Transcript Slide 1

ARTIFICIAL INTELLIGENCE
FOR AUTOMATED FITTING OF COCHLEAR IMPLANTS
Paul J Govaerts, MSc, MD, PhD
B Vaerenberg , G De Ceulaer, W Kowalczyk, J Diez, I Bermejo
The Eargroup (Antwerp, Belgium)
&
Universities of Antwerp (BE), Leiden (NL), UNED (ES)
CI FITTING

State of the art

Eargroup approach

Comfort based

Outcome based

No systematic
approach, no
universal GCP,
huge variability

Systematic
approach

Tailoring from
the start

Start with “One fits
all”, postpone
tailoring
FITTING FOR PERFORMANCE
-10
0
10
20
30
Measure
outcome
40
dB
50
60
70
80
100
90
100
80
110
phoneme score (%)
120
125
250
500
1000
2000
4000
8000
60
40
20
0
10
20
30
40
50
db SPL
60
70
80
90
Interpret
MAP &
outcome
Modify
MAP
OUTCOME BASED
Identification
Discrimination
Detection
intensity
Intensity
spectral
Spectral
content
content
S
temporal
Temporal
content
content
FITTING FOR PERFORMANCE
N > 60
Measure
outcome
N > 150
Interpret
MAP &
outcome
Modify
MAP
audiometry
speech audiometry
Fitting to
Outcome
eXpert
Govaerts, et al. Otol Neurotol 2010; 31(6):908-18.
FOX 1.1
 SW
opens in the background
 3 active maps ready to be foxed ...
FOX 1.1

User interface
 Password
protected log-in
 User friendly patient-selection
 SW opens in the back
 3 active maps ready to be foxed ...
FOX 1.1

Typical procedure
 Open
FOX – Select MAP
 Perform 2 outcome measures (20 ‘)
 Request advice – Judge – Accept recommendations
 Put new map in processor
CASE

Case1: 4 months after switch-on
SWITCH-ON
400
350
300
P75
250
P50
P25
200
150
100
50
0
E1
E2
E3
E4
E5
E6
E7
E8
E9
E10
E11
E12
E13
E14
E15
E16
FITTING SCHEME
0
24
1
2
3
4
5
6
7
8
9
10
11
12
18
AutoMaps
Switch on
Silver 1
Silver 2
Silver 3
Gold 1
Gold 2
Gold 3
Ivory 1
Ivory 2
Ivory 3
FITTING SCHEME
0
1
2
3
4
5
6
7
8
9
10 11 12
Source MAP
18
24
AutoMaps
Map
modifications
Switch on
Gold 3#1
Silver 1
Silver 2 Gold 3#2, …
Silver 3
Gold 1
Gold 2
Gold 3
Ivory 1
Ivory 2
Ivory 3
- Audiogram
- A§E phoneme discrimination
- A§E Loudness Scaling
-Speech Audiogram
FITTING SCHEME
0
1
2
3
4
5
6
7
3 hours
8
9
10 11 12
18
24
2,0
PRELIMINARY RESULTS
1
2
3
4
5
6
7
8
9
1,0
0,5
10 11 12
RMS
0
1,5
18
24
0,0
-0,5
-1,0
3 hours

Switch on: N=8, Fox
-1,5
-2,0
250 Hz
1.1(EG0910)
1000 Hz
4000 Hz
100
 3 months postop (2,5 hours)
90
80
phoneme score (%)
70
60
50
40
30
20
10
0

Ongoing trial Europe, India
10
20
30
40
50
60
70
80
90
SPL
Vaerenberg, et al. Int J dbAudiol
2011; 50:50-8.
FOX
EUROPEAN MULTICENTRIC STUDY
Andreas Büchner, Thomas Lenarz, MHH, Hannover, Germany
Rolf-Dieter Battmer, Romy Goetze, UKB, Berlin, Germany
Isabelle Mosinier, Stephanie Borel, Beaujon, Paris, France
Huw Cooper, Claire Fielden, University hospital, Birmingham, UK
Zebunissa Vanat, Joanne Muff, Adenbrookes, Cambridge, UK
Terry Nunn, Anzel Britz, Guy’s and St.Thomas’, London, UK
Filiep Vanpoucke, Advanced Bionics Europe
Dzemal Gazibegovic, Advanced Bionics Europe
Paul Govaerts, Eargroup, Antwerp, Belgium
PRELIMINARY RESULTS
~ 36000 outcome
points
Identification
in 275 CI users
in 15 CI centres
Discrimination
Detection
intensity
Intensity
spectral
Spectral
content
content
S
temporal
Temporal
content
content
PRELIMINARY RESULTS: AUDIOGRAM
Target = 30 dB (35 for 250 Hz)
 Tolerance = 40 dB

