Tech

University of Michigan AI Diagnoses Hard-to-Detect Heart Disease Using ECG Data Alone

April 4, 2026 · From the archive, written in the Professor's voice before our October 2026 redesign from crescendo.ai

What happened

Researchers at the University of Michigan developed an AI model that diagnoses coronary microvascular dysfunction (CMVD) by analyzing standard electrocardiogram (ECG) signals. This model uses deep learning techniques to detect subtle markers of CMVD, which traditional diagnostic methods often miss, enabling non-invasive, early detection.

Why it matters

This work demonstrates how AI can extract clinically relevant insights from routine medical data without invasive tests. For healthcare AI practitioners, it highlights the potential of domain-specific deep learning models to improve diagnostic accuracy and patient outcomes by leveraging underutilized data sources like ECGs.

Who's doing it

The research group at the University of Michigan deployed this AI model in clinical trials, reporting a 15% increase in early CMVD detection rates and reducing reliance on costly imaging techniques.

Try it

1. Obtain ECG datasets from publicly available repositories such as PhysioNet (https://physionet.org).
2. Use a deep learning framework like TensorFlow to implement the University of Michigan's model architecture as described in their 2025 publication.
3. Train and validate the model on your ECG data to identify CMVD markers, aiming to replicate the reported diagnostic accuracy improvements.

Read the original at crescendo.ai