Pooya Eini, MD

The heart

Imaging intelligence for chamber and valve

Echocardiography, cardiac magnetic resonance, and cardiac CT produce quantitative signals that are rich yet operator dependent. My work synthesizes the evidence for automated assessment of right ventricular function and valvular regurgitation, and for models that anticipate atrial fibrillation and heart failure before they declare themselves clinically. In original work, I develop explainable models that phenotype mitral valve prolapse and read cardiovascular diagnoses from pediatric electrocardiograms.

• Right ventricular function
• Mitral regurgitation
• Tricuspid regurgitation
• Postoperative atrial fibrillation
• Heart failure risk
• Cardiac MRI
• Mitral valve prolapse phenotyping
• Pediatric ECG

The great vessels

Aorta, carotid bifurcation, and cerebrovascular risk

Conduit arteries fail catastrophically, yet rarely without warning. I study machine learning for detecting and characterizing carotid plaque on ultrasound, for predicting stroke in hypertensive populations, and for diagnosing acute aortic dissection, together with the circulating microRNA signatures that may identify abdominal aortic aneurysm before rupture.

• Carotid plaque
• Stroke prediction
• Aortic dissection
• Abdominal aortic aneurysm
• Vascular ultrasound

Pulmonary microcirculation

Embolism, venous thrombosis, and alveolar flooding

At the alveolar capillary interface, obstruction and congestion decide survival. Oxygenated blood (red) crosses the capillary bed and returns desaturated (blue); an embolus silences everything downstream, and failing hemodynamics flood the air spaces. I have evaluated machine learning for mortality prediction in acute pulmonary embolism, for detecting embolism in patients with pneumonia, for recurrence after venous thromboembolism, and for automated detection of pulmonary edema.

• Pulmonary embolism
• Recurrent VTE
• Pulmonary edema
• Risk stratification

Inside the lumen

Method: every cell is a data point

Blood carries the signal; methodology decides whether it is read correctly. Across this body of work, algorithms are judged on discrimination, calibration, and validation rather than headline accuracy, and pooled estimates are only as credible as the studies beneath them.

Atherosclerosis

From fatty streak to flow-limiting plaque

Atherosclerosis begins when apolipoprotein B containing lipoproteins are retained in the intima. Recruited monocytes engulf oxidized lipid and become foam cells, and the lesion matures into a fibroatheroma whose fibrous cap either stabilizes or thins. Notice how red cells accelerate through the stenosis. My work examines how artificial intelligence reads plaque on intravascular ultrasound, how coronary CT angiography grades left main stenosis against IVUS, and how immune activation accelerates atherogenesis.

• Intravascular ultrasound
• Coronary CT angiography
• Left main disease
• Carotid plaque
• Inflammation

Rupture and thrombosis

The final common pathway

When a thin fibrous cap ruptures, tissue factor within the necrotic core meets flowing blood. Platelets adhere and aggregate, fibrin polymerizes into a mesh, and erythrocytes are trapped within it. The same biology underlies myocardial infarction, ischemic stroke, and venous thromboembolism, the events that my prognostic work seeks to anticipate.