Nighttime features derived from topic models for classification of patients with COPD

Nighttime symptoms are important indicators of impairment for many diseases and particularly respiratory such as chronic obstructive pulmonary disease (COPD). The use wearable sensors to assess sleep in COPD has mainly been limited the monitoring limb motions or duration continuity sleep. In this paper we present an approach concisely describe patterns subjects with without COPD. methodology converts multimodal data into a text representation uses topic modeling identify across dataset composed more than 6000 assessed nights. This enables discovery higher level features resembling unique characteristics that then used discriminate between healthy those evaluate patients’ severity dyspnea level. Compared standard features, discovered latent structures nighttime seem capture aspects sleeping behavior related effects dyspnea.