16
SPATIO-TEMPORAL ANALYSIS OF THE SIGNIFICANT CHANGES IN CARTILAGE MORPHOLOGY: DATA FROM THE OSTEOARTHRITIS INITIATIVE Jose Tamez-Pena 1 , Patricia Gonzalez 2 , Edward Schreyer 2 , Saara Totterman 2 , 1 Biomedicine, Tec de Monterrey, Monterrey, Nuevo Leon, Mexico; 2 Qmetrics Technologies, Rochester, NY, USA

Jose Tamez-Pena 1 , Patricia Gonzalez 2 , Edward Schreyer 2 , Saara Totterman 2 ,

  • Upload
    ion

  • View
    26

  • Download
    0

Embed Size (px)

DESCRIPTION

SPATIO-TEMPORAL ANALYSIS OF THE SIGNIFICANT CHANGES IN CARTILAGE MORPHOLOGY: DATA FROM THE OSTEOARTHRITIS INITIATIVE. Jose Tamez-Pena 1 , Patricia Gonzalez 2 , Edward Schreyer 2 , Saara Totterman 2 , - PowerPoint PPT Presentation

Citation preview

Page 1: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

SPATIO-TEMPORAL ANALYSIS OF THE SIGNIFICANT CHANGES IN

CARTILAGE MORPHOLOGY: DATA FROM THE OSTEOARTHRITIS

INITIATIVEJose Tamez-Pena1, Patricia Gonzalez2, Edward Schreyer2, Saara

Totterman2, 1Biomedicine, Tec de Monterrey, Monterrey, Nuevo Leon, Mexico;

2Qmetrics Technologies, Rochester, NY, USA

Page 2: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Objective

• Visualize, Follow and Quantitate the Areas of Cartilage Loss in an OA population

Page 3: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

IntroductionProblem:

• Cartilage Thickness Changes are focal, spatially heterogeneous and bi-directional: Thinning and thickening

• Cohort studies look at population averages• The average of this heterogeneous data

has a very small responsivenessSolution:

• For a subject: Localize and isolate the changes

• For a cohort: Count map of the significant changes in cartilage thickness

Page 4: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Material & Methods

• Osteoarthritis Initiative (OAI) 3D DESS data sets: – Releases 0.C.2, 1.C.2 and 3.C.1 from Progression

cohort.• Three time points: Baseline, 12 month and 24 month.• 138 subjects with 3 time points

– Nonexposed Data Release 0.E.1, 1.E.1 and 3.E.1 (n=108)• Three time points: Baseline, 12 month and 24

month.

– OAI Pilot Scan-Rescan Longitudinal Data for the estimation of scan-rescan variability (n=24)

Page 5: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Multi-Atlas-Based Segmentation• Generate Atlas• Register and Segment Each MRI to the

Atlas (ITK registration modules)• Postprocess the segmentation to

match underlying MRI information.• Visually score the quality of the

segmentation.• Use the registration data to map each

segmentation to the atlas space• Subtract each mapped segmentation

to compute change in cartilage thickness

• Compute Significant Changes• Compute Cohort Averages

Page 6: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Quantitation

Page 7: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Standardized Analysis: Changes in Cartilage Thickness

-0.5 +0.5 -0.5 +0.5

Medial Lateral

Page 8: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Change Measurement: Significant Change Maps

Minus

Change Map

Scan-Rescan SDD MapSignificant Change Map

(Activation Map)

-0.5 +0.5

0.0 +0.5

-0.5 +0.5

-0.5 +0.5

Baseline24 Month

-0.75 +0.75-0.5 +0.5

=

The Scan-RescanStandard Deviation of the Differences (SDD) is used to mark changes in thickness values that higher than the scan-rescan paired errors ( Delta < -1.96*SDD )

Page 9: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Population Maps

Average ReferencedThickness

12 Month Change Map

24 Month Change Maps

Average Change Map

Significant ChangesPrevalence Map

-0.5 +0.5 -0.5 +0.5

0.0 0.3 0.0 0.3

-0.5 +0.5

+ =

Page 10: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

24 Month Results

-0.25

+0.25

0.0

0.3

-0.5

+0.5

Nonexposed No Denuded Low Denuded High Denuded

n=103 n=51 n=43 n=43 20.6%

9.3%8.1% 11.6%

17.3% 14.7%

3.4%

6.8%

Baseline

24 MonthChange

24 MonthHeat Map

-0.5

+0.5

Page 11: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Nonexposed No-Denuded Low Denuded High Denuded

12M-BL 3.6% 5.4% 7.1% 9.1%

24M-BL 3.4% 7.8% 8.6% 11.0%

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

Rela

tive

Area

Average Area of Femur Affected by Significant Changes

SRM=0.73P<0.001 P=0.007 P=0.054SRM=0.36

SRM=0.36

SRM=0.39

Nonexposed No-Denuded Low Denuded High Denuded

12M-BL 4.3% 5.9% 7.50% 8.20%

24M-BL 4.6% 7.9% 8.80% 9.50%

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%Re

lativ

e Ar

ea

Average Area of Tibia Affected by Significant Changes

P<0.001

SRM=0.31

Page 12: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Nonexposed No Denuded Low Denuded High Denuded

12M-BL 22.3% 43.1% 76.7% 88.8%

24M-BL 19.4% 58.9% 76.7% 93.0%

0.0%10.0%20.0%30.0%40.0%50.0%60.0%70.0%80.0%90.0%

100.0%

Prop

ortio

n of

Sub

ject

s

Proportion of Subjects Whose Significant Changes were Greater Than

5% of the Femur Area

P=0.009 P=0.001

No Exposed No Denuded Low Denuded High Denuded

12M-BL 35.0% 51.0% 72.1% 72.1%

24M-BL 37.9% 56.9% 74.4% 67.4%

0.0%10.0%20.0%30.0%40.0%50.0%60.0%70.0%80.0%90.0%

100.0%Pr

opor

tion

of s

ubje

cts

Proportion of Subjects Whose Significant Changes were Greater Than

5% of the Tibia Area

P=0.038

Fisher’s Exact Test

Page 13: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Limitations

• Small OA population• Multi-atlas based segmentation is biased

towards atlas models– Less accurate at advanced OA cases– Higher noise at advanced OA cases

Page 14: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Conclusion

• The automated analysis methodology enabled the localization and mapping of the significant changes in cartilage Thickness.

• The significant changes are heterogeneous– Non Exposed OAI cohort did not change– OA subjects with no denuded areas had 2.4% of new

areas of cartilage loss every year. SRM=0.73

• The methodology indicated that not all subjects are affected by loss, and that the prevalence of loss is greater at more advanced OA groups.

Page 15: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

Acknowledgements

• The OAI for all the imaging and clinical data

Page 16: Jose Tamez-Pena 1 , Patricia  Gonzalez 2 ,  Edward Schreyer 2 ,  Saara  Totterman 2 ,

-0.5 +0.5 -0.5 +0.5

Baseline 24 Month Change Map

Change Measurement: Significant Change Maps

-0.75 +0.75