Lecture 5 Flashcards
1
Q
Why do we have to use ABR in babies?
A
- Babies can’t tell you when they can and can’t hear
- Can’t use behavioural audiometry
- Need to use objective measures
2
Q
What is the prevalence of infant hearing loss?
A
1/1000
3
Q
What is considered a significant loss in a baby?
A
40 dB or greater is what we call a significant loss in a baby (prevalence is about 1/1000)
4
Q
Why is 40dB considered a significant loss in a baby?
A
It is very hard to pick up a mild loss in a baby from these measures
5
Q
What is the JCIH time frame?
A
- Screening by 1 month
- Full audiologic/medical evaluation by 3 months
- Intervention by 6 months
6
Q
For babies that require a cochlear implant, when do they get this?
A
- Cochlear implants are not implanted until a year
- Usually a hearing aid is fit prior to receiving cochlear implant surgery
7
Q
What is UNHS?
A
- Universal newborn hearing screening
- This used to be the push
- BUT, having screening, but not having anything to follow up isn’t good (that’s why EHDI is better)
8
Q
What is EHDI?
A
- Early hearing detection intervention
- We want this (it catch’s hearing loss and has the right support services)
9
Q
What are the two components of UNHS?
A
- Otoacoustic emissions (OAEs)
- Automated auditory brainstem response (AABR)
10
Q
Explain OAEs
A
- TEOAE or DPOAE
- Very efficient
- Used for screening well-babies
11
Q
Explain AABR
A
- Slightly more time to conduct
- Requires more expertise
- Assesses more of the auditory system
- Used for screening in NICU (risk factors for AN/AD)
12
Q
What is the difference between AABR and ABR?
A
AABR is doing an ABR, but it is being scored for you
13
Q
Why are AABRs better than OAEs, but aren’t used as often?
A
- AABR is less common than OAE
- AABR picks up neural problems (OAE doesn’t)
- AABR is better, but OAE is more accessible
14
Q
AABRs are the best way to screen, but there are ____
A
Limitations
15
Q
We need to be able to categorize results into three possible categories (as a function of ____ and ____)
A
Frequency, level
16
Q
What are the 3 possible categories to categorize results?
A
- Response (hearing)
- No refer
- No response (not hearing)
- Refer
- Noisy result (inconclusive)
- Baby isn’t very relaxed
17
Q
What wave matters most with screening?
A
Wave V
18
Q
We want an electrode on either side of the ____
A
Dipole
19
Q
Where do we want to put the active electrode?
A
Active (non-inverting) at vertex (Cz) or forehead (FPz)
20
Q
Where do we want to put the reference electrode?
A
Reference (inverting) at mastoid (TP7 or TP8)
21
Q
Where do we want to put the ground electrode?
A
Ground on lower forehead / contralateral mastoid / clavicle
22
Q
What is the best impedance for babies?
A
- Target is often < 10 kOhm and within 1 kOhm
- Adults is less than 5 kOhm
23
Q
Low-pass filter generally ____Hz for threshold ABR
A
1500
24
Q
High-pass generally ____Hz for threshold ABR
A
20-30
25
Filter settings are all about one thing, which is...
Seeing wave V
26
Explain noise when recording ABR?
- Noise is not stable over a run (most of the EEG size is noise)
- On each sweep, most of what you are recording is noise
27
Variation in sweep size because of ____
Noise
28
A big sweep is ____
Noisy
29
A small sweep is ____
Quiet
30
Explain weighted vs non-weighted averaging
Non-weighted averaging:
- Every sweep (whether its noisy or not), gets included
Weighted averaging:
- A rejection level is added to get rid of very noisy sweeps
- Accepting less of what happens on the noisy trials gives a cleaner result
31
Use ____ in assessment, especially thresholds.
Weighted averaging
32
What are the two objective assessment and detection methods?
1. Response correlation
2. SNR-based methods
33
When doing a screening ABR, the system ____ decides if an ABR is there
Objectively
34
When doing a diagnostic ABR, you need ____ criteria. Why?
- Objective
- Statistics help you know if ABR is there or not because sometimes you are unsure
35
What is correlation?
The correlation between two waves (because you need to do two sweeps)
36
What is cross correlation?
- Same thing as correlation, but you are moving one of the things
- Calculate correlation, then move one wave forward in time a bit, recalculate, move again, recalculate, etc.
- Gives you a bunch of correlation values at different lags
37
When does cross correlation matter most?
- This matters the most when trying to correlate things at 2 different levels (ABR gets later and lower levels)
- This is a good way to determine if a wave is delayed
38
What are two ways to look at correlation or cross-correlation?
- Replications
- Split-halves of a single run
39
A correlation of ____ or greater is strongly suggestive of a response
0.46 (if there is a high enough correlation, the baby is not referred)
40
What is template matching?
- Cross-correlation with a template
- They have a template of what wave V should look like (bump and negativity)
- Slide the template around at different times to see if wave V is present
41
What is the SNR-based method looking at?
