Sleep Fragments

HomeProfessionalsClinical ResourcesSleepSleep Fragments ▶ Fun With Filters
Fun With Filters

Contributed by Susheel Patil, MD, PhD and Hartmut Schneider, MD, PhD, Johns Hopkins University, Baltimore, MD

The fragment below is a 2 minute recording from a CPAP titration study. What respiratory events are shown on the recording?

Click here for a larger view.


The respiratory events observed in this fragment are apneas. There are features that are suggestive of a central apnea (i.e. cardiac oscillations in the airflow channel). There are also features suggestive of an obstructive apnea (i.e. persistent respiratory effort, particularly in the abdominal effort channel). Review of the study by the sleep physician revealed that the effort channels were filtered inappropriately. The low frequency filter was set at 0.1 Hz and the high frequency filter was set at 0.5 Hz. When the high frequency filters was increased to 15Hz, the fragment appears as in the Figure.

In this appropriately filtered fragment, cardiac oscillations are now apparent in both the thoracic and abdominal effort signals. The lack of increasing thoraco-abdominal movements or paradox effort during the apnea confirms the presence of a central apnea.

The artifact created by the initial filter settings is due to an inappropriately low, high frequency filter setting. A high frequency filter attenuates fast frequency waveforms. The setting of a high frequency filter sets a cutoff frequency above which the amplitude of that frequency is reduced and higher frequencies become increasingly attenuated. The frequency of the cardiac oscillations in the abdominal effort channel is at 1.2 Hz. The high frequency filter of 0.5 Hz eliminates the cardiac oscillation and results in aliasing by misrepresenting the true frequency of the signal observed in this channel.

The effect of increasing the filter setting is seen in the following figure. At 5 Hz, the frequency of the underlying signal is preserved, however at lower filter settings, the cardiac oscillation artifact is lost. At a high frequency filter setting of 0.5 or 1.0 Hz, the waveform takes on a frequency compatible with respiration that can lead to the false conclusion of the presence of abdominal effort. Current guidelines from the American Academy of Sleep Medicine’s Manual for the Scoring of Sleep and Associated Events suggest that the appropriate filter settings for respiratory signals should be set at 0.1 Hz for the low frequency filter and 15 Hz for the high frequency filter. The minimal sampling frequency recommended for respiratory signals is 25 Hz, with 100 Hz recommended as most desirable for the visualization of cardiac oscillations.

The AASM Manual for the Scoring of Sleep and Associated Events. Westchester, IL: American Academy of Sleep Medicine, 2007.

Fisch BJ. Digital and analog EEG instruments: parts and functions. In: Fisch BJ, editor. Fisch and Spehlmann's EEG primer: Basic Principles of Digital and Analog EEG. Amsterdam: Elsevier, 1999: 35-72.


Click here for a larger view.