Transcript DtiStudio

DtiStudio
Susumu Mori
Johns Hopkins University
Overall direction of DTI research
Smaller b (<1,200), fewer directions (<30), 2mm, 5 min
DWIs
Tensor calculation
Larger b (> 3,000), more directions (>60), lower resolution, longer time
High Angular-Resolution Diffusion Imaging
Overall direction of DTI research
Smaller b (<1,200), fewer directions (<30), 2mm, 5 min
DWIs
Tensor calculation
Overall direction of DTI research
Smaller b (<1,200), fewer directions (<30), 2mm, 5 min
DWIs
•
•
•
•
•
Tensor calculation
Automated and quantitative quality control pipeline
Fully automated cloud-based web-based pipeline
Integration of image quantification schemes
Dynamic programming for fiber tracking
Automated fiber tracking
DtiStudio
How DTI studies could go wrong
Motion
Registration
Eddy-current distortion
Motion monitoring
Eddy current monitoring
Outlier pixels (dropout, ghost)
Slice rejection
Pixel rejection
Motion monitoring
Non-physiological motion
XMR-SNR-test-AH: dtRegVoltranslationy
XMR-SNR