“Stacked ABR: Fundamentals and Use
in Small Tumor Screening”
Presentation prepared by Manuel Don, Ph.D.
Electrophysiology Department
House Ear Institute, Los Angeles, CA.
Presented by Kathy Murphy, M.A. CCC-A
Bio-logic Systems Corp.
HOUSE EAR INSTITUTE
Manny Don:
Tôi sẽ
Talk Outline
1) Giới thiệu tổng quan về điện thính giác thân não mở rộng( Stacked
ABR)
I.
Background: Evolution of the Stacked ABR
2) Bàn luận về phương pháp đo và phân tích ABR
3)Điểm qua những nghiên cứu đã công bố và giới thiệu những số liệu
của những đề tài nghiên cứu gần đây chưa được công bố với những khối
u nhỏ
II.
Stacked ABR: Rationale and Method
III.
Stacked ABR: Studies
4) Đưa ra những biên bản ngắn gọn sàng lọc những khối u nhỏ
5) Những điểm nổi bật nhất của những nghiên cứu mở rộng của
Stacked ABR trong tương lai.
IV. Tumor Screening Protocol
V.
Future Stacked ABR Work
Talk Outline
HOUSE EAR INSTITUTE
Manny Don:
Tôi sẽ điểm qua:
Background: Evolution of the Stacked ABR
1) Khối u dây VIII tự nhiên 2) ABR tiêu
chuẩn đánh giá khối u.
3) Giải thích những gì biểu hiện trên ABR
tiêu chuẩn.
A. Eighth Nerve Tumors
4) Và thảo luận về các giới hạn của ABR
tiêu chuẩn.
B. Standard ABR Tumor
Detection
C. What Do Standard ABR Measures Represent?
D. Limitations of Standard ABRs
I. Background
HOUSE EAR INSTITUTE
Manny Don:
First, a brief
discussion of
eighth nerve
tumors
Background: Evolution of the Stacked ABR
A. Eighth Nerve Tumors
B. Standard ABR Tumor Detection
C. What Do Standard ABR Measures Represent?
D. Limitations of Standard ABRs
I. Background
HOUSE EAR INSTITUTE
Manny Don:
•This is a cross section through a human auditory meatus taken from
Cross Section: Human Auditory Meatus
some work published by Spoendlin and Schrott in 1989.
It shows the spatial relationship of the auditory and vestibular nerve
bundles. This relationship is important because acoustic tumors are,
in reality, vestibular Schwannomas, i.e. tumors that arise from the
Schwann cells of the vestibular nerves.
I want to emphasize that cochleotopic representation in the auditory
nerve as deduced from the representation of the turns shown in roman
numerals, demonstrates that tumors arising from the vestibular nerves
(superior
and inferior) can affect auditory nerve fibers from both low
(2 kHz)
and high frequencies. Note, for example that the fibers from turn II
representing 1 kHz are, at this level, very close to the inferior
vestibular nerve.
(1 kHz)
(6 kHz)
Spoendlin and Schrott (1989)
HOUSE EAR INSTITUTE
I. Background: Eight Nerve Tumors
Manny Don:
Cross-section of Internal Auditory Canal (IAC)
For the remainder of the talk, I will be
using this schematic drawing of the
cross-section through the internal
auditory canal (IAC) which shows the
four
major nerve
bundles: the acoustic
Facial
Nerve
nerve, the superior and inferior
vestibular nerves, and the facial nerve.
Acoustic Nerve
Sup.
Vest. Nerve
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE
I. Background: Eight Nerve Tumors
Manny Don:
Medium or Large Tumor in IAC
This is a
schematic
illustration of
how a medium
orFacial
relativelyNerve
large tumor
arising from the
superior
vestibular nerve
might encroach
on the acoustic
nerve fibers in
the internal
auditory canal.
Acoustic Nerve
Sup.
Vest. Nerve
Tumor
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE
I. Background: Eight Nerve Tumors
Manny Don:
Next, I would like to review very briefly the standard ABR measures for
tumor detection. In a publication a few years ago, Dr. Bauch and his
colleagues showed that the best two standard ABR measures for tumor
detection was the IT5 and the I-V delay.
Background: Evolution of the Stacked ABR
A. Eighth Nerve Tumors
B. Standard ABR Tumor Detection
C. What Do Standard ABR Measures Represent?
D. Limitations of Standard ABRs
I. Background: Standard ABR
HOUSE EAR INSTITUTE
Manny Don:
Standard ABR Measures for Acoustic Tumor Detection
IT5 is a measure of the interaural time delay for wave V and
IT5 = Interaural time delay for wave V
was developed many years ago by Selters and Brackmann.
One simply compares the latency of wave V between
the
6.4
tumor suspected ear and the non-involved ear. If theNon-Tumor
latency Side
in the tumor suspected ear exceeds that of the non-involved
ear by a certain criterion, the test is positive for a tumor.
There is some correction factor for hearing loss.
L1
IT5 = L2 - L1 = 0.9 ms
L2
7.3
Tumor Side
0
2
4
6
8
ms
10
12
14
HOUSE EAR INSTITUTE
I. Background: Standard ABR
Manny Don:
Standard ABR Measures for Acoustic Tumor Detection:
I and V
The I-V delay is simply the latency difference between wave I and
Delay
= Latency
Difference
Between
Wave
wave VI-V
of the
ABR response
in the
suspected ear.
If this delay
exceeds a certain criterion value, this measure is positive for a
tumor.
I - V = 4.85 ms
I-V Delay
I-III Delay
6.55
4.90
Acoustic Tumor
1.70
V
I
0
2
III
4
6
8
10
12
14
ms
HOUSE EAR INSTITUTE
I. Background: Standard ABR
Manny Don:
Some 20 years ago, we concluded from a study of these two
standard ABR measures in a large series of tumor cases, that
these measures detected medium and large size tumors but
that many tumors smaller than 1 cm would be missed.
Studies over the last 10 years have confirmed this finding that
standard ABR measures, frequently miss small tumors.
“ABR yields high detection scores: up
to 90%. The larger the tumor, the
easier its detection. It is likely that small
tumors (< 1 cm) will be missed.”
Eggermont JJ, Don M, Brackmann DE.
Electrocochleography and auditory brainstem
responses in patients with pontine angle tumors.
Ann Otol, Rhinol, and Laryngol, Suppl. 1980; 75: 1-19.
I. Background: Standard ABR
HOUSE EAR INSTITUTE
Summary of Standard ABR Test
Manny Don:

Detects nearly
all medium
and large
acoustic
In summary
of the use of standard
ABR measures
for
tumors. detecting acoustic tumors, studies have shown that:
(bullet 1)

Misses 30-50%
(bullet 2) of small (<1 cm) acoustic tumors.
The consequence of this failure and the advent of
Magnetic Resonance Imaging or MRI is that for many
clinics…
Consequence of failure to detect small tumors
All patientsIn with
suspicious
hearing
essence,
MRIs are used clinical
to screen for
acoustic and
tumors.
balance symptoms
are sent for Magnetic
Resonance Imaging (MRI).
I. Background: Standard ABR
HOUSE EAR INSTITUTE
Manny Don:
However, there are some...
Drawbacks of Screening with MRI

Relatively expensive ($2100)

Not available everywhere

Invasive, anxiety producing, and uncomfortable test
for some patients
 Cannot be used on patients with implanted metal
devices or materials

Most patients tested do not have a tumor
I. Background: Standard ABR
HOUSE EAR INSTITUTE
Why do standard ABR measures often fail
to detect small tumors ?


