Some Open Problems in Large Volume Data Mining in Biomedical Applications

Опубликовано 6 сентября 2016, 5:37

Recent advances in sensor technologies have enabled long term recordings of numerous physiologic parameters in patients, generating very large data sets.  This phenomena extends to implantable sensors as well as non-invasively applied external sensors.  A cursory review of applications, issues, and opportunities will be presented, with examples drawn from electroencephalographic (EEG) recordings, hyperspectral imaging of skin and cardiovascular systems, mass spectrometry, and similar uses.  Key properties of biosignals (such as smoothness, low rank data, etc) allow novel methods to be applied to these very large data volumes for extraction of actionable clinical information and practical application in intelligent bedside diagnostic instruments.  These advances can propagate to consumer electronics health-related applications.