In a paper presented at the Machine Learning for Healthcare conference, researchers from Massachusetts Institute of Technology (MIT) demonstrated a model that automatically learns features predictive of vocal cord disorders.

The features come from a dataset of 100 subjects, each with a week’s worth of voice-monitoring data and several billion samples. The dataset contains signals captured from a small accelerometer sensor mounted on subjects’ necks. In experiments, the model used features automatically extracted from the data to classify, with high accuracy, patients with and without vocal cord nodules (lesions that develop in the larynx, often because of patterns of
voice misuse).

Importantly, the model accomplished this task without a large set of hand-labelled data. The researchers now want to monitor how various treatments — such as surgery and vocal therapy — impact vocal behavior. They also hope to use a similar technique on electrocardiogram data, which is used to track muscular functions of the heart.