Strain measures myocardial deformation throughout the cardiac cycle. Tissue Doppler imaging calculates strain by measuring tissue velocities at a fixed point. Speckle tracking uses ultrasonic backscatter (speckles or kernels) to track tissue movement in 2 and 3 dimensions. Speckle tracking has gained acceptance because it allows more comprehensive vector analysis and is less dependent on the angle of the ultrasound beam. Longitudinal, radial, circumferential, and rotational vectors are used to quantify segmental, territorial, and global longitudinal strain. Global longitudinal strain is the most frequently reported measurement. Measurements vary by vendor.1Farsalinos K.E. Daraban A.M. Unlu S. Thomas J.D. Badano L.P. Voigt J.U. Head-to-head comparison of global longitudinal strain measurements among nine different vendors: the EACVI/ASE Inter-Vendor Comparison Study.J Am Soc Echocardiogr. 2015; 28: 1171-1181.e2Abstract Full Text Full Text PDF PubMed Scopus (333) Google Scholar This limitation largely results from postacquisition software, artificial intelligence, and machine learning.2Negishi K. Lucas S. Negishi T. Hamilton J. Marwick T.H. What is the primary source of discordance in strain measurement between vendors: imaging or analysis?.Ultrasound Med Biol. 2013; 39: 714-720Abstract Full Text Full Text PDF PubMed Scopus (66) Google Scholar Machine learning (ML) can be broadly subdivided into supervised and unsupervised learning.3Gandhi S. Mosleh W. Shen J. Chow C.M. Automation, machine learning, and artificial intelligence in echocardiography: a brave new world.Echocardiography. 2018; 35: 1402-1418Crossref PubMed Scopus (40) Google Scholar Supervised learning includes human-derived algorithms. Each step is typically hand coded and can be easily analyzed for scientific validity. Unsupervised ML is more effective for pattern recognition. It is especially valuable for interpretation of images. Deep learning is a subcategory of unsupervised ML that uses electronic neural networks to mimic the human brain. The machine is presented with most of a large data set for training. The remaining data are used to test and validate training. Deep learning is ideal for strain. Unfortunately, there are limitations. The algorithms generated by deep learning cannot be interpreted by humans. They are not scientifically transparent. We can only analyze results. Each machine-derived strain measurement depends on software configuration and the group of patients originally presented to the machine. The latter is a major limitation in the global arena.4Asch F.M. Miyoshi T. Addetia K. et al.Similarities and differences in left ventricular size and function among races and nationalities: results of the World Alliance Societies of Echocardiography Normal Values Study.J Am Soc Echocardiogr. 2019; 32 (1396.e2-1406.e2)Abstract Full Text Full Text PDF Scopus (38) Google Scholar Finally, it is essential to recognize that there are statistically significant differences for normal values between vendors and these values may change with software updates.1Farsalinos K.E. Daraban A.M. Unlu S. Thomas J.D. Badano L.P. Voigt J.U. Head-to-head comparison of global longitudinal strain measurements among nine different vendors: the EACVI/ASE Inter-Vendor Comparison Study.J Am Soc Echocardiogr. 2015; 28: 1171-1181.e2Abstract Full Text Full Text PDF PubMed Scopus (333) Google Scholar In this issue of The Annals of Thoracic Surgery, Kislitsina and associates5Kislitsina O.N. Thomas J.D. Crawford E. et al.Predictors of left ventricular dysfunction after surgery for degenerative mitral regurgitation.Ann Thorac Surg. 2020; 109: 669-677Abstract Full Text Full Text PDF PubMed Scopus (8) Google Scholar provide further evidence to support the use of strain to predict ventricular dysfunction after mitral intervention. The European Association of Cardiovascular Imaging/American Society of Echocardiography Industry Task Force consensus statement has reduced intervendor differences.6Potter E. Marwick T.H. Assessment of left ventricular function by echocardiography: the case for routinely adding global longitudinal strain to ejection fraction.JACC Cardiovasc Imaging. 2018; 11: 260-274Crossref PubMed Scopus (149) Google Scholar However, results still vary, especially with older software. As strain and other forms of artificial intelligence are incorporated into clinical practice, we must remain vigilant and recognize the limitations and risks of unsupervised machine learning and uninterpretable algorithms, and the results they generate. Predictors of Left Ventricular Dysfunction After Surgery for Degenerative Mitral RegurgitationThe Annals of Thoracic SurgeryVol. 109Issue 3PreviewThis study was performed to determine whether strain can supplement the ability of left ventricular (LV) ejection fraction (LVEF) to predict postoperative ventricular dysfunction in patients undergoing mitral valve surgery for degenerative mitral regurgitation (DMR). Full-Text PDF