Applications of Artificial Intelligence for the Prediction and Diagnosis of Cancer Therapy-Related Cardiac Dysfunction in Oncology Patients

Figures

• 

  Figure 1. Example of previously validated artificial intelligence models applied to single 12-lead electrocardiogram to predict risk of (A) low left ventricular ejection fraction and (B) new atrial fibrillation [24,26,27]. Blue dots represent individual ECG recordings with a risk score below the pre-determined cutoff (reflected as the red dashed line), red triangle represent individual ECG recordings with risk scores above the pre-determined cutoff, and purple X represents the currently selected ECG.

• 

  Figure 2. Example of AI model for automated measurement of left ventricular ejection fraction (LVEF), reported on apical 4-chamber (A4C) and apical 2-chamber (A2C) views. Model utilizes endocardial contour auto-generation to trace end-diastolic LV cavity volume (EDV), subsequently reporting LVEF by Simpson’s method.

• 

  Figure 3. Example of artificial intelligence model to automate global longitudinal strain (GLS) on transthoracic echocardiogram. Through automated contouring of endocardial and epicardial borders on apical 4-chamber (A4C), apical 2-chamber (A2C), and apical 3-chamber (A3C) views, generation of peak endocardial and mid strain is presented.

• 

  Figure 4. A machine learning model applied to a transthoracic echocardiographic 4-chmaber view to help differentiate constrictive pericardial disease from cardiac amyloidosis, which is the prototype for restrictive cardiac pathology; the GradCam images demonstrate focus of the AI model on the septum which has abnormal motion in constrictive pericarditis (panel (A)), and mitral annulus in cardiac amyloidosis (panel (B)), features which are important in the clinical differentiation of these entities. The generated heatmap depicts the focus of AI, whereby the red areas of those of highest focus. This figure is reproduced with permission [79].

• 

  Figure 5. Artificial intelligence automation of vessel segmentation and plaque quantification in case of left anterior descending coronary artery stenosis, with noncalcified, soft, cholesterol rich plaque is denoted in red and calcified plaque in yellow. Figure reproduced with permission [103].