantown, WV, USA
- 2Division of Cardiology, Daejeon St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, Seoul, Korea
- ↵∗Address for correspondence: Partho P. Sengupta, MD, DM Heart & Vascular Institute, West Virginia University, 1 Medical Center Drive, Morgantown, WV 26506-8059. Telephone: +1 304 598 4651 Fax: +1 304 285 1986
Background Recent efforts in collecting multi-omics data open numerous oppurtunities for comprehensive integration of highly heterogenous data to classify a patient's cardiovascular state, eventually leading to tailored therapies. We present a method that focuses on cohort matching algorithms for performing patient-to-patient comparisons along multiple echocardiographic parameters for predicting meaningful patient subgroups.
Methods A total of forty-two echocardiography features including 2-dimensional and Doppler measurements, left ventricular (LV) and atrial speckle-tracking and vector flow mapping data were obtained in 297 patients. We developed a similarity network to delineate distinct patient phenotypes and then trained a neural network models for discriminating the phenotypic presentations.
Results The patient similarity model identified four clusters (I-IV), with patients in each cluster showed distinctive clinical presentations based on ACC/AHA heart failure stage and the occurrence of short-term major adverse cardiac and cerebrovascular events (MACCE). Compared to other clusters, cluster IV had a higher prevalence of stage C or D heart failure (78%, p<0.001), NYHA classes III or IV (61%, p<0.001), and a higher incidence of MACCEs (p<0.001). The neural network model showed robust prediction of patient clusters with area under the receiver operating characteristic curve ranging from 0.82 to 0.99 for the independent hold out validation set.
Conclusion Automated computational methods for phenotyping can be an effective strategy to fuse multidimensional parameters of LV structure and function. It can identify distinct cardiac phenogroups in terms of clinical characteristics, cardiac structure and function, hemodynamics, and outcomes.
- Topological data analysis
- High-dimensional echocardiographic parameters
- Heart failure
- deep phenotype
- deep phenotype
Conflicts of interest: Partho P. Sengupta is a consultant for Heart sciences, Ultromics. Jung Sun Cho, Sirish Shrestha, Nobuyuki Kagiyama, Muhammad Ashraf, Muhammad Khalil, Yasir Abdul Ghaffar, and Grace Casaclang-Verzosa have nothing to disclose.
This study was supported by a grant from Hitachi Healthcare America.
Declaration of Helsinki: Our study complies with the tenets of the Declaration of Helsinki. In addition, the locally appointed ethics committee has approved the research protocol used in this study, and informed consent has been obtained from the subjects (or their legally authorized representative).
- Received November 4, 2019.
- Revision received February 19, 2020.
- Accepted February 20, 2020.
This article requires a subscription or purchase to view the full text. If you are a subscriber or member, click Login or the Subscribe link (top menu above) to access this article.