The obvious reason: Small tumors exert less
pressure and affect a smaller number of neural
fibers than larger tumors.
But, these not the only factors because:
1. Many of these small tumors exert enough
pressure to cause clinical symptoms.
2. Many small tumors are detected by standard
ABR measures.
I. Background: Standard ABR
HOUSE EAR INSTITUTE
Manny Don:
Why do standard ABR measures often fail
Our hypothesis is:...
to detect small tumors ?
So, the question is,…
Hypothesis: Standard ABR measures often fail to
detect small tumors because these measures are
dominated by activity from a subset of 8th nerve
fibers that may not be affected by the small tumor.
Thus, the limitation is not with ABRs per se, but with
the ABR measures used.
I. Background: Standard ABR
HOUSE EAR INSTITUTE
Manny Don:
In order to understand the failure of the standard
ABR measures, we need to understand what
these measures represent.
Background: Evolution of the Stacked ABR
A. Eighth Nerve Tumors
B. Standard ABR Tumor Detection
C. What Do Standard ABR Measures Represent?
D. Limitations of Standard ABRs
I. Background/What Do Standard ABRs Represent?
HOUSE EAR INSTITUTE
Manny Don:
(self-explanatory)
The IT5 and I-V Delay Use Wave V
Latency Measures:
What does the latency of the ABR
wave V represent?
I. Background/What Do Standard ABRs Represent?
HOUSE EAR INSTITUTE
Manny Don:
To begin our discussion of what wave V latency in
the ABR represents, I want to clear up ...
Two Prevalent Misconceptions About
Click-evoked ABRs
1. Clicks have only high-frequency energy.
2. ABRs can only test cochlear function from
2 to 4 kHz.
I. Background/What Do Standard ABRs Represent?
HOUSE EAR INSTITUTE
Manny Don:
First, let us take a look at the spectral energy of a click.
Here we see the amplitude spectrum of a click that is
produced by applying a 100 µsec pulse to a TDH-49
earphone. You can see that there is considerable energy
from 5 kHz down to at least 200 Hz. Energy falls off above
5 kHz in part because of the response characteristics of the
earphone and because of the width of the pulse. We will
not discuss this further as there are several technical issues
that go beyond the scope of this talk. The point here is that
there is considerable low-frequency energy in a click
stimulus.
HOUSE EAR INSTITUTE
Manny Don:
V
60 dB nHL Clicks
In this slide we have a series of ABR waveforms. The top
trace (white)
is the standard
Standard
ABR ABR obtained with wide -band
click stimuli. The succeeding traces represent ABRs from
octave wide regions of the cochlea. The center
frequencies of
V
these cochlear
are noted to the left ofV each trace. I will
CF regions
= 11.3 kHz
discuss later how these ABRs from place-specific regions of
CF =obtained.
5.7 kHz Adding these traces together result
the cochlea are
in the standard ABR waveform shown in the top trace. The
main point to see here is that, while the wave V latency of the
V
standard ABR at the top is dominated by cochlear activity from
CF = 2.8
kHz
the high frequency
regions,
there is considerable evoked
activity from the lower frequency regions as well. However,
V
this evoked activity from the lower frequency cochlear
regions
is phase cancelled
and
CF = 1.4
kHzonly the high-frequency contributions
are evident.
V
CF = 0.7 kHz
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15
ms
HOUSE EAR INSTITUTE
Manny Don:
This slide illustrates this point. The top trace is again
the standard ABR to wide-band click stimuli. The
next two traces
are ABRs
Standard
ABRfrom octave-wide regions
centered at 11.3 and 5.7 kHz. If these two ABRs are
added together, the resultant waveform is shown in
the bottom trace. Note that the latency as well as
the amplitude of the standard ABR at the top is very
CF = 11.3 kHz
close to that of the sum of just these two highfrequency ABRs.
CF = 5.7 kHz
60 dB nHL Clicks
11.3 kHz + 5.7 kHz
0
1
2
3
4
5
6
7
8
ms
9 10 11 12 13 14 15
HOUSE EAR INSTITUTE
I. Background/What Do Standard ABRs Represent?
Manny Don:
Self-explanatory
Take Home Messages

A click stimulus is a wideband acoustic signal with
as much low-frequency energy as there is highfrequency (HF) energy.

The click-evoked ABR contains neural activity
representing all frequency regions of the cochlea,
not just the HFs.

In the standard ABR, wave V latency is dominated
by HF regions because lower frequency
contributions are phase-cancelled.
I. Background/What Do Standard ABRs Represent?
HOUSE EAR INSTITUTE
Manny Don:
Let us now focus on the limitations of the
standard ABRs with respect to small
tumor detection.
Background: Evolution of the Stacked ABR
A. Eighth Nerve Tumors
B. Standard ABR Tumor Detection
C. What Do Standard ABR Measures Represent?
D. Limitations of Standard ABRs
I. Background/Standard ABRs Limitations
HOUSE EAR INSTITUTE
Manny Don:
Normal Internal Auditory Canal (IAC)
Let us now represent standard ABR in terms of
the auditory nerve fibers in the internal auditory
canal or IAC. I am not trying to show the
actual distribution of the high-frequency fibers
here
in the auditory
Facial
Nervenerve bundle. The
important point is that the high-frequency fibers
which dominate the wave V latency of the
standard ABR measure is only a subset of the
fibers in the nerve bundle.
Sup.
Vest. Nerve
Standard
ABR
High-frequency
Acoustic Nerve
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE
I. Background: Standard ABRs Limitations
Manny Don:
Medium or Large Tumor in IAC
This illustrates how a large tumor will usually affect
a large number of auditory nerve fibers especially
those from the high-frequency regions. As a result,
the standard ABR measure (latency) is abnormal.
Facial Nerve
Abnormal
Standard
ABR
Acoustic Nerve
Sup.
Vest. Nerve
Tumor
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE
I. Background: Standard ABRs Limitations
Manny Don:
Small Tumor in IAC
A small tumor affects fewer fibers but could affect sufficient
high-frequency fibers to cause an abnormal standard ABR
latency measure. Abnormal standard ABR measures are
obtained in about half of the small tumors cases.
Facial Nerve
Sup.
Vest. Nerve
Abnormal
Standard
ABR
Acoustic Nerve
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE
I. Background: Standard ABRs Limitations
Manny Don:
Small Tumor in IAC
However, it is easy to see the possibility that a small tumor
could affect a small number of fibers that do not include very
many high-frequency fibers. As a result, the wave V latency of
the standard ABR could still be normal. Our hypothesis is
that
this is the
reason the standard ABR fails to detect many
Facial
Nerve
small tumors.
Sup.
Vest. Nerve
Normal
Standard
ABR
Acoustic Nerve
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE
I. Background: Standard ABRs Limitations
Summary

The wave V latency used in standard ABR IT5 and
I-V delay measures is dominated by neural activity
from high-frequency (HF) regions of the cochlea.

If the tumor does not affect these HF fibers
sufficiently, the standard ABR latencies will be
normal.

Small tumors do not always affect HF fibers, so
they may be missed by standard ABR measures.
I. Background/What Do Standard ABRs Represent?
HOUSE EAR INSTITUTE
Manny Don:
What would be the advantages of developing
an ABR measure for tumor screening instead of
using the MRI test?
Advantages of ABRs over MRIs
The ABR is:

Much less expensive

More widely available

Non-invasive

More comfortable than MRI
II. Stacked ABR: Rationale
HOUSE EAR INSTITUTE
The Challenge
Can we develop an ABR test that:
1. Detects small tumors that cause symptoms, and
2. Significantly reduces the number of patients
sent for MRI who do not have a tumor?
II. Stacked ABR: Rationale
HOUSE EAR INSTITUTE
Manny Don:
Small Tumor in IAC
In order to develop an effective ABR test to screen for small
tumors, we have to solve this problem where the tumor does
not affect the nerve fibers that are critical for the ABR
measure. In the case of the standard ABR, the high-frequency
subset
of fibers.
Facial
Nerve
Normal
Standard
ABR
Acoustic Nerve
II. Stacked ABR: Rationale
Sup.
Vest. Nerve
High-frequency
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE

Hypothesized that tumor detection fails with
standard ABR measures because these measures
are dominated by high-frequency activity and small
tumors may not always affect the high-frequency
fibers.

