View
214
Download
0
Category
Tags:
Preview:
Citation preview
Agenda:
- Axes check- Validation vs Pathology- Paper Publications- Congress presentations & Meetings
VII PMT meeting – Feb 29, 2012
Project method
1. Delphi is required to take all decisions based on experience + evidence
2. Empirical data on SUs taken on ACPC is fed to Delphi (external & internal landmarks)
3. Delphi decides on which SUs make up harmo hippo (internal landmarks)
4. Delphi decides that best segmentation is on hippo axis (10-5-1 in favour-neutral-against)
5. Empirical evidence collected that total hippo volume on ACPC = hippo axis
1. Delphi is required to take all decisions based on experience + evidence
2. Empirical data on SUs taken on ACPC is fed to Delphi (external & internal landmarks)
3. Delphi decides on which SUs make up harmo hippo (internal landmarks)
Internal Landmarks (n) External Landmarks (n)
SUs:
Minimum Hippocampus 5 1
Alveus/Fimbria 5 1
Subiculum - Oblique Line 5 1
Subiculum – Horizontal Line 5 1
Subiculum – Morphology 5 1
Tail - Crura 4 2
Tail – Tail End 5 1
Harmonized Hippo 6 0
2. Empirical data on SUs taken on ACPC is fed to Delphi (external & internal landmarks)
3. Delphi decides on which SUs make up harmo hippo (internal landmarks)
Hippo AC-PC same
Round 1: Which plane
Fisher’s exact
Round 2: Agree on hippo plane
Disagree same Agree
Rounds 3 and 4 converged on
mean angle between R and L
axes
Fisher’s
Binomial
4. Delphi decides that best segmentation is on hippo axis (10-5-1 in favour-neutral-against)
DATA SHOWN IN
PARIS, JULY 2011
Delphi: www.hippocampal-protocol.net/SOPs/delphipanel.html , access with user: prova , pw: prova
5. Empirical evidence collected that
total hippo volume on ACPC = hippo axis
Tracer Hippo_Left Hippo_Right
Intra-rater: same 10 subjects (20 images).different orientation (ACPC vs Hippocampal Axes)
MB 0.989 0.993
GP 0.985 0.991
LA* 0.971 0.967
Inter-rater: all 20 images (ACPC+Hippocampal Axes)
MB-GP-LA* 0.944 0.961
Inter-rater: only for ACPC orientation (10 images)
MB-GP-LA* 0.959 0.961
Inter-rater: only for Hippocampal Axes orientation (10 images)
MB-GP-LA 0.954 0.970
LA*= LA traced only 5 images oriented along ACPC line
Benchmark Harmonized hippos:1.5T ADNI scans2 x each of the 5
Scheltens’s atrophy score x 2 sides(SAME on 3T ADNI scans)
(total for each rater: 40 hippos)
5 expert tracers
20 naive tracers
The best 5 naive tracers
Harmonized Protocol:1.5T ADNI scans
2 sides x 5 Scheltens’s atrophy scores x 3 time points (0-12°month-24°month)
x 3 scanners + retracing for timepoint 1(SAME on 3T ADNI scans)
(total for each rater: 240 hippos – including40 hippos already traced)
GOLD STANDARD
Local Protocol:1.5T 3D T1-weighted scans from (Bobinski et al., 2000)pathologically verified set(total for rater: 30 hippos)
1 tracer
Qualification(20 tracers)
Qualification
global and local 95% confidence intervals
RM-ANOVA: test of rater and rater by center terms
RM-ANOVA: test of main effects side, trace-retrace, atrophy, time, scanner, rater
Local Protocol:Experimental set (1.5T ADNI):2 x each of the 5 Scheltens’s
atrophy score x 2 sides (SAME on 3T ADNI scans)
(total for each rater: 40 hippos)
Harmonized Protocol:Experimental set (1.5T ADNI): 2 x each of the 5 Scheltens’s
atrophy score x 2 sides(SAME on 3T ADNI scans)
(total for each rater: 40 hippos)
Harmonized Protocol:1.5T 3D T1-weighted scans from (Bobinski et al., 2000)pathologically verified set(total for rater: 30 hippos)
RM-ANOVA: test of protocol main effect
VARIABILITY EVALUATION
VALIDATION vsPATHOLOGY
Tra
inin
g(t
raci
ng 2
0 hi
ppos
on
1.5T
AD
NI
scan
s w
ith e
ach
SU)
(SA
ME
on
3T A
DN
I sc
ans)
Del
phi p
anel
→ h
arm
oniz
ed p
rot
VALIDATION ON 1800 HIPPOCAMPI
IN TOTAL
Pros & Cons
Methodological
1) Violate general method of the project (decisional tree)
2) Interfere with the aim of the project (define gold standard as best possible accuracy)
3) Disregard and contradict the opinion of 20 international experts of hippocampus
Methodological
1) Contrast with the orientation on which evidence was gathered
Note that this contrast :did not constitute a problem for 20
international hippo expertsdid not correspond to quantitative
evidence showing any impact on the Harmo Prot
Accepting the consensus on hippo axes would:
Practical
Reduce partial volume effects and facilitate manual tracing (tracers and panelists’ subjective reports)
Changing to AC-PC would:
Practical
Facilitate automatic preprocessing
Validation versus pathology
Local Protocol:1.5T 3D T1-weighted scans from (Bobinski et al., 2000)pathologically verified set(total for rater: 30 hippos)
1 tracer
Qualification on benchmark images(total for each rater: 40 hippos)
Harmonized Protocol:1.5T 3D T1-weighted scans from (Bobinski et al., 2000)pathologically verified set(total for rater: 30 hippos)
RM-ANOVA: test of protocol main effect
WHICH SAMPLE?
