Artificial Intelligence in Cardiac Illness

“ We could build in a system that would take every missed diagnosis — a patient who developed lung cancer eventually — and feed it back to the machine. We could ask, What did you miss here? Could you refine the diagnosis? There’s no such system for a human radiologist. If you miss something, and a patient develops cancer five years later, there’s no systematic routine that tells you how to correct yourself. But you could build in a system to teach the computer to achieve exactly that.”

— Geoffrey Everest Hinton, The New Yorker

1. Huge Data Sets from Echocardiography! Can Humans Use it all to Differentiate Similar Illnesses?

Figure 1: Data normalization for speckle tracking echocardiography-based data. Each measurement was subjected first to temporal normalization followed by spatial normalization. All normalized variables were then binned in to quintiles based on similar data derived from normal subjects. CP indicates constrictive pericarditis; LV, left ventricular; RCM, restrictive cardiomyopathy; and STE, speckle tracking echocardiography (Sengupta PP, 2016).

In cardiology, one of the most common imaging modality for patients with suspected cardiac illness is echocardiography. This imaging modality generates several thousand data points during the exam, and with new emerging technology this number is likely to grow. Clinicians are faced with a problem in integrating and interpreting this data. In most cases, only a fraction of the data is used to make a clinical decision. The challenge lies when clinicians are to differentiate between cardiac illnesses that look similar on echocardiography such constrictive pericarditis and restrictive cardiomyopathy.

The human brain to differentiate between the two illnesses will rely on memories from previous and repeated experiences. The brain will also pick on recognizing patterns in order to assess the cardiac illness. In a similar process, a group of researchers worked on a cognitive computing tool that would be able to learn and recall these patterns found in echocardiography data sets. They performed highly complex analyses that allowed them to differentiate between the two illnesses with an accuracy approaching 90%, a discrimination that exceeds the human ability.

Read more: Cognitive Machine-Learning Algorithm for Cardiac Imaging

2. Predicting Outcomes in Cardiac Surgery

Fig 1. Receiver operating characteristic curves showing the performance of EuroSCORE I, EuroSCORE II, and the ML model in predicting post-operative mortality (Allyn J., 2017).

Cardiovascular disease imposes a huge burden in terms of morbidity, mortality, disability, functional decline, and health care cost. Taking care of cardiac ill patients before, during and after surgery have raised public health concerns because of the high risk of complications. As a result, predicting the outcome of cardiac surgery through risk stratification models has become important to aid in the clinical decision for surgery versus conservative treatment. Two main models in use today are the European EuroSCORE II and the American Society of Thoracic Surgeons (STS) scores. However, many studies have pointed the limit of these scoring systems. Machine Learning, a subfield of artificial intelligence, demonstrated superiority over the traditional risk stratification models.

A group of researchers have looked at comparing the different models of prediction of mortality after cardiac surgery, including EuroSCORE II, a logistic regression model and a machine-learning (ML) model, using ROC and DCA. The machine-learning model outperformed the EuroSCORE I and the EuroSCORE II with area under curve values of 0.719 (EuroSCORE I), 0.737 (EuroSCORE II), and 0.795 (ML model).

Read more: A Comparison of a Machine Learning Model with EuroSCORE II in Predicting Mortality after Elective Cardiac Surgery: A Decision Curve Analysis

3. Machine Learning in the World of Cardiac Magnetic Resonance Imaging — 3D Motion Analysis of the Heart

Figure 1: Example of computational modeling for a patient with idiopathic pulmonary arterial hypertension. A, Cine MR images were segmented by using prior knowledge from a set of disease-specific atlases. Here, the intensity image in the short-axis of the heart is overlaid with labels for left ventricular blood pool (red), myocardium (green), RV blood pool (yellow), and free wall (blue). B, A 3D model at end-diastole (gray) and end-systole (blue, right ventricle; and red, left ventricle) was used to determine the direction and magnitude of systolic excursion at each corresponding anatomic point in the mesh by using a deformable motion model. C,A statistical model of RV endocardial motion was used for feature selection to determine functional patterns associated with survival (relative weightings shown for the RV free wall) (Dawes TJW., 2017)

One kind of hypertension is pulmonary hypertension that affects arteries of the lungs and is often associated with impaired exercise tolerance and shortness of breath. The disease progresses fast, ultimately leading to heart failure and death. To diagnose and investigate the disease, invasive techniques are in use by inserting a catheter in the pulmonary artery to measure right-sided heart pressure. Cardiac magnetic resonance (CMR) imaging is also routinely done to assess cardiac function.

A radiologist will read the CMR image providing information about right-sided heart failure. However, the most important prognostic features are complex and require other methods that can integrate data and identifying those that are relevant.

Therefore, computational imaging analysis combined with machine learning has been utilized to filter through prognostic features and pick on those that are meaningful to predict eventual right-sided heart failure and death. These researchers developed a machine-learning algorithm that looks at right ventricle function by analyzing motion of the heart from CMR images and demonstrated a more accurate prediction of patient outcomes in pulmonary hypertension.

Read more: Machine Learning of Three-dimensional Right Ventricular Motion Enables Outcome Prediction in Pulmonary Hypertension: A Cardiac MR Imaging Study