In order for a new ABR measure to detect small
tumors, it must measure activity from essentially all
fibers, not just a subset.
II. Stacked ABR: Rationale
HOUSE EAR INSTITUTE
Manny Don:
Normal IAC
In other words, we need a new ABR measure
that is based on essentially all the nerve fibers
(encompassed by the black circle), not just a
subset.
Facial Nerve
Sup.
Vest. Nerve
New
ABR
Acoustic Nerve
II. Stacked ABR: Rationale
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE
Manny Don:
Normal IAC
A solution would be to divide the whole
auditory nerve into five groups and use the
activity from these groups in a new ABR
measure. These five groups could
represent
Facialfrequency
Nerve regions of the cochlea
that span the whole frequency range and,
therefore, the the whole auditory nerve.
Sup.
Vest. Nerve
New
ABR
2
1
3
5
Acoustic Nerve
ABR: Rationale and Method
II. StackedStacked
ABR: Rationale
4
Inf . .
Vest. Nerve
HOUSE EAR INSTITUTE
Manny Don:
An example of a...
Diagnostic Test: If you add the activity from
each of the five areas, is the amplitude normal?
Activity from area 1
+
Activity from area 2
+
Activity from area 3
+
Activity from area 4
+
Activity from area 5
1
2
3
4
Normal Amplitude
5
II. Stacked ABR: Rationale
HOUSE EAR INSTITUTE
Manny Don:
Medium or Large Tumor in IAC
Let’s see how this new ABR test would work with the various
tumor examples presented earlier: First, in example of the
medium or large tumor, we can see that many fibers from areas
New ABR: Abnormal
2,3, and 4 would be compromised. In addition to the fibers
obscured by the tumor, stippled fibers represent fibers that are
also affected by the encroaching tumor. Reduction of the
contributions from these affected areas would result in an
abnormally low overall amplitude of the added activity.
1
2
3
4
Tumor
3
5
Normal
Tumor
Acoustic Nerve
II. Stacked ABR: Rationale
HOUSE EAR INSTITUTE
Manny Don:
Small Tumor in IAC
In the case where the small tumor affected sufficient
high-frequency fibers to produce an abnormal standard
ABR
latency
measure,
it would also reduce the
New
ABR:
Abnormal
contribution from areas 2 and 3 and produce an
abnormally low amplitude of the added activity from all
areas.
1
2
3
4
5
Normal
Tumor
Acoustic Nerve
II. Stacked ABR: Rationale
HOUSE EAR INSTITUTE
Manny Don:
Small Tumor in IAC
Finally, let’s look at the case where the
small tumor did not affect a sufficient
number of high-frequency fibers, and
therefore, the standard ABR latency
measures
normal.
Facialwere
Nerve
Sup.
Vest. Nerve
Normal
Standard
ABR
Acoustic Nerve
II. Stacked ABR: Rationale
Inf.
Vest. Nerve
HOUSE EAR INSTITUTE
Manny Don:
Small Tumor in IAC
For this troublesome case, the new ABR measure would
still be abnormal because the contributions from areas 3
andNew
4 would
be reduced,
espcially area 4. Because this
ABR:
Abnormal
new measure involves contributions from essentially all
the nerve fibers, it doesn’t matter where the tumor is
located or which fibers are affected as long as there has
been compromise of a sufficient number of fibers.
1
2
3
4
5
Normal
Tumor
Acoustic Nerve
II. Stacked ABR: Rationale
HOUSE EAR INSTITUTE
New ABR Measure
Manny Don:
Requirements
Proposed
This new ABR measure that involves essentially
all of Methods
the
auditory nerve fibers is the Stacked ABR measure. The
Wideband Click
1. An auditory
signal
following
is athat
list of the requirements of such a measure and
stimulates
all
theessentially
proposed methods
for fulfilling the requirements. First
need ... of the
frequencyweregions
cochlea
2. A method for separating
the responses from
different frequency regions
of the cochlea
The Derived-band
ABR Technique
3. A procedure for summing
the responses to
approximate total neural
activity
The Stacking
Technique
II. Stacked ABR: Rationale
HOUSE EAR INSTITUTE
Stacked ABR: Rationale and Method
A. Derived-band ABRs
1. Stimuli
Manny a.
Don:Clicks
b.ourHigh-pass
masking
We begin
Stacked ABR discussion
withnoise
the derived-band ABR
method
which fulfills the subtraction
stimulus requirement and the procedure
2.
Response
for separating out activity from different parts of the cochlea, and
therefore,
differentderived-band
subsets of auditory nerve
fibers. The stimuli
B.
Stacking
ABRs
are clicks and high-pass masking noise. The responses to various
combinations
of clicks and high-pass noises are subtracted from
C.
ABR recordings
each other to obtain the derived-band ABRs representing activity
initiated from different frequency regions of the cochlea. I’ll
demonstrate this later.
II. Stacked ABR: Method
HOUSE EAR INSTITUTE
Manny Don:
TDH-49
Recall earlier that a click produced by a
100 µsec pulse results in a broad-band
stimulus with energy in both high and low
frequencies. This spectrum is shown
again in the top trace. Thus, when
presented at aboutClick
60 dB NHL, this click
will stimulate most of the cochlea.
The traces below show the spectrum of
the high-pass masking noise that is used
to mask activity from various parts of the
cochlea.
High-pass
Masking Noise
II. Stacked ABR: Derived-band ABRs/Stimuli
HOUSE EAR INSTITUTE
Base
Manny
Don:
8
Frequency kHz
4
2
1
Apex
0.5
Now I would like to demonstrate how we can use click
stimuli and high-pass masking noise to obtain ABRs that
are related to activity from a specific place and frequency
region of the cochlea. In this slide, we see a schematic
showing the nerve fibers coming from five regions of the
cochlea. The five regions are color-coded and going from
ABR alone
to
Click
Alone
response
left to right, frequency goes from high to low. Click
Shown
below
(Standard
(Standard ABR)
ABR)
is a cross-section of the auditory nerve. When stimulated
by
a click,
essentially all the fibers are activated and the
Nerve
Fibers
resultant ABR (shown on the right) represents activity from
all parts of the cochlea. This is the click-evoked standard
ABR with which we all are familiar. I will refer to this as the
standard unmasked ABR.
HOUSE EAR INSTITUTE
II. Stacked ABR: Derived-band ABRs/Response Subtraction
Manny Don:
Frequency kHz
Apex
If we now present 8the clicks
masking
4 with 82 kHz high-pass
1
0.5
Base
noise, the resultant response comes from the unmasked
regions of the cochlea, i.e., below 8 kHz. The masked portion,
8 kHz and above, is shown as blackened fibers.
M
Click + 8 kHz HighABR masking
to Click +noise
8 kHz
pass
High-pass Masking Noise
Nerve Fibers
HOUSE EAR INSTITUTE
II. Stacked ABR: Derived-band ABRs/Response Subtraction
Manny Don:
Bas e
8
Fre quency kHz
4
2
1
Ape x
0.5
If we now subtract the 8 kHz masked response from the
unmasked response, we obtain a derived-band response. Note
that because the activity from below
the
8 kHz region is in both the
Click
response
ABRalone
to Click
Alone
(Standard
ABR)
(Standard ABR) this activity is removed
unmasked
and 8 kHz masked conditions,
Nerve Fibers
by the subtraction process. Thus, the derived-band response
represents activity only from above the 8 kHz region. Theoretically,
this derived-band activity represents an octave wide region
Derived-band ABR
centered at about 11.3 kHz.
CF = 11.3 kHz
Bas e
8
Fre quency kHz
4
2
1
Ape x
0.5
Bas e
8
Fre quency kHz
4
2
1
Ape x
0.5
M
Click
+ 8 kHz
ABR
to Click
+ 8 HighkHz
pass masking noise
High-pass masking noise
Nerve Fibers
Nerve Fibers
HOUSE EAR INSTITUTE
II. Stacked ABR: Derived-band ABRs/Response Subtraction
Manny Don:
Frequency
Apex masking noise,
Base
Next, if we present
the clicks
with kHz
4 kHz high-pass
0.5 Thus, the
1
2 are removed.
4
contributions from84 kHz and
above
resultant ABR represents activity from below the 4 kHz region.
M
M
ABR to Click + 4 kHz
High-pass masking noise
Nerve Fibers
HOUSE EAR INSTITUTE
II. Stacked ABR: Derived-band ABRs/Response Subtraction
Manny Don:
Bas e
8
Fre quency kHz
4
2
1
Ape x
0.5
If we now subtract
the 4 kHz high-pass masked
M
response from the 8 kHz high-pass masked
response, we obtain the next derived-band
ABR to
8 kHz
Click
+ 8Click
kHz +Highresponse. By subtraction, the
activity
below
pass
masking
noise
High-pass
masking
noise 4 kHz
Fibers
which isNerve
common
to both the 4 kHz and 8kHz
masked response, is removed. The region above
8 kHz is masked in both conditions. Thus, the
derived-band response represents activity
between 4 and 8 kHz; activity that was unmasked
in the 8 kHz condition but masked in the 4 kHz
high-pass Bas
condition.
Theoretically,
this derivedFre quency kHz
Ape x
e
8
4
2
1
0.5
band activity represents
an
octave
wide region
M 5.7 kHz.
centered atMabout
Derived-band ABR
CF = 5.7 kHz
Bas e
8
Fre quency kHz
4
2
1
Ape x
0.5
M
Nerve Fibers
ABR to Click + 4 kHz
High-pass masking noise
Nerve Fibers
HOUSE EAR INSTITUTE
II. Stacked ABR: Derived-band ABRs/Response Subtraction
Manny Don:
Next, if we present
the clicks
with kHz
2 kHz high-pass
Frequency
Apex masking
Base
noise, contributions
2 kHz and
Thus,
8 from 4
2 above
1 are removed.
0.5
the resultant ABR represents activity from below the 2 kHz
M
M
M
region.
ABR to Click + 2 kHz
High-pass masking noise
Nerve Fibers
HOUSE EAR INSTITUTE
II. Stacked ABR: Derived-band ABRs/Response Subtraction
Manny Don:
Frequenc y k Hz
4
2
Base
8
1
Apex
0.5
As before, if we now subtract the 2 kHz high-pass masked response from the 4
M
M
kHz high-pass masked response, we obtain the next derived-band response. By
subtraction, the activity below 2 kHz, which is common to both the 2 kHz and 4k
Hz masked responses, is removed.
region above 4 kHz is masked in both
ABR to ClickThe
+ 4 kHz
High-pass masking noise
conditions.
the derived-band response represents activity between 2 and 4
Nerve FiThus,
bers
kHz; activity that was unmasked in the 4 kHz condition but masked
in the 2 kHz
Derived-band
ABR
CF = 2.8 an
kHzoctave
high-pass condition. Theoretically, this derived-band activity represents
wide region centered at about 2.8 kHz.
Frequenc y k Hz
4
2
Base
8
Frequenc y k Hz
4
2
Base
8
M
M
1
Apex
0.5
M
1
Apex
0.5
M
M
ABR to Click + 2 kHz
High-pass masking noise
Nerve Fi bers
Nerve Fi bers
HOUSE EAR INSTITUTE
II. Stacked ABR: Derived-band ABRs/Response Subtraction
Manny Don:
Summary:...
Note that the latency of wave V of the
Derived-band
ABR Summary
derived-band
ABRs increase
in latency
astheNeural
frequency
region it represents
contributions
from
becomes
lower.
This
progressive
different frequency regionsdelay
down the cochlea is the basis for the
of the cochlea can be
phase cancellation of the low frequency
obtained
using
the derivedcontributions
to the
unmasked
band ABR method.
response.

