Sam Friedman: ML4CVD: Machine learning for cardiovascular diseases

Cardiovascular diseases (CVD) are the number 1 cause of death globally yet current therapies and methods to assess risk cannot treat or detect many kinds of CVD. Using data from many modalities (e.g. MRI, ECG, and EHR) we are training deep models to segment cardiac anatomy, to predict clinical measures of cardiac health, and to empower GWAS to find genetic associations with CVD.

Gregory Way: Cell painting measurements predict cell health

Cells respond to perturbations in various ways. Traditionally, these responses are measured using a series of individual, image-based assays. However, it turns out that Cell Painting assays measure this information for free: Training simple machine learning models can be used to predict many cell health outcomes.

Beth Cimini: A web application for high content screening

Piximi is a new classification tool being designed here at the Broad, which allows users to classify pictures of objects (cells, nuclei, etc) directly within their browser. The app helps users to annotate their objects into different classes, then train a neural network to distinguish classes; this neural network can then be run on many images on the users' local machine, or the parameters of the network can be easily exported to run on larger machines elsewhere. We will demonstrate the current functionality of the app, discuss features still in development, and discuss the future directions we are contemplating.