- How big is the ABR compared to the noise (SNR)
- You can't just measure the ABR without noise
- We are always measuring EP + BN
- As we average, noise goes down (but its always present)
42
What is the formula for SNR-based methods?
- We can’t get ABR without noise, but we can get noise without ABR (SNR+1)
- SNR + 1 is found from ABR+BN / noise estimate (so to figure out SNR – 1)
43
if we divide the amplitude of the ____ (which includes noise) by the amplitude of the ____, we have SNR + 1.
Response, noise
44
What are the 3 steps to estimate SNR?
1. Estimate amplitude of BN
2. Divide response amplitude (which includes noise) by this
3. Subtract 1
45
What are the 3 basic methods of estimating background noise?
1. Signal absent recording (e.g., a pre-stimulus basseline)
2. The plus/minus average
3. Trial-to-trial variance
46
How does the signal absent method work?
- Compare amplitude of trial with stimulus present (EP+BN) with amplitude of trial with no stimulus present (BN)?
- This requires two runs, one with stimulus and one without (or a prestimulus baseline)
47
What is the problem with the signal absent method?
- The brain changes after stimulation (i.e., even the noise is not the same)! --> Induced Activity!
- Not all is time locked (some is induced)
- The BN on either side of the stimulus do not equal each other
48
Explain with the equation why the signal absent method doesn't work
49
Explain the +/- average method
- Create an estimate of the noise, by inverting every other response before averaging.
- Anything consistent from trial to trial should get subtracted away
- If you have a wave V on sweep 1 and 2 (but 2 is flipped), if they are the same, they will cancel out and only the noise will be left (left with everything that is different between sweep 1 and 2)
50
What is the advantage of the +/- averaging method?
The background noise is obtained from the same time period as the response, so the following equation holds
51
What is the problem with the +/- method?
Much better than the signal-absent method, but its contaminated by slight differences (variability) in the actual EP that aren't subtracted away
52
Explain the trial-trial variance method
1. Pick a single time point (wave V peak)
2. Record response amplitude at this time point over individual trials
3. Find the difference between each trial amplitude and the across-trial amplitude (average)
4. Since these differences will be positive and negative, the average differences will be zero, so we square them to make them positive
- How to calculate variance
5. The sum of the squared values (SS) divided by the number of trials is the mean squared value (MS)
- This is variance (or power) of the noise
6. The square root of the mean squared value is the standard deviation (amplitude of the noise)
53
What is residual noise?
Residual noise is simply the background noise estimate (i.e., the square root of the trial to trial variance)
54
What is the best method for BN?
Trial-to-trial variance is similar to the plus/minus average, but it uses actual trial-trial differences instead of just average differences (some of the noise gets averaged away)
- This is the best approach
55
1. Signal-to-noise ratio is a ratio of ____
2. If we square this value, we have a ratio of ____
3. A ratio of variances can be tested for significance using an ____
1. Amplitudes
2. Powers
3. F distribution
56
SNR^2 is a ratio of ____, and therefore is an F statistic
Powers (or variances)
57
What is the Fsp?
- When the noise is estimated using trial-to-trial variance, measured at a single time point this is called the Fsp (usually where wave V happens)
- Fsp = a test to see if an ABR is significant at a single point in time
58
When you do the Fsp, you are getting the ____, but as an F statistic
SNR
59
What gives you the Fsp?
1. Square ABR amp to get the power
2. Then divide that by the trial-to-trial variance at that point
3. This gives you the Fsp
60
An Fsp of ____ corresponds to a confidence interval of 99%
- 3.1
- Target Fsp is 3.1 because it gives you a 99% that ABR is present
61
Most clinical systems now include an ____ measurement
- Fmp (multiple points)
62
The RN should be ____nV before deciding that no response is present
~ 20 (Ontario)
63
When do you have enough averages - subjective?
- When response can be seen
- Replicate to visualize variability (better: calculate cross- correlation)
- Run at multiple levels to visualize latency-intensity function
64
When do you have enough averages - objective?
a) When response is sufficiently large relative to background noise (e.g., SNR > 1.2, Fsp > 3.1)
- BN can be estimated by baseline, +/- averaging or trial-trial variance or amplitude (but baseline is a poor choice)
b) When background (i.e. residual) noise is sufficiently low (20 nV), even if no response
65
You use ____ when an ABR response isn’t clear
Stats
66
Every response needs to be categorized as:
1. Present
2. Absent (absent with low noise)
3. Undetermined (absent with high noise)
67
Need ____ to determine between present or absent (this is where stats comes in)
Low noise
68
When screening with the ABR, how many levels is the test conducted at?