Derived-band ABRs represent
activity from more specific
frequency regions than
moderate-to-high level
toneburst-evoked ABRs.
V
Unmasked
(Standard) ABR
V
CF = 11.3 kHz
V
CF = 5.7 kHz
V
CF = 2.8 kHz
V
CF = 1.4 kHz
V
CF = 0.7 kHz
0
2
4
6
8
ms
10
12
14
II. Stacked ABR: Method/Derived ABRs
HOUSE EAR INSTITUTE
Stacked ABR: Rationale and Method
A. Derived-band ABRs
1. Stimuli
a. Clicks
b. High-pass masking noise
2. Response subtraction
B. Stacking derived-band ABRs
C. ABR recordings
II. Stacked ABR: Method/Stacking
HOUSE EAR INSTITUTE
Manny Don:
The Stacked ABR Technique
The temporal shifting and aligning of the responses removes the
phase-canceling
effects.
thebyresultant
Stacked ABR
 The Stacked
ABR isThus,
formed
first
represents
contributions
temporally
aligning from
waveall
V parts
of theof the cochlea.
derived-band ABRs, then summing
the responses.


Aligning the derived-band ABRs
eliminates phase cancellation of
lower frequency activity. Thus, the
Stacked ABR amplitude
reflects activity from all frequency
regions of the cochlea, not just the
high frequencies.
Reduction of any neural activity
due to a tumor, even a small
tumor, will result in a reduction
of the Stacked ABR amplitude.
Sum of
Shifted
Waveforms
Stacked
ABR
Shi fted to 5.7 kHz
Wa ve V l aten cy
CF = 11.3 kHz
CF = 5.7 kHz
CF = 2.8 kHz
CF = 1.4 kHz
CF = 0.7 kHz
0
2
4
6
8
ms
10
12
14
II. Stacked ABR: Method/Stacking
HOUSE EAR INSTITUTE
Manny Don:
The next series of
slides will illustrate
how amplitude
measures of the
standard ABR fail to
reflect all the neural
activity whereas the
Stacked ABR does.
This slide shows for
a non-tumor normal
hearing individual
the standard ABR
and the derived
bands with their
CF = 11.3 kHz
natural delays in the
left panel.CF
The
rightkHz
= 5.7
panel shows the
same waveforms but
= 2.8
aligned toCF
form
ths kHz
Stacked ABR.
Stacked ABR
Derived-Bands
Derived-Bands Aligned
(Actual timing)
(Shifted and summed)
Stacked ABR
Standard ABR
CF = 1.4 kHz
CF = 0.7 kHz
0 2 4 6 8 10 12 14
ms
II. Stacked ABR: Rationale and Method--Stacking derived-band ABRs
0 2 4 6 8 10 12 14
ms
HOUSE EAR INSTITUTE
Manny Don:
Example
This illustrates
what 1
happens to the
amplitude of the
standard ABR if
activity from the
lowest two
frequency bands are
removed. The
amplitude drops by
only 10%.
Standard ABR
Standard ABR minus 2 bands
}
10% Reduction
CF = 11.3 kHz
CF = 5.7 kHz
CF = 2.8 kHz
CF = 1.4 kHz
CF = 0.7 kHz
0
2
4
6
8
10 12 14
0
ms
II. Stacked ABR: Rationale and Method--Stacking derived-band ABRs
2
4
6
8
10 12 14
ms
HOUSE EAR INSTITUTE
Manny Don:
Example 1
For the same
individual in the
previous slide, this
illustrates how the
removing the lowest
two frequency bands
reduced the Stacked
ABR by 33%. This
shows how the
Stacked ABR is
more sensitive to the
reduction of neural
activity evoked by
the click.
Stacked ABRs
Stacked ABR minus 2 bands
33 % Reduction
CF = 11.3 kHz
CF = 5.7 kHz
CF = 2.8 kHz
CF = 1.4 kHz
CF = 0.7 kHz
0
2
4
6
8
ms
10 12 14
0
2
4
II. Stacked ABR: Rationale and Method--Stacking derived-band ABRs
6
8 10 12 14
ms
HOUSE EAR INSTITUTE
Manny Don:
2
This is a Example
more
dramatic illustration
from another nontumor normalhearing individual as
to what happens to
the amplitude of the
standard ABR if
activity from the
lowest two
frequency bands are
removed. The
amplitude actually
increases by 12%.
The reason for an
increase instead of
an expected
decrease is due to a
variation in the
phase canceling
effects of the two
lower frequency
bands.
Standard ABR
12% Increase
Standard ABR minus 2 bands
}
CF =11.3kHz
CF =5.7kHz
CF =2.8kHz
CF =1.4kHz
CF =0.7kHz
0 2 4 6 8 10 12 14
0 2 4 6 8 10 12 14
ms
II. Stacked ABR: Rationale and Method--Stacking derived-band ABRs
ms
HOUSE EAR INSTITUTE
MannyExample
Don: 2
Stacked ABR
For the same
individual in the
previous slide, this
illustrates how the
removing the lowest
two frequency bands
reduced the Stacked
ABR by 23%
whereas the
standard amplitude
had increased by
12%. It is logical to
expect a reduction
instead of an
increase when
activity is removed.
This again shows
how the Stacked
ABR is more
sensitive and
reflects reduction of
neural activity
evoked by the click.
Stacked ABR minus 2 bands
23 % Reduction
CF =11.3kHz
CF =5.7kHz
CF =2.8kHz
CF =1.4kHz
CF =0.7kHz
0 2 4 6 8 10 12 14
0 2 4 6 8 10 12 14
ms
II. Stacked ABR: Rationale and Method--Stacking derived-band ABRs
ms
HOUSE EAR INSTITUTE
Manny Don:
Normal IAC
Thus, in summary, the Stacked ABR represents activity
from all parts of the cochlea and should be more
sensitive to small tumors.
Facial Nerve
STACKED
New
ABR
2
1
3
5
Acoustic Nerve
Stacked ABR: Rationale and Method
II. Stacked ABR: Stacking
Sup.
Vest. Nerve
4
Inf. .
Vest. Nerve
HOUSE EAR INSTITUTE
Manny Don:
I would like to take a few moments to discuss some
aspects about ABR recordings that are critical to
the Stacked ABR.
Stacked ABR: Rationale and Method
A. Derived-band ABRs
1. Stimuli
a. Clicks
b. High-pass masking noise
2. Response subtraction
B. Stacking derived-band ABRs
C. ABR recordings
II. Stacked ABR: Method/ABR Recordings
HOUSE EAR INSTITUTE
Manny Don:
ABR Recordings
The recording parameters in terms of
amplification, filter passband, and filter slope
 Electrode
Recording
are fairly standard.
To facilitate
consistencyMontage
in
our recordings and to reduce variability in our
database, our Cz electrode is located at the
intersection that is midway between the nasion
