Homepage | Links | Contacts
Projects

Project Contact:
José Manuel Fonseca
(jmf@uninova.pt)

Ali Abdollahi Gharbali
(a.gharbali@campus.fct.unl.pt)

Shirin Najdi
(s.najdi@campus.fct.unl.pt)

Financing:
Internal Project

Deep Sleep
Project Summary
Automatic Sleep Stage Classification using Advanced Artificial Intelligence Techniques
Sleep occupies significant part of human life. It is fundamental for physical and mental health. Although most people think that sleep is a passive and constant process, as a matter of fact, it is an active state. Human body moves frequently during the night and the brain is sometimes more active during sleep than in the period of the normal wake state.

Healthy human sleep generally consists of two distinct stages with independent functions known as Non-Rapid Eye Movement (NREM) and Rapid Eye Movement (REM) sleep. In the ideal situation, NREM and REM states alternate regularly, each cycle lasting 90 minutes on average. NREM sleep accounts for 75-80% of sleep duration. According to American Academy of Sleep Medicine (AASM), NREM is subdivided into three stages: stage 1 or light sleep, stage 2 and stage 3 or Slow Wave Sleep (SWS). Classification of sleep stages is vital for diagnosing many sleep related disorders and Polysomnography (PSG) is an important tool in this regards.

The automatic sleep stage classification mainly consists of three steps: pre-processing, feature extraction and classification. The pre-processing step includes artefact rejection and artefact correction. In the feature extraction step, researchers try to compactly represent PSG recordings by means of a feature vector. This feature vector should be informative and non-redundant enough to ease the subsequent classification step. Finally, in the classification step, the extracted feature vectors are assigned to one of the five categories using a proper classifier.
Although much has been done for automatic sleep stage classification, still there is lack of a method for efficient sleep monitoring which can be considered reliable from medical expert’s point of view. The aim of this project is to fill this gap by using advanced data mining and artificial intelligence techniques.



Overall Objectives

The project will contribute to the investigating the potential of advanced artificial intelligence methods in reliable automatic sleep stage classification.

The specific objectives of the project include:

  • De-noising the PSG recordings using adaptive methods without distorting the original signal.
  • Analysis the capability of traditional features from the consistency and effectiveness point of view.
  • Propose new feature extraction and selection scheme based on advanced data mining techniques.
  • Applying recent and advanced classification methods for achieving more robust and accurate classification results compared to traditional methods.

Research Topics
  • Artificial Intelligence
  • Deep Learning
  • Adaptive De-noising
Partners