Cardiac Imaging

May 2023
EN.510.433 Medical Image Analysis

Objectives:

Given a sequence of cardiac ultrasound images(2-channel and 4-channel views), use medical image analysis techniques to locate the left ventricle (LV) of the heart

Then, re-construct a 3D volume of the LV according to the orthogonal image views, then estimate the stroke volumes during a full heartbeat

Evaluate the algorithm on test data of 10 new patients (as shown in the animations).

Skills Applied:

Computer vision implementation wtih Python: dynamic thresholding, image denoising,template matching, and interpolation

Algorithm design and testing

GitHub, CI/CD data pipeline

Comments:

The project was largely open-ended, allowing me to apply any image analysis technique from course materials. This project adopted classic computer vision techniques , rather than ML-based models, where I gained a much deeper understanding of their practical applications. In the future, I'm curious about ML-based approachesas well, where the model setting accounts for a-priori constraints.