and inion and midway between the two ears.
Cz - Positive
We record differentially between Cz and the
50% (nasion to inion)
ipsilateral mastoid.
50% (ear to ear)
Ipsi Mastoid - Negative
Contra Mastoid - Ground
 Amplification: 5 X 105
 Filter Passband: 0.1 - 3 kHz
 Filter Slope: 12 dB/octave
II. Stacked ABR: Method/ABR Recordings
HOUSE EAR INSTITUTE
Manny Don:
ABR Recordings
An important aspect of the Stacked ABR is to make
sure that the derived-ABR responses that are “stacked”
have minimal physiological noise. We want the
measure to represent neural activity related to
stimulation rather than background physiological noise.
To accomplish this, we perform estimates of the
residual noise in the ABR, perform an average that is
1. Estimation of Unaveraged Noise (Fsp estimation)
weighted towards blocks of sweeps that have the least
physiological background noise, and terminate the
2. Weighted
Averaging
weighting)
averaging
when the estimate
is a low 20(Bayesian
nV.
 Minimization of Physiological Noise in ABRs
3. Termination of Averaging When Residual
Noise Level is low ( 20 nV)
II. Stacked ABR: Method/ABR Recordings
HOUSE EAR INSTITUTE
Manny Don:
This slide is taken from some work by Claus Elberling
demonstrating the effect of the averaged noise on the recorded
ABR. It should be remembered that the recorded ABR contains
both the “true ABR” and the averaged background noise. In
reality, we cannot separate out these two contributions to the
recorded ABR. The hope is that we perform sufficient averaging
to reduce the averaged background noise to a level where it is
contributing little to the recorded ABR so that the recorded ABR is
a good representation of the ‘true ABR”. In this slide, we see
how on one run the averaged noise has a waveform that has
random peaks that coincide with the true ABR wave V peak.
Thus, the recorded ABR has a an amplitude that is larger than
the true amplitude. On another run, a random noise peak is
antiphasic with the true ABR wave V peak and the recorded ABR
wave V is much smaller. This is the major reason that wave V
amplitude seems to vary from run to run. Thus, it is important to
average until the residual noise in the waveform is very small so
that the recorded ABR is a close reflection of the true response.
HOUSE EAR INSTITUTE
II. Stacked ABR: ABR Recordings/Noise
ABR Recordings
 Minimization of Physiological Noise in ABRs
1. Estimation of Unaveraged Noise (Fsp estimation)
2. Weighted Averaging (Bayesian weighting)
3. Termination of Averaging When Residual
Noise Level is low ( 20 nV)
II. Stacked ABR: Rationale and Method--ABR recordings
HOUSE EAR INSTITUTE
Manny Don:
This illustrates the
difference in
waveform
repeatability when a
fixed number of
sweeps are used
(left panel) vs when
a fixed signal to
noise ratio (Fsp
value- from paper by
Elberling & Don,
1984).
Demonstrates that
the run to run
variability is mainly
due to the averaged
noise.
II. Stacked ABR: Rationale and Method--ABR recordings
HOUSE EAR INSTITUTE
ABR Recordings
 Minimization of Physiological Noise in ABRs
1. Estimation of Unaveraged Noise (Fsp estimation)
2. Weighted Averaging (Bayesian weighting)
3. Termination of Averaging When Residual
Noise Level is low ( 20 nV)
II. Stacked ABR: Rationale and Method--ABR recordings
HOUSE EAR INSTITUTE
Estimated Residual Noise as a
Function of Accepted Sweeps
140
120
Subject 1 Theoretical
Subject 1 Estimated
Subject 2 Estimated
100
80
60
40
20
0
0
2000
4000
6000
8000
Manny Don:
Plots of the
estimated residual
noise in an ABR as
a function of the
number of sweeps
averaged. Illustrates
that the key to low
residual noise is a
quiet person.
Subject 1 is noisy
but the reduction of
the noise with
averaging is
predicted by the
averaging principle
(√N). At end of
nearly 10,000
sweeps noise in the
average is barely
equivalent to subject
2 after 500 sweeps.
Averaging can only
do so much.
10000
Accepted Sweeps
II. Stacked ABR: Rationale and Method--ABR recordings
HOUSE EAR INSTITUTE
Normal Averaging
Weighted Averaging
1
(0.5)
2
(0.8)
3
(1.0)
4
Figure from Elberling
and Wahlgreen
1 (2.03)
(Scand. Audiol., 1985)
comparing results of
2 normal
(1.35) averaging (all
sweeps equally
weighted, left panel)
3 versus
(1.18) weighted
averaging using the
4 Bayesian
(3.20)
estimation
technique to weight
blocks of sweeps
5 according
(0.04)
to the
amount of estimated
residual noise in that
6 block
(0.05)of sweeps (right
panel). Get better
with
7 reproducibility
(0.07)
weighted average.
This minimizes the
(0.07)
8 destructive
effect of
episodic noise shown
Avg.
in(X4)
previous slide.
Avg.
(0.3)
5 (31.7)
6 (19.8)
7
(17.2)
(17.4)
8 Avg.
(X4)
Avg.
0
2 4
6
8 10 12 14 16
II. Stacked ABR: Rationale and Method--ABR recordings
Manny Don:
0
2 4
6
8 10 12 14 16
HOUSE EAR INSTITUTE
40
Manny Don:
Effect of Bayesian Weighting
on Estimated Residual Noise
Illustrates how the
35
Bayesian estimation
for weighted
averaging
30
essentially removes
the deleterious effect
of the episodic noise
shown two slides
25
previously.
Gold filled triangles
show how the
20
estimated residual
noise nearly mimics
the theoretical noise
levels when using
15
the Bayesian
estimation technique
for weighted
10
averaging.
Actual
Theoretical
Bayes
5
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
Accepted Sweeps
II. Stacked ABR: Rationale and Method--ABR recordings
HOUSE EAR INSTITUTE
Summary of Noise in ABRs

