Hi there! The human body is one of the marvellous creations of nature, right? Have you ever come across the word, "EEG"? Of course, if you are somewhat close to medical jargon, you must. Others, let's say it's some kind of an electrical signal that represents the brain activity. Sounds good?
So today, I'm going to talk about something called "Auditory Brainstem Response" (ABR).
ABR falls under "Evoked Potentials" (Eps). Wait a second. I know, I'm throwing out a lot of complex words towards you. Let me explain briefly.
When there's an external stimulus to our brain, either auditory, visual, somatosensory, etc, there arises a potential in our scalp as the brain's sensory processing to that stimulus. Those are what we call "Evoked Potentials". So when the stimulus is auditory, neural activity elicited by the cochlea of the ear in response to that specific auditory stimulus, is called "Auditory Brainstem Response". Hope it's clear. 🤗
Usually this stimulus can be a 'click' or a 'CE-chrip'. This image shows, time domain representations of those two stimlus.
The next image you see is a time-domain representation of such ABRs at different stimulus intensity levels. As you can see, there are five main peaks for an ABR but it's not that clearly noticeable in every case. Because these ABRs lie within the (µV) microvolt range and are comparably very small concerning the noise we encounter while recording these potentials. Also note that the frequency of ABRs lies within 100Hz to 3000Hz and ABR appears within 10 ms usually.
Now, let's see the anatomical generators of each wave of this ABRs. If you are want to go straight to the Signal processing part, the engineering stuff, you may skip this to here.😉
Wave I : Distal portion of cranial nerve VIII
Wave II : Cochlear nucleus and proximal VIII nerve
Wave III : Superior olivary complex in the caudal portion of the auditory pons
Wave IV : Pontine third-order neurons mostly located in the superior oliv
Wave V : From the vicinity of the inferior colliculus
Cool. Now it's time for the biomedical engineering stuff. 🤩
We did an experiment to record ABR from a healthy person inside our biomedical engineering lab at the University of Moratuwa. For the rest of this article, I am referring to the procedure we followed. We used ADInstruments Octal Bio-Amp™ and ADInstruments PowerLab 16/35™ as our amplifier setup. Also, we used Etymotic Research ER-3C™ insert earphones to provide the auditory stimulus to the subject. The stimulus was provided to the right ear while keeping the left ear sealed by a foam ear tip. The reason for selecting these insert earphones is that it has a high noise exclusion (30+ dB external noise exclusion) while 70+db isolation between two ears. (Interaural attenuation)
The data acquisition was performed with two electrode montages.
We used disposable Ag/AgCl electrodes and skin cleansing alcohol swabs, and skin preparation abrasive gel was used to prepare the skin. I don't pay that much attention to explaining that because from this article what I expect to talk about is signal processing and results. 🙂
Oh! If you are lost middle of nowhere trying to understand the above pic of electrode montages, just think of A1 as the left ear and A2 as the right ear. So, Nasion is a bone near the nose. Inion is from the back of the scalp.
We provided the stimulus, each click and chirp, to the subject via the insert earphones at three different intensity levels 40dB, 60dB and 80 dB. Every stimulus was provided repeatedly thoughout a window of one minute and both the audio stimulus and the bio signal were recorded with a sampling rate of 10K Hz. Since ABR could have a frequency up to 3000Hz, the sampling rate must be more than twice that of it from the Nyquist theorem.
The full code, I made to analyse these results, can be found in my git repo. 👨🏻💻👈🏻
Shall we take a look at our recordings? 😉
Okay, that seems like a lot of data, huh!
Let's zoom in. Remember I told before, that this ABR appears within 10 ms of the stimuli.
"Oh c'mon man, what is this? 🥴", is that what you thinking at the moment? I zoomed the above recording to the time between 12 to 12.09. So we should have multiple ABRs present here because we were providing the stimulus repeatedly over this time frame. But is it visible here? Hell, no. It is because the ABRs are in the range of µVs and the noises play a huge role here.
Let's apply our engineering knowledge here.
Since we know that the ABRs, the signals we are interested about, are lying in the frequency range of 100Hz to 3000Hz, we can apply a bandpass filter to remove the frequency components that are out of this range.
Now I have filtered the signal using a bandpass filter. But still, the ABRs are not visible. 😢
What is the reason? Any guesses? Okay, the problem is, that the noise and the information we are looking for, have frequency overlappings. Hence, a frequency-based filtering, like a bandpass filter alone, will not do the job. We need some kind of a trick to remove the noise, that appears in the frequency range of the useful information. Tedious, right!😨
This is where, the signal processing technique, "Ensemble Averaging", comes into action.
We have several assumptions here. We assume that the signal and noises are uncorrelated. Also, we assume that the noise is a broadband stationary process with a population mean of zero and non-zero variance, say σ2. (sigma squared).
It can be shown that, if we average the signal over N number of repetitions or realisations, we can reduce the noise variance σ2 to σ2/N. Showing the proof in this article will make this article a long one. So if needed, I can cover it with a new article. ✍️
So when the number of repetitions/realisations increases, we can reduce the noise variance to a greater extent. Since the noise mean is zero, it will do the job perfectly.
To find the times to separate each stimulation window, we can use a threshold for the stimulus voltage, and a few lines of code will help us to separate each time window. That means each repetition window. It is like counting the number of containers in a train. The code can be found in my GitHub repo.
Are you ready to see the results?
For the above electrode montage 1, when using CE-chirp as the stimulation and 80 dB as the intensity, I ended up achieving this result. 👏
This is a very good result under our practical conditions. Wave V, the most often analyzed component in clinical applications of the ABR, is clearly visible here. As you can see in the title, 1281 repetitions are averaged to obtain this result. So the noise variance is 1281 times smaller than its normal case.
Why stopping here? Shall we do some analysis too.
Below is the result comparison when we provide click and chirp stimulus, keeping other conditions (electrode montage and intensity level) constant.
As we can see from the results, when using chirp as our stimulus, we have a higher amplitude and a lower latency in our ABR. It's because, when compared with the click stimulus, chirp signal causes synchronous displacement along a larger portion of the basilar membrane. [1]
What if we change the intensity level, keeping other factors constant.
As the intensity increases, the amplitude rises and the latency decreases. The reason behind this behaviour is, that when we increase the intensity of the stimuli, the number of neuron firings increases causing the net amplitude to rise up. [2]
Let's keep others constant while changing the electrode montage
Here I have provided the chirp 40dB signal as the stimulus. Since the intensity is low, ABR has a lower amplitude in both. But configuration 2 has a higher amplitude compared to the configuration 1. The reason is that the auditory cortex of the brain is more close to Cz than the mastoid. [3]
So that's it. Hope this finds you interesting. Have a great day! Tada 👋
PS:
👈🏻 Behine the scenes
References
[1] M. Chertoff, J. Lichtenhan, and M. Willis, “Click- and chirp-evoked human compound action potentials,” The Journal of the Acoustical Society of America, vol. 127, no. 5, pp. 2992–2996, May 2010, doi: https://doi.org/10.1121/1.3372756
[2] R. M. Stelmack and K. G. Wilson, “Extraversion and the effects of frequency and intensity on the auditory brainstem evoked response,” Personality and Individual Differences, vol. 3, no. 4, pp. 373–380, Jan. 1982, doi: https://doi.org/10.1016/0191-8869(82)90003-4
[3] A. J. King and Y. S. Sininger, “Electrode Configuration for Auditory Brainstem Response Audiometry,” American Journal of Audiology, vol. 1, no. 2, pp. 63–67, Mar. 1992, doi: https://doi.org/10.1044/1059-0889.0102.63
Comments