AUDIOGRAM
100
90
80
% on target
70
60
50
40
30
71
69
56
64
55
55
20
10
0
250
500
1000
2000
4000
8000
AUDIOGRAM
Audiogram
100
90%
90
83%
80
60%
70
60
50
40
% on target
30
Median Interval
between first and last test session (days)
20
74
10
0
AutoMAP
FOX (almost) on target
FOX on target
22
18
14
250
500
1000
30
27
2000
4000
8000
PRELIMINARY RESULTS
~ 36000 outcome
points
Identification
in 275 CI users
in 15 CI centres
Discrimination
Detection
intensity
Intensity
spectral
Spectral
content
content
S
temporal
Temporal
content
content
SPECTRAL DISCRIMINATION
100
PD:EaSI: r2 = 0,2602; y = -20,717 + 3,7695*x
90
80
70
EaSI
60
50
40
30
20
10
0
-10
0
2
4
6
8
10
12
PD
14
16
18
20
22
SPECTRAL DISCRIMINATION
Spectral Discrimination (A§E PD)
100
99%
97%
90
80
70
82%
60
50
40
% on target
30
20
10
0
AutoMAP
FOX (almost) on target
FOX on target
SPECTRAL DISCRIMINATION
100
90
80
% on target
70
60
50
40
97 99 98 94 99 99 98
79 95
95
71
89
97
95
84
96 95
80
67
74
30
20
10
0
a-r u-ʃ u-a u-i i-a o-a i-ε m-z s-ʃ ε-a u-o ə-a ə-o ə-ε ə-i z-s v-z ə-u u-y y-i
SPECTRAL DISCRIMINATION
100
90
80
% on target
70
60
50
40
97 99 98 94 99 99 98
79 95
95
71
89
97
95
84
96 95
80
67
74
30
20
10
0
a-r u-ʃ u-a u-i i-a o-a i-ε m-z s-ʃ ε-a u-o ə-a ə-o ə-ε ə-i z-s v-z ə-u u-y y-i
SPECTRAL DISCRIMINATION
400
Median Interval
373
between first and last
test session (days)
335
350
300
247
250
200
151
149
150
106
100
60
50
60
28
28
75
38
0
a-r u-ʃ u-a u-i
i-a o-a i-ε m-z s-ʃ ε-a u-o ə-a ə-o ə-ε ə-i z-s v-z ə-u u-y y-i
m-z
ɛ-a
v-z
y-i
ə-ɛ
z-s
PRELIMINARY RESULTS
~ 36000 outcome
points
Identification
in 275 CI users
in 15 CI centres
Discrimination
Detection
intensity
Intensity
spectral
Spectral
content
content
S
temporal
Temporal
content
content
LOUDNESS SCALING
Loudness Scaling
100
85%
90
72%
80
70
48%
60
50
40
% on target
30
20
10
0
AutoMAP
FOX (almost) on target
FOX on target
LOUDNESS SCALING
600
Median Interval
507
between first and last test session (days)
500
445
422
375
400
300
350
260
200
121
125
100
38
35
37
80
35
50
67
0
35
50
65
65
80
35
50
65
80
PRELIMINARY RESULTS
~ 36000 outcome
points
Identification
in 275 CI users
in 15 CI centres
Discrimination
Detection
intensity
Intensity
spectral
Spectral
content
content
S
temporal
Temporal
content
content
SPEECH AUDIOMETRY
100
90
80
% on target
70
60
50
73
40
61
30
20
16
10
0
40-55
55-70
70-85
SPEECH AUDIOMETRY
Speech Audiometry
100
90
80
85%
70
60
61%
50
40
% on target
30
20
10
0
AutoMAP
FOX (almost) on target
FOX on target
SPEECH AUDIOMETRY
600
524
500
Median Interval
between first and last test session (days)
400
292
300
216
200
100
0
40-55
55-70
70-85
OVERALL (41 OUTCOME POINTS)
All 41 outcome points
100
90%
90
86%
80
71%
70
60
50
40
% on target
30
20
10
0
AutoMAP
FOX (almost) on target
FOX on target
CONCLUSIONS

Measure performance
- Feasible in daily clinical practice (<10’ per test)
- Language independent
- Target = normal values
audiometry
speech audiometry
http://otoconsult.com

Artificial Intelligence




Assists the audiologist to navigate
Optimises results
Systematises procedure
Allows for quality control