All ABRs contain unaveraged background
physiological noise (i.e. residual noise)

Methods for estimating the amount of
noise in the ABR can be used to:
1. Give greater weight to sweeps with
the lowest noise, and
2. Determine when to stop averaging.
II. Stacked ABR: ABR Recordings/Noise
HOUSE EAR INSTITUTE
CAUTION!
 Since the Stacked ABR is an amplitude
measure, the amount of noise left in the
response is critical.

For a more accurate estimate of neural
activity, the residual noise level must be low.

If the residual noise level is high, then the
Stacked ABR must be interpreted with caution.
II. Stacked ABR: ABR Recordings/Noise
HOUSE EAR INSTITUTE
ABR Recordings
 Minimization of Physiological Noise in ABRs
1. Estimation of Unaveraged Noise (Fsp estimation)
2. Weighted Averaging (Bayesian weighting)
3. Termination of Averaging When Residual
Noise Level is low ( 20 nV)
II. Stacked ABR: Rationale and Method--ABR recordings
HOUSE EAR INSTITUTE
Manny Don:
To insure that the
stacked ABR is
composed of
averages that
contain mostly
evoked neural
responses,
averaging is
terminated when the
residual noise is 20
nV or less. Based
on this conservative
criterion, subject 1
would not have
achieved criterion in
over 10000 sweeps
whereas subject 2
achieved this level in
about 1500 sweeps.
Now exploring if the
criterion noise level
can be raised
without
compromising the
effectiveness of the
Stacked ABR.
Estimated Residual Noise as a
Function of Accepted Sweeps
140
120
Subject 1 Theoretical
Subject 1 Estimated
Subject 2 Estimated
100
80
60
40
Residual Noise Criterion
20
0
0
2000
4000
6000
Accepted Sweeps
II. Stacked ABR: Rationale and Method--ABR recordings
8000
10000
HOUSE EAR INSTITUTE
Clinical Judgement

There will be cases where the residual noise
level remains high even after thousands of
sweeps.

In these cases, make sure the patient is
comfortable and as relaxed as possible before
you continue.

If the patient remains noisy, then the Stacked
ABR must be interpreted with caution.
II. Stacked ABR: ABR Recordings/Noise
HOUSE EAR INSTITUTE
Take Home Message 5
 Residual noise levels must be low so that the
Stacked ABR amplitude mostly reflects total neural
activity, not noise.
 Stacked ABRs formed from noisy recordings may
be unreliable and should be interpreted with
extreme caution.
Questions?
II. Stacked ABR: Rationale and Method--ABR recordings
HOUSE EAR INSTITUTE
Talk Outline
I. Background: Evolution of the Stacked ABR
II. Stacked ABR: Rationale and Method
III. Stacked ABR: Studies
IV. Tumor Screening Protocol
V. Future Stacked ABR Work
Talk Outline
HOUSE EAR INSTITUTE
Stacked ABR: Studies
A. Interpretation of Results
1. Basics of Data Presentation
2. Sensitivity vs. Specificity
B. Published Study
C. Preliminary Work
III. Stacked ABR/Interpretation of Results
HOUSE EAR INSTITUTE
Manny Don:
50%
Percentile
In the top graph is the well know normal distribution curve. It
simply describes how the values
of X are distributed
in a
25%
75%
population. The mean X value of this normal distribution is here
90% 50% of
10%
where the peak of the distribution
occurs. In essence,
the population have values less than this mean and 50% of the
population have values greater.
X The x value where 75% occurs,
means that 75% of the population have values at or below and
25% have values above. This normal distribution can be
replotted in terms of a cumulative distribution shown at the
bottom. These percent values now form the y-axis with the X
values on the x-axis.
90% From this cumulative curve, we can pick
any x value and determine what percent of the population have
75%
values equal or lower,
or vice versa.
Normal
Distribution
Cumulative
Distribution
50%
25%
10%
X
HOUSE EAR INSTITUTE
III. Stacked ABR: Results/Data Presentation
Manny Don:
Cumulative Distribution Curve
Percentile
Let’s use the example of the distribution of Stacked ABR values plotted on
a cumulative curve. If we ask, 50% of the test population have Stacked
ABR amplitudes that fall below what value? We simply find where the
50% point intersects the curve and from that point where it intersects on
the amplitude
100 axis. The answer in this case is 880 nV.
50% of the test
population have
Stacked ABR
amplitudes that
fall below what
value?
80
60
50
40
20
880 nV
0
200
400
600
800
1000 1200 1400
Stacked ABR (nV)
III. Stacked ABR: Results/Data Presentation
HOUSE EAR INSTITUTE
Cumulative Distribution Curve
Manny Don:
Percentile
Going the other direction, we can ask, “What percentage of
the test population has Stacked ABR amplitudes less than
1000 nV. From the 1000 nV point we draw a line to the
100
82 %
cumulative distribution curve and from that intersection with
the curve we draw a line to where it intersects the What percentage of
80
percentage axis. In this case, it is determined that 82% of
the test
population
the
test
population
have
Stacked
ABR
amplitudes
equal
to
60
has Stacked ABR
or less than 1000 nV.
amplitudes < 1000 nV?
40
20
0
200
400
600
800
1000 1200 1400
Stacked ABR (nV)
III. Stacked ABR: Results/Data Presentation
HOUSE EAR INSTiTUTE
Sensitivity vs. Specificity of an ABR Measure
Sensitivity
Specificity
For a given criterion of
an ABR measure, the
sensitivity is defined as
the percentage of tumor
cases detected (i.e., the
true-positive rate).
For a given criterion of
an ABR measure, the
specificity is defined as
the percentage of nontumor cases that are
correctly identified (i.e.,
the true-negative rate).
III. Stacked ABR: Results/Sensitivity vs. Specificity
HOUSE EAR INSTITUTE
Manny Don:
Percentile
At the top of this figure are the distributions of an ABR test meas
for the non-tumor and tumor populations. We can see that the
values for the non-tumor population are greater than for the tumo
Tumor
Non-Tumor
population.
These
distributions are replotted below as cumulativ
distributions. If we wanted to use this test measure to detect all o
the tumors in this population, we would choose the ABR test valu
100
indicated by the vertical dash line. You can see that all values for
100% Tumors Detected
Criterion
the tumor population lie below the value indicated by this vertical
100%
Sensitivity
line. Thus, for this criterion, the
sensitivity
is 100% in that all
75
tumors would be detected (i.e.
because
all tumors
had values that we
all tumors
detected)
75% False-Positives
less than the criterion value. However, selecting this criterion
would correctly identify only 25% of the non-tumor population. Th
50
specificity would be only 25%. That is 75% of the non-tumor
population also have values that fall below the selected criterion
25%
Specificity
and would have a positive (false
positive)
result.
25
(i.e. 75% non-tumors
misdiagnosed)
0
Criterion
ABR Test Measure
HOUSE EAR INSTITUTE
III. Stacked ABR: Results/Sensitivity vs. Specificity
Manny Don:
Percentile
If you wanted to select another criterion that would provide better
specificity,for example 90%, Tumor
the criterion value
required is shown
Non-Tumor
here by the dashed vertical line. With this criterion, 90% of the nontumor cases would be correctly identified. However, we can see that
for this criterion value, only 75% of the tumor case would be
100That is, the sensitivity is now only 75%. It is important to
detected.
remember that for any given ABR test measure, the distributions are
Criterion
defined. Selecting
any criterion
value determines the sensitivity
and
75% Tumors
Detected
specificity.
90%notSpecificity
75 By selecting a different criterion value, one does
change the separation of the distributions. Therefore, (i.e.
if one90% non-tumors
changes the criterion to improve sensitivity, the specificity
will be identified)
correctly
compromised;
if one changes the criterion to improve specificity,
50
then the sensitivity will be worse.
25
10
0
10% False-Positives
75% Sensitivity
(i.e. 25% tumors missed)
Criterion
ABR Test Measure
HOUSE EAR INSTITUTE
III. Stacked ABR: Results/Sensitivity vs. Specificity
Take Home Message

