1
|
Kennedy L, Bates K, Therrien J, Grossman Y, Kodaira M, Pressacco J, Rosati A, Dagenais F, Leask RL, Lachapelle K. Thoracic Aortic Aneurysm Risk Assessment: A Machine Learning Approach. JACC. ADVANCES 2023; 2:100637. [PMID: 38938360 PMCID: PMC11198590 DOI: 10.1016/j.jacadv.2023.100637] [Citation(s) in RCA: 1] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 09/20/2022] [Revised: 05/25/2023] [Accepted: 06/20/2023] [Indexed: 06/29/2024]
Abstract
Background Traditional methods of risk assessment for thoracic aortic aneurysm (TAA) based on aneurysm size alone have been called into question as being unreliable in predicting complications. Biomechanical function of aortic tissue may be a better predictor of risk, but it is difficult to determine in vivo. Objectives This study investigates using a machine learning (ML) model as a correlative measure of energy loss, a measure of TAA biomechanical function. Methods Biaxial tensile testing was performed on resected TAA tissue collected from patients undergoing surgery. The energy loss of the tissue was calculated and used as the representative output. Input parameters were collected from clinical assessments including observations from medical scans and genetic paneling. Four ML algorithms including Gaussian process regression were trained in Matlab. Results A total of 158 patients were considered (mean age 62 years, range 22-89 years, 78% male), including 11 healthy controls. The mean ascending aortic diameter was 47 ± 10 mm, with 46% having a bicuspid aortic valve. The best-performing model was found to give a greater correlative measure to energy loss (R2 = 0.63) than the surprisingly poor performance of aortic diameter (R2 = 0.26) and indexed aortic size (R2 = 0.32). An echocardiogram-derived stiffness metric was investigated on a smaller subcohort of 67 patients as an additional input, improving the correlative performance from R2 = 0.46 to R2 = 0.62. Conclusions A preliminary set of models demonstrated the ability of a ML algorithm to improve prediction of the mechanical function of TAA tissue. This model can use clinical data to provide additional information for risk stratification.
Collapse
Affiliation(s)
- Lauren Kennedy
- Department of Chemical Engineering, McGill University, Montreal, Quebec, Canada
- Division of Cardiac Surgery, McGill University Health Centre, Montreal, Quebec, Canada
| | - Kevin Bates
- Department of Chemical Engineering, McGill University, Montreal, Quebec, Canada
- Division of Cardiac Surgery, McGill University Health Centre, Montreal, Quebec, Canada
| | - Judith Therrien
- Division of Cardiology, McGill University Health Centre, Montreal, Quebec, Canada
| | - Yoni Grossman
- Division of Cardiology, McGill University Health Centre, Montreal, Quebec, Canada
| | - Masaki Kodaira
- Division of Cardiology, McGill University Health Centre, Montreal, Quebec, Canada
| | - Josephine Pressacco
- Division of Diagnostic Radiology, McGill University Health Centre, Montreal, Quebec, Canada
| | - Anthony Rosati
- Department of Chemical Engineering, McGill University, Montreal, Quebec, Canada
- Division of Cardiac Surgery, McGill University Health Centre, Montreal, Quebec, Canada
| | - François Dagenais
- Institut Universitaire de Cardiologie et de Pneumologie de Québec, Québec, Quebec, Canada
| | - Richard L. Leask
- Department of Chemical Engineering, McGill University, Montreal, Quebec, Canada
| | - Kevin Lachapelle
- Division of Cardiac Surgery, McGill University Health Centre, Montreal, Quebec, Canada
| |
Collapse
|
2
|
Emmott A, Alzahrani H, Alreshidan M, Therrien J, Leask RL, Lachapelle K. Transesophageal echocardiographic strain imaging predicts aortic biomechanics: Beyond diameter. J Thorac Cardiovasc Surg 2018; 156:503-512.e1. [DOI: 10.1016/j.jtcvs.2018.01.107] [Citation(s) in RCA: 14] [Impact Index Per Article: 2.3] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 11/08/2017] [Revised: 01/09/2018] [Accepted: 01/16/2018] [Indexed: 02/07/2023]
|
3
|
Emmott A, Garcia J, Chung J, Lachapelle K, El-Hamamsy I, Mongrain R, Cartier R, Leask RL. Biomechanics of the Ascending Thoracic Aorta: A Clinical Perspective on Engineering Data. Can J Cardiol 2016; 32:35-47. [DOI: 10.1016/j.cjca.2015.10.015] [Citation(s) in RCA: 28] [Impact Index Per Article: 3.5] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/09/2015] [Revised: 10/17/2015] [Accepted: 10/18/2015] [Indexed: 12/14/2022] Open
|
4
|
Sokolis DP. Effects of aneurysm on the directional, regional, and layer distribution of residual strains in ascending thoracic aorta. J Mech Behav Biomed Mater 2015; 46:229-43. [DOI: 10.1016/j.jmbbm.2015.01.024] [Citation(s) in RCA: 32] [Impact Index Per Article: 3.6] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/09/2014] [Revised: 01/15/2015] [Accepted: 01/28/2015] [Indexed: 11/28/2022]
|