Predicting Heart Disease

This collection shows an example of using unsupervised machine learning to predict heart disease using images derived from echocardiogram data (work done in conjunction with Paul Leeson and Ross Upton, University of Oxford).

The echocardiogram data is reduced from a 3D to a 2D image using a technique called principal strain analysis and the colours shown in the final image represent blood flow in the heart (dark = poor, red = medium, yellow = good). Using these data we can do Principal Component Analysis (PCA) to project each image into 2D space.

Other Demos You Should Try


DICOM Mammograms

This mammogram dataset consists of 3486 DICOM (Digital Imaging and Communications in Medicine) images. It contains normal, benign, and malignant cases with verified pathology information.


The Muscle Atlas

The Muscle Atlas created for the Institure of Myology, Sorbonne University, contains over a thousand of muscle biopsy images from people of different ages as well as animals.