- Test is conducted at one level
- 35 dB nHL is most common (not getting a wave V at this level indicates hearing loss)
69
Screening with ABR - automated
- Must use an objective criterion (F test or cross-correlation)
70
Screening with ABR - non-automated
- Usually with an objective criterion
- SNR
- Fsp
- Residual noise
71
ABR considered most accurate test of hearing for infants ____
< 6 months
72
ABR stopping criteria (for sininger et al research)
1. 70 dB nHL stimulus stop when response observed
2. 30 & 50 dB nHL stimuli; stop if one of the following:
a) Fsp criterion reached AND at least 768 sweeps AND observed (pass)
b) Low residual noise (<15 nV) was reached without response (refer)
c) Maximum number of sweeps (6144) without response (refer)
73
Explain the difference of ABR and OAE at low and high frequencies
- ABR is the best at detecting HL at lower frequencies (better than OAEs)
- At high frequencies, OAEs performed similarly to ABR
74
If the prevalence is 0.2%...
- 2 children out of 1000 have hearing loss
- 998 children do not have hearing loss
75
If you test 1000 babies, how many refer results should you get?
- Hit rate = 80% --> 80% of 2 = 1.6
- FA rate = 10% --> 10% of 998 = 99.8
- That means you will have 1.6 + 99.8 = 101.4 positive results
76
Given a positive result, what is the chance that the child has hearing loss?
1.6/101.4 = 1.6%
- Most refers with an ABR screening are going to be wrong
77
If you test 1000 babies, how many pass results should you get?
- Miss rate = 20% 20% of 2 = 0.4
- Correct-rejection rate = 90% --> 90% of 998 = 898.2
- That means you will have 0.4 + 898.2 = 898.6 negative results
78
Given a negative result, what is the chance that the child can hear normally?
898.2/898.6 = 99.96%
79
How many peaks are visible in infant ABR?
- Wave I is adult-like by 3 mos
- At birth: 2 ms, about .35 uV
- Wave III is adult-like by 8-16 mos
- Wave V is adult-like by 18-36 months
- at birth: 7 ms, about .4 uV
80
What is the latency-intensity function for infants?
Latency-intensity function is –35 µS/dB (compared to –44 µS/dB in adults)
81
In infants, there tends to be a steeper slope for ____
SNHL (especially sloping)
82
Latency changes with ____
Age
83
What wave settles first and last?
- Wave I settles first, wave V settles later (mature from the outside in)
- By 3 months, you have already reached mature levels of wave I
84
Wave V is late at ____ levels
Soft (when doing actual testing, you are looking at 15ms for wave V)
85
Threshold rate
- Threshold is not affected by rates up to 50 Hz
- Usually we use a rate ca. 40/sec
86
Threshold phase
- Stimulus phase does not affect thresholds
- Use alternating to eliminate cochlear microphonic
- Or combine rarefaction and condensation after recording (i.e., to allow for visualization of CM)
87
What stimuli do we use for threshold tests?
We use tone bursts because they are more frequency specific
88
Explain what is important to do with a baby before testing
- Newborn babies naps are short (roughly an hour at a time)
- Sedation with chloral hydrate when necessary, usually not required with newborns
- Prepare baby before feeding
- Make sure baby is not fed the hour before the appointment
- Ideally you want a hungry and tired baby!
89
What level do you begin testing at with a baby?
- Begin testing at lower level (30 dB)
- Less chance of waking baby
- Less wasted time
90
General approach is all about ____
Efficiency
91
What are the 4 testing steps of the ABR procedure?
1. Start with 2000 Hz at 30 dB nHL, than 60 dB nHL
2. If no response, try bone conduction at 30 dB nHL etc. More important to know type of loss than threshold at first.
3. Then find threshold in 10 dB steps
4. Get 2 kHz in both ears first (and maybe 4 kHz) before going on to 500 Hz, which is more difficult
92
Switch to ____ very quickly is not seeing ____ response
BC, AC
93
Always use ____ channels
Two
94
Is masking required?
- Masking is generally not required; if the other ear is responding, the response will be larger contralaterally.
- This is even true for bone conduction in infants. They generally have 10–15 dB IA for bone
95
Goal is to Estimate ____ Threshold
Behavioural
96
Estimated behavioural threshold is called ____
eHL
97
ABR thresholds are typically expressed in ____
dB nHL
98
ABR measure of ____ needs to be correct to ____ to fit a HA
nHL, eHL
99
ABR threshold estimates have standard deviations that range from about ____dB across various studies
5 to 15
100
Recording parameters - electrodes
FPz to TP7 / TP9 (mastoids)
101
Recording parameters - impedance
< 5 kΩ adults, < 10 kΩ children
102
Recording parameters - filter
30 --> 1500 Hz
103
Recording parameters - average
- As many as necessary (e.g. 1000-6000)
- Waves do not need to be pretty!
104
Recording parameters - rate
~40/s
105
Recording parameters - window
20-30 ms (!!!!) - need a big window
106
Recording parameters - stimulus
- 100 µS click for neurodiagnostics
- 2-1-2 tone-bursts for thresholds
107
Recording parameters - rejection
+/- 25 uV
108
Recording parameters - criteria
- Objective criteria if possible
- Fmp, residual noise,
- SNR, correlation
Generally targeting residual noise ~ 20 nV
109
Recording parameters - sedation
- < 6 mos (no sedation)
- > 6 mos (sedation)
110
Recording parameters - weighted vs. non-weighted
Weighted averaging if possible