For any ABR test, the relationship between sensitivity and
specificity is determined by how well the measure separates the
tumor and non-tumor populations.

Therefore, for any ABR test:
(1) Changing the diagnostic criterion to increase sensitivity will
decrease specificity. In other words, changing the criterion to
detect more small tumors will increase the number of
misdiagnosed non-tumor patients (i.e. increase the falsepositive results).
(2) Changing the diagnostic criterion to increase specificity will
decrease sensitivity. In other words, changing the criterion to
increase the number of correctly identified non-tumor patients
(i.e. to decrease the number of false-positives) will increase the
number of small tumors that are missed.
HOUSE EAR INSTITUTE
III. Stacked ABR: Results/Sensitivity vs. Specificity
Manny Don:
The only way to increase both sensitivity and specificity
“The…better”. Here we show a new ABR test
at the
time
is population
to use an
ABR
measure
thatsame
separates
tumor
from
the measure that
non-tumor
population
than the example
separates
thebetter
non-tumor
and tumor populations better.
shown earlier. Now, the criterion that yields 100%
sensitivity, i.e. detection of all tumors, also has a
specificity of 80%, i.e. correctly identifies 80% of the
non-tumor population. Tumor
Non-Tumor
Percentile
100
Criterion
100% Sensitivity
(i.e. all tumors detected)
100% Tumors Detected
75
50
25
20
80% Specificity
(i.e. 80% non-tumors
correctly identified)
20% False-Positives
0
Criterion
NEW ABR Test Measure
HOUSE EAR INSTITUTE
III. Stacked ABR: Results/Sensitivity vs. Specificity
Manny Don:
Standard ABR vs. Stacked ABR
We can view the poor performance of the standard ABR measures
as a problem of highly overlapped distributions of these measures
for small tumor patients and non-tumor patients. We now want to
demonstrate how the distributions
of the Stacked
ABR measures
Standard
ABR Measure
for these populations are more
separated
sensitivity
(e.g.,
IT5 allowing
or I-V high
Delay)
and relatively high specificity.
Small
Tumor
Non-Tumor
Criterion
Stacked ABR Measure
(Wave V Amplitude)
Small
Tumor
Non-Tumor
Criterion
HOUSE EAR INSTITUTE
III. Stacked ABR: Results/Sensitivity vs. Specificity
Manny Don:
I will review our results with 47 tumor cases. The
population criteria were tumors not detected by the
Acoustic
Subjects
(N of
= 47)
standard ABR measures
(IT5 Tumor
and I-V delay)
irrespective
tumor size or that the tumor was equal to or less than 1
 Small (< 1 cm) tumors, irrespective of standard ABR results
cm irrespective
of standard ABR results. This table shows
Tumors not
detected
by standard
ABR measures,
irrespective
the 
distribution
of the
tumor patients
with respect
to
gender,ofear,
agesize
of patient.
theand
tumor
SUBJECTS
Males
14
Females
7
Total
21
Left Ear
15
11
26
Individuals
29
18
47
Right Ear
Age Range
32-68 years
Mean Age
48 years
25-66 years
49 years
III. Stacked ABR: Studies/Preliminary Work
HOUSE EAR INSTITUTE
Manny Don:
The control reference group was composed of 73 non-tumor
normal-hearing subjects. These included patients with negative
MRI results Non-tumor
and other normal-hearing
subjects recruited
from (N = 73)
Normal-hearing
Subjects
employees of the House Ear Institute and House Ear Clinic and
 House
Ear and
Clinic
patients
MRI
results
their family
members
friends.
Thiswith
tablenegative
shows the
distribution
for the 
non-tumor
population
with
respect
gender, ear, and
age ofand friends
House Ear
Institute
and
Clinictoemployees,
families,
patient.
SUBJECTS
Males
13
Females
27
Total
40
Left Ear
17
16
33
Individuals
30
43
73
Right Ear
Age Range
18-37 years
18-51 years
Mean Age
26 years
30 years
III. Stacked ABR: Studies/Preliminary Work
HOUSE EAR INSTITUTE
Manny Don:
In this slide we show the cumulative distribution curve for the
Pure
TonewithAverages
47 small
tumor cases
respect to the for
Pure47
ToneSmall
Average Tumor
(PTA) of the standard audiometric frequencies from 500 Hz to
8 kHz. Note that nearly
45% of this tumor population had
100
PTA values 20 dB or less and about 33% had PTAs 15 dB or
90
lower. In essence, many of the small tumor cases have fairly
80
good audiometric thresholds.
Cases
Percent
70
60
50
40
30
20
10
0
-10
0
10
20
30
PTA
40
50
60
III. Stacked ABR: Studies/Preliminary Work
HOUSE EAR INSTITUTE
Manny Don:
IT5
Of these 47 tumor2.6
cases, we were able to obtain the IT5 values in 35 cases. Only 54%
2.4
N = 35
(19/35) of the tumors
by the IT5 test. It can be seen that several of the
2.2 were detected
 Detected
(19)wave V latencies (negative IT5s) in the tumor
tumor cases missed
had shorter
2 actually
Abnormal

Missed
(16)
ear. To change the
criterion
to
detect
most
of
the
tumors would result in a negative IT5
1.8
value and, therefore,
1.6 would include most of the non-tumor cases, (i.e. very poor
specificity). In this1.4
slide, we also show the relationship between the PTA and the IT5
value. Using the 1.2
standard 0.2 ms IT5 as the criterion for a positive tumor test, note that
when a tumor was 1
missed by the IT5 (filled red circles), the PTA was generally less than
.8 the cases where the PTA was greater than about 32 dB, the IT5
32 dB. Moreover, all
.6
detected the tumor. So, it appears that patients with small tumors that are missed by the
.4
standard ABR, have
.2 no more than a mild hearing loss. This observation will be of
importance to our future
work with then Stacked ABR.
0
-.2
-.4
-.6
Normal
-10
0
10
20
30
40
50
60
PTA
III. Stacked ABR: Studies/Preliminary Work
HOUSE EAR INSTITUTE
Manny Don:
5.6
5.4
5.2
5.0
4.8
4.6
4.4
4.2
4.0
3.8
3.6
3.4
3.2
3.0
(a)
N = 16

