In this episode Jen and Meshia welcome back Petra Davidson to discuss questions we’ve received on filters.

Petra gives a very broad definition of filters as electronic curtains that we employ to cut out unwanted electrical signals. An EEG is a drawing, the graph part, of the electrical signals originating from the cortex of the brain.  Our equipment, no matter how well made, cannot record only brain electrical signals. It picks up all electrical signals in the preset frequency ranges. This includes other electronic devices, involuntary and voluntary muscle activity, electrical signals from eye movement and tongue movement and any other electrical discharges.  Anything from 0 cycles per second or Hz to 120 Hz.  We then use filters or curtains to block unwanted signals, similar to the way we use curtains to cut out unwanted light sources.  There are three basic types of filters, low frequency, high frequency and notch filters. 

Aliasing: changing of the signal because of undersampling. When the analog to digital converter brings in the signal, it has to choose discreet data points, consequently it is limited to only a finite number of data it can properly assess. It then samples and retransmits the data. It is similar to playing the telephone game and only relaying every other word. 

Nyquist theory: In order to obtain a proper output signal the incoming sample must be at a rate twice the highest frequency you wish to record.

Bandwidth:  refers to the range of frequencies between the established filters.

Electrode impedance: the total effective resistance to alternating current.  We need some resistance to AC current which is one reason we do NOT want the impedance to be less than 1.

Paul asked – Do the filters cut out all unwanted signals or only at those specific frequencies? 

The filters attenuate or flatten the specified signal’s amplitude by a maximum of 20-30% of its original amplitude. For low pass, also known as, high frequency filters, the signals above the set limit are attenuated or reduced in amplitude progressively higher than 20-30% up to eliminated. For high pass or low frequency filters, the signals below the set limit are attenuated or reduced in amplitude progressively higher than 20-30% up to eliminated. Filters are not perfect, per industry standards, filters can allow up to a maximum of 20-30% of the original signal for the specific named filter to be displayed. 

Ron asked – Would it be best then to set your low filters as high as you can and the high filters as low as you can to get a narrow band of activity? Wouldn’t that allow the cleanest signal? 

Absolutely not. As EEG machines have improved over the last several decades, research has shown that we have only begun to tap into the full frequencies the brain can reach. It is best practices to keep the filters as open as possible to allow for maximum visualization of the frequencies produced by the brain. 

Cyndi asked – What can happen if the filters are too narrow? Does that affect the recording? 

If the filters are too narrow, the doctor’s ability to properly diagnose the patient is at risk. If the low frequency filters are set too high, slower frequencies are blocked.  It is important in cases of encephalopathies, comas and altered mental status to record the slow activity in order to properly assess the patient’s outcome.  If the high frequency filters are set too low, the high frequencies such as spikes and sharps and even the overall background rhythm.  

Connor asked – When performing an EEG study, should we feel free to use the notch filter any time we are in the ICU? 

In EEG and other Neurodiagnostic modalities, the initial instinct to clearing up results should always be troubleshooting as you mentioned. Always apply troubleshooting first. Then when you have eliminated or labeled and monitored sources of noise, appropriately applying filters is acceptable. Make certain the impedances are balanced prior to employing the notch filter. If impedances are not balanced, the notch filter can block the appearance of a bad electrode to surface interface.  This means that the notch filter has the ability to disguise a bad electrode.  This disguise appears on an EEG as a normal waveform, an alpha coma, a sinusoidal theta discharge, or even intermittent bursts of spindles. The artifact from a bad electrode does not always come in the form of high amplitude, thick, black fuzzy lines. Particularly in those patients who are most critical, the bad electrode may appear as either focal or generalized alpha.  

Does that mean that if the impedances are high, but balanced, the test quality will be adequate? 

Most likely, with modern amplifiers, high but balanced impedances will result in an adequate test, although it would be better to have low and balanced impedances. With that being said, impedances should never be below 1 kOhms. 

In EEG, what are the typical sources of noise and signal you seek? 

Common noise in EEG or the unwanted signal typically arises from the following sources: EOG or eye movement, EMG or muscle artifact and power line artifact. Common cortical signals in EEG  or our desired signals are tiny signals that only originate from the cortex, or brain itself. The noise signals and brain signals share common frequencies which is why filters are necessary but must be used cautiously.  Noise signals from eye movement are typically 1-3 Hz, the same as delta cortical activity.  Noise signals from EMG or muscle and from power lines are 50-120 Hz which are in the same range as seizure like sharps and spike wave activity.  

Amanda asked – Are there any other Disadvantages and advantages to each type of filter?

According to UC Davis, Physics Department, there are many disadvantages and advantages to each type of filter. I feel confident in saying that by and large electronics specialists prefer digital filters for the following reasons: 

  1. Digital filters are programmable. Analog filters require physical restructuring to handle different settings.
  2. Digital filters are more stable with respect to time and temperature.
  3. Digital filters are more easily tested, designed and utilized.
  4. Digital filters are more versatile.
  5. Digital filters can handle a wider range of frequencies more easily.

With all of those advantages, it appears that digital filters are the only way to go. Now, it is important to understand that analog filters have the advantage that they can sample the signal in its entirety.  Digital filters must select discreet samples. This is where aliasing is important. It can make or break all of the advantages we just listed. Under-sampled signals, no matter how well filtered are not an accurate representation of the cortical activity.

As you study for your boards it is important to study many resources as you might find one that really speaks to you more than another.  For example, there is a fantastic YouTube video by UCL Medical Physics and Biomedical Engineering called, “Understanding Frequency Filters in 10 minutes”.  If you have spent any time on the Trusted Academy’s website, you know that we offer  CEU classes, foundations, board prep classes for a number of different neurodiganostic fields and tons of resources to further understand filters.