Echocardiography is a non-invasive diagnostic technique that provides information about cardiac anatomy, function, and hemodynamics. In the last 40 years, the evolution of this technique has made it an integral part of cardiovascular medicine.

There have been many important developments in the field of cardiac ultrasound. In the mid and late 60s, M-mode echocardiography was first used to diagnose pericardial effusion and played a specific role in the measurement of cardiac dimensions and time intervals. In the late 70s and early 80s, a major advancement was the clinical application of 2D echo to provide real-time images of the heart. In addition, Doppler techniques were being used for accurate hemodynamic measurements of aortic valve stenosis and estimation of pulmonary artery pressure. In the early ’80s, a phased-array, 2D transducer was incorporated into a flexible gastroscope and TEE entered the modern age, starting with intraoperative echo. Color doppler was being evaluated in Japan and allowed for assessment of valvular regurgitation.

Size matters: 

LV hypertrophy has been found to be a risk factor for myocardial infarction, ventricular arrhythmias, sudden cardiac death, and heart failure. Accurate non-invasive measurement of the LV mass and volume allow quantitative assessment and follow-up of hypertrophy in patients with hypertension, valvular heart disease, and cardiomyopathies.

In collaboration with his wife, Ellen, Nelson Schiller validated a method of LV mass measurement using real-time 2D echocardiograms and a truncated ellipsoid algorithm.

It can be used to follow adaptive and maladaptive changes occurring in response to normal physiological demands in the case of long-distance runners as well as follow the myocardial response to changes occurring post-renal transplantation in patients with hypertension and renal failure. Also, it could be used in the timing of surgery in patients with chronic valvular insufficiency where the relationship between volume and mass could potentially indicate future decompensation.

Most importantly, 2D echocardiography could be applied to the characterization of disease and give us insight into diabetic cardiomyopathy with a preliminary look at diastolic heart failure.

Since then, 3D echocardiography has been developed and provides intuitive recognition of cardiac structures from any spatial point of view with complete information about heart chamber volumes and function.

The future: Could the association of machine learning to echocardiographic data predict patient outcomes?

Machine learning algorithms can reveal hidden patterns in big data. Unsupervised learning, such as cluster analysis, can evaluate complex interactions among different variables, without knowledge of the outcome. These Artificial Intelligence methods have been used to analyze LV function to characterize HFpEF, discover phenotypic clustering of LV diastolic dysfunction, and classify prognostic categories using exercise echocardiography in patients with HFpEF and phenogroup patients with systolic heart failure to identify response to cardiac resynchronization.

Schiller’s group has studied the association of machine learning-derived phenogroupings of echocardiographic variables with heart failure in stable coronary artery disease in the Heart and Soul Study. In this cohort of 1,000 participants, automated model-based unsupervised clustering on 15 echocardiographic measures identifies four distinct phenogroups, which successfully partition subjects into categories of risk of Heart failure hospitalization. Of relevance to heart failure hospitalization, the higher risk group had the highest history of heart failure, myocardial infarction, and revascularization such as PCI/CABG, the lowest renal function, and the highest CRP and NT-proBNP.