Missed (12)
I-V delay in ms
I-V delay in ms
With regard to the other standard ABR measure, the I-V delay, reliable
recordings were obtained in 25 of the I-V
smallDelay
tumor cases (16 males and 9
females). Using 2 standard deviations from the mean of non-tumor normal
hearing individuals, only about 25% of the tumor cases were detected.
Tumorsresult
- Females
- Males
Again, to change theTumors
criterion
to detect most of the tumors would
in a
N=9
N = 16
very short I-V delay value and, therefore, would include most of the non6
6
T umors
-Females
umors
Males
tumor cases,
(i.e. very poorTor
no
specificity).
5.8
5.8
 Detected (4)
 Detected (2)
(b)
5.6
5.4
5.2
5.0
4.8
4.6
4.4
4.2
4.0
3.8
3.6
3.4
3.2
3.0
Observations
N=9

Missed (7)
Observations
III. Stacked ABR: Studies/Preliminary Work
HOUSE EAR INSTITUTE
Manny Don:
Percentile
Recall that to achieve high sensitivity
and goodNon-Tumor
specificity, we need
Tumor
a new ABR measure that 100
better separates the small tumor and
non100%
Sensitivity
(i.e.
all
tumors
100%
Tumors
tumor populations. We believe
that
theDetected
Stacked ABR measure detected)
75 much better to allow a criterion value
separates these populations
that results in high sensitivity and good specificity for small tumors.
50
80% Specificity
25
20
20% False-Positives
0
(i.e. 80% non-tumors
correctly identified)
Criterion
NEW ABR Test Measure
(i.e. all tumors
detected)
Non-Tumor
100
0
75
25
50
50
25
75
80
0
Criterion
Specificity Percentile
100% Sensitivity
Sensitivity Percentile
Tumor
80% Specificity
(i.e. 80% non-tumors
correctly identified)
100
Stacked ABR Amplitude
HOUSE EAR INSTITUTE
III. Stacked ABR: Studies/Preliminary Work
Manny Don:
30
60
40
50
50
Target = 50%
Specificity
40
60
30
70
20
80
10
90
Specificity Percentile
Sensitivity Percentile
 Tumors
In this figure are plotted the cumulative distributions
of the Stacked ABR values for
 Normal-Hearing
Males
the subjects. The open light blue circles are for
the non-tumor normal-hearing
Non-Tumor
subjects (N=30) and the filled orange circles represent the small tumor cases
(N=29). If we choose a criterion Stacked ABR value that yields 95% detection (left y0
200
400
600
800 1000 1200 1400 1600
axis) or sensitivity, this value results in a specificity of 93% (right y-axis). If we
100
0
Target
= 95%
change the criterion to
detect
all Sensitivity
of the tumors in100%
this population, the specificity
would
90 This would suggest that we could
Sens screen and catch10all the tumors
still be over 50%.
and still reduce a80significant number of non-tumor patients sent for imaging
using this
20
Stacked ABR measure.
70
93% Specificity
0
100
0
200
400
600
800
1000
1200 1400
1600
Stacked ABR nV
HOUSE EAR INSTITUTE
III. Stacked ABR: Studies/Preliminary Work
Manny Don:
 Tumors
In this figure, we plot the cumulative distribution curves of the Stacked ABR
Females  Normal-Hearing
for the female subjects (small tumor N=18 and non-tumor = 43). The Stacked
Non-Tumor
ABR value that results in 95% sensitivity (left y-axis) or detection, yields a
specificity of 78% (right y-axis). Again, the criterion value for 100% detection
yields a specificity of0 over200
50%.400 600 800 1000 1200 1400 1600
0
Target = 95% Sensitivity
100%
Sens
Sensitivity Percentile
90
10
80
20
70
30
60
40
50
Target = 50%
Specificity
40
30
50
60
70
20
78% Specificity
80
10
90
0
100
0
200
400
600
800
1000
1200
Specificity Percentile
100
1400 1600
Stacked ABR nV
HOUSE EAR INSTITUTE
III. Stacked ABR: Studies/Preliminary Work
Manny Don:
Self-explanatory
Take Home Message
The Stacked ABR appears to have better
sensitivity and specificity than the Standard
ABR for small (<1 cm) tumors.
In other words, the Stacked ABR is better at:
1. detecting small tumors, and
2. decreasing the number of misdiagnosed
non-tumor patients (i.e. decreasing the
number of false-positives referred for MRI).
III. Stacked ABR: Studies/Preliminary Work
HOUSE EAR INSTITUTE
Manny Don:
I want to briefly outline a simple screening
protocol for acoustic tumors.
Talk Outline
I. Background: Evolution of the Stacked ABR
II. Stacked ABR: Rationale and Method
III. Stacked ABR: Studies
IV. Tumor Screening Protocol
V. Future Stacked ABR Work
Talk Outline
HOUSE EAR INSTITUTE
Manny Don:
Recall that the simple standard ABR
measure
detect nearly all
ABR
SCREENING
Perform Standard
medium and large
size acoustic tumors
and close FOR
to half the small
ABR Analyses
PROTOCOL
(IT5,an
I-V,efficient
etc.)
tumors. Thus, for
screen,
the following
protocol could be
ACOUSTIC
TUMORS
performed.
1) Measure the standard IT5 and or I-V delay. If either of these
No
For An
values are abnormal, the patient
is Send
referred
for imaging.
There No
is no
Tumor?
Normal?
MRI
need to perform the Stacked ABR.
2) If these standard Yes
values are abnormal, and suspicion isYes
still
aroused, the Stacked ABR should be performed. If it is abnormally
low, then the patient is referred for imaging.
Perform Stacked
ABRABR
Analyses
Stacked
is normal,
3) If the
and followed.
Normal?
No
Yes
then the patient should be
Send For An
MRI
Observe?
Follow?
Tumor?
Evaluate for
auditory
neuropathy
and/or
refer for
neurological
evaluation
Prescribe
observed Treatment
(e.g., surgery)
Yes
No
HOUSE EAR INSTITUTE
IV. Stacked ABR: Screening Protocol
Talk Outline
I. Background: Evolution of the Stacked ABR
II. Stacked ABR: Rationale and Method
III. Stacked ABR: Studies
IV. Tumor Screening Protocol
V. Future Stacked ABR Work
Talk Outline
HOUSE EAR INSTITUTE
Manny Don:
Future Stacked ABR Work
In our NIH funded work we are investigating the following
optimizations and refinements to the Stacked ABR method:
1)We believe that interaural comparisons, when obtainable, will

improve
the sensitivity of the Stacked ABR.
Optimization and Refinement of the Stacked ABR
2) We
that specificity
might be improved by taking into
1.believe
Interaural
comparisons
consideration the effect of hearing loss.
2.alsoHearing
compensation
3) We
believe thatloss
using faster
click rates will increase the
separation between the small tumor from non-tumor populations.
3. Faster stimulus rates

Development of an automated clinical prototype
V. Future Work
HOUSE EAR INSTITUTE
Development of an Automated Clinical Prototype
 Bio-logic Systems Corp.: Licensed by the House Ear
Institute to collaborate on development of an automated
version of the Stacked ABR.
 Multi-Center Study: For validation of the Stacked ABR
method and evaluation of the automated prototype.
(NIH SBIR Phase II Awarded to Bio-logic Systems Corp.)
V. Current Developments and Future Stacked ABR Work
HOUSE EAR INSTITUTE
Staff Acknowledgements
Department of Electrophysiology
Betty Kwong, M.S., CCC-A
Chiemi Tanaka, M.A., CCC-A
Michael Waring, Ph.D.
Department of Clinical Studies
Ann Masuda, M.S., CCC-A
Department of Histopathology
Fred Linthicum, M.D.
Physicians at the House Ear Clinic
Support
NIH/NIDCD 1R43 DC04141 Raviv (PI)
NIH/NIDCD 2R44 DC04141 Raviv (PI)
NIH/NIDCD R01 DC03592 Don (PI)
HOUSE EAR INSTITUTE