a perimetric re-test algorithm that is significantly more accurate than current procedures

30
A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES Andrew Turpin School of Computer Science and Information Technology RMIT University, Melbourne Darko Jankovic Department of Optometry and Vision Science University of Melbourne Allison McKendrick Department of Optometry and Vision Science University of Melbourne

Upload: yoshi-fuller

Post on 03-Jan-2016

31 views

Category:

Documents


0 download

DESCRIPTION

A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES. Allison McKendrick Department of Optometry and Vision Science University of Melbourne. Andrew Turpin School of Computer Science and Information Technology RMIT University, Melbourne. Darko Jankovic - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE

ACCURATE THAN CURRENT PROCEDURES

Andrew TurpinSchool of Computer Science and Information TechnologyRMIT University, Melbourne

Darko JankovicDepartment of Optometry and

Vision ScienceUniversity of Melbourne

Allison McKendrick Department of Optometry

and Vision ScienceUniversity of Melbourne

Page 2: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

Possible re-test algorithms

1. Use individual presentation information…2. Use test Hill of Vision to bias re-test3. Continue the previous test with “next”

termination criteria– More reversals (staircase, MOBS)– Tighter PDF standard deviation (Bayesian) – Fixed number of presentaitons

4. Use test thresholds to seed re-test– Starting point for staircase (FT From Prior)– Initialisation of MOBS stacks – Centre a PDF around threshold (ZEST, SITA)

Page 3: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

Prior distribution (before first presentation)

Page 4: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 1 presentation

Page 5: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 2 presentations

Page 6: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 3 presentations

Page 7: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 4 presentations

Page 8: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 5 presentations

Gaussian with standard deviation 3dB

Page 9: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

1. Continued ZEST• Termination Criteria

- Fixed # presentations 4,5,6- Standard deviation 0.7, 0.8, 0.9, 1.0

• LF- Steep, steeper, steepest

2. Seeded ZEST• PDFs

- Gaussian standard deviation 2,3,4 dB- Step function, width 4,6,8,10 dB

• LF- Steep, steeper, steepest

• Termination criteria- Fixed # presentations 4,5,6- Standard deviation 0.7, 0.8, 0.9, 1.0

3. MOBS- Stack initialisation 2, 3, 4 dB- Termination criteria: 2, 3 reversals

2, 3, 4 width

486 procedure

s

Patient set False+

False-

Normal-1 0% 0%

Normal-2 3% 15%

Normal-3 15% 3%

Normal-4 20% 20%

Glaucoma-1 0% 0%

Glaucoma-2 3% 15%

Glaucoma-3 15% 3%

Glaucoma-4 20% 20%

8 Patientmodels

Computer Simulations

350 realpatients

Page 10: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

Performance: No Error, No Change

4 5 6 7Mean number of presentations

Mea

n ab

solu

te e

rror

(dB

)

2

1Z

F ull Threshold

estS eeded ZestC ontinued Zest

S ITA

Bengtsson et al, ACTA ‘97

Page 11: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

Performance: General Height -3dB

4 5 6 7

Mea

n ab

solu

te e

rror

(dB

)

2

1

C

S

Z

F ull Threshold

est

eeded Zest

ontinued ZestS ITA

Mean number of presentations

Page 12: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

ProblemsContinue• General Height change ignored, need many

presentations to get right answer if GH changes, and there is still a bias towards original test value

Seed• Could adjust seed if GH change known

– Estimate with “primary points” algorithm– Would be slower than Full Threshold (and SITA)

Katz et al, IOVS 1632

Page 13: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

Speeding up GH-corrected Seed

• Spend 2 or 3 presentations per location checking if threshold not less than last time (multi-sample supra-threshold)

• If so, then do no more for that location

• Otherwise, assume threshold decreased, and seed a ZEST accordingly

McKendrick & Turpin, OVS 2005

Page 14: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

29 28 28 30

30 31 31 31 32 29

27 29 33 31 30 33 30 30

26 33 35 34 32 29 28 25

27 32 33 35 33 32 31 26

27 30 32 31 33 31 30 29

29 30 27 28 27 28

28 27 26 24 26

1 2 3 4 5 6 7 8 9 10 11 12 13 14 …

35 35 34 33 33 33 33 33 33 32 32 32 32 32 …

Automated Static Perimetry, 1999, Anderson & Patella

Page 15: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

29 28 28 30

30 31 31 31 32 29

27 29 33 31 30 33 30 30

26 33 35 34 32 29 28 25

27 32 33 35 33 32 31 26

27 30 32 31 33 31 30 29

29 30 27 28 27 28

28 27 26 24 26

31

3031

31

General Height decrease of 2dBSupra-threshold decrement of 2dB

So multi-sample all locations at previous less 4dB

If see this 2 of 3 times, then just use previous threshold - 2dBelse do a full ZEST on the location

27 29

Test Re-Test

Page 16: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

29 28 28 30

30 31 31 31 32 29

27 29 33 31 30 33 30 30

26 33 35 34 32 29 28 25

27 32 33 35 33 32 31 26

27 30 32 31 33 31 30 29

29 30 27 28 27 28

28 27 26 24 26

31

3031

31

31 29

33

General Height decrease of 2dBSupra-threshold decrement of 2dB

So test all locations at previous less 4dB

If see this 2 of 3 times, then just use previous threshold - 2dBelse do a full ZEST on the location

Page 17: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

Performance with no error

4 5 6 7

Mea

n ab

solu

te e

rror

(dB

)

2

1 CS Z

F ull Threshold

esteeded Zestontinued Zest

N ew

S ITA

Mean number of presentations

Page 18: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

General height -3dB

4 5 6 7

Mea

n ab

solu

te e

rror

(dB

)

2

1

C

S

Z

F ull Threshold

est

eeded Zest

ontinued Zest

N ew

S ITA

Mean number of presentations

Page 19: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

Conclusions• Continuing previous procedure doesn’t work

• Seeding a ZEST with a Gaussian pdf about previous threshold works, but is slow

• Adding multi-sampling supra-threshold step gives speed and accuracy gains

• The resulting re-test procedure is as fast, but more accurate, than existing test algorithms BUT does not detect an isolated increase in threshold

Page 20: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

Hill of Vision Approach

• Alter eccentricity adjustments in growth pattern based on individual’s HoV

• Takes into account General Height change

• Very small gains, but not really worth the effort

Page 21: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

C C C C

B B B B B B

C B A B B A B C

C B B B B B C D

C B B B B B C D

C B A B B A B C

B B B B B B

C C C C C

Page 22: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

C C C C

B B B B B B

C B 31 B B 31 B C

C B B B B B C D

C B B B B B C D

C B 33 B B 29 B C

B B B B B B

C C C C C

Page 23: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

C C C C

26 27 30 B B B

C 30 31 32 B 31 B C

C 32 33 B B B C D

C B B B B B C D

C B 33 B B 29 B C

B B B B B B

C C C C C

Page 24: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

C C C C

26 27 30 B B B

C 30 31 32 B 31 B C

C 32 33 B B B C D

C B B B B B C D

C B 33 B B 29 B C

B B B B B B

C C C C C

Page 25: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

25 26 C C

26 27 30 B B B

26 30 31 32 B 31 B C

25 32 33 B B B C D

C B B B B B C D

C B 33 B B 29 B C

B B B B B B

C C C C C

Page 26: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

25 26 C C

26 27 30 B B B

26 30 31 32 B 31 B C

25 32 33 B B B C D

C B B B B B C D

C B 33 B B 29 B C

B B B B B B

C C C C C

Page 27: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 6 presentations

Page 28: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 7 presentations

Page 29: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 8 presentations

Page 30: A PERIMETRIC RE-TEST ALGORITHM THAT IS SIGNIFICANTLY MORE ACCURATE THAN CURRENT PROCEDURES

After 9 presentations