Paper OperationalizationAlzheimer’s and Dementia
Papers describing the project
Survey of protocols (preliminary phase; published, JAD 2011)
Operationalization (preliminary phase; to be completed)
Delphi consensus (Brescia Team)
Master tracers’ practice and reliability (Brescia Team)
Development of certification platform (Duchesne and coll?)
Validation data and Protocol definition (Brescia Team)
Validation vs pathology (TBD)
Benchmark Harmonized hippos:1.5T ADNI scans2 x each of the 5
Scheltens’s atrophy score x 2 sides(SAME on 3T ADNI scans)
(total for each rater: 40 hippos)
5 expert tracers
20 naive tracers
The best 5 naive tracers
Harmonized Protocol:1.5T ADNI scans
2 sides x 5 Scheltens’s atrophy scores x 3 time points (0-12°month-24°month)
x 3 scanners + retracing for timepoint 1(SAME on 3T ADNI scans)
(total for each rater: 240 hippos – including40 hippos already traced)
GOLD STANDARD
Local Protocol:1.5T 3D T1-weighted scans from (Bobinski et al., 2000)pathologically verified set(total for rater: 30 hippos)
1 tracer
Qualification(20 tracers)
Qualification
global and local 95% confidence intervals
RM-ANOVA: test of rater and rater by center terms
RM-ANOVA: test of main effects side, trace-retrace, atrophy, time, scanner, rater
Local Protocol:Experimental set (1.5T ADNI):2 x each of the 5 Scheltens’s
atrophy score x 2 sides (SAME on 3T ADNI scans)
(total for each rater: 40 hippos)
Harmonized Protocol:Experimental set (1.5T ADNI): 2 x each of the 5 Scheltens’s
atrophy score x 2 sides(SAME on 3T ADNI scans)
(total for each rater: 40 hippos)
Harmonized Protocol:1.5T 3D T1-weighted scans from (Bobinski et al., 2000)pathologically verified set(total for rater: 30 hippos)
RM-ANOVA: test of protocol main effect
VARIABILITY EVALUATION
VALIDATION vsPATHOLOGY
Tra
inin
g(t
raci
ng 2
0 hi
ppos
on
1.5T
AD
NI
scan
s w
ith e
ach
SU)
(SA
ME
on
3T A
DN
I sc
ans)
Del
phi p
anel
→ h
arm
oniz
ed p
rot
Congress presentations
- AAN 2012, New OrleansOral Presentation S04.003, April 24, Aging and Dementia: Therapeutic Interventions, 1.30 pmM Boccardi, M Bocchetta, L Apostolova, J Barnes, G Bartzokis, G Corbetta, C DeCarli, L DeToledo-Morrell, M Firbank, R Ganzola, L Gerritsen, W Henneman, R Killiany, N Malykhin, P Pasqualetti, J Pruessner, A Redolfi, N Robitaille, H Soininen, D Tolomeo, L Wang, C Watson, H Wolf, S Duchesne, CR Jack, GB Frisoni.Delphi consensus on landmarks for the manual segmentation of the hippocampus on MRI: preliminary results from the EADC-ADNI Harmonized Protocol working group.
- AAIC 2012, Vancouver
Abstract n. 26291, submitted to AAIC 2012M Boccardi, M Bocchetta, A Redolfi, P Pasqualetti, R Ganzola, N Robitaille, S Duchesne, CR Jack, GB Frisoni and the EADC-ADNI Hippocampal Harmonization Group.Definition of Harmonized Protocol for Hippocampal Segmentation
Abstract n. 26292, submitted to AIC 2012M Boccardi, M Bocchetta, A Redolfi, P Pasqualetti, R Ganzola, N Robitaille, S Duchesne, CR Jack, GB Frisoni and the EADC-ADNI Hippocampal Harmonization Group.Definition of Harmonized Protocol for Hippocampal Segmentation
AAN – April 24, 2012AGENDA: 1. Delphi Paper; 2. Publication policy
AAIC – July 2012AGENDA: 1. Presentation of Harmonized Protocol; 2. Reliability figures
SOPs periodical meetings
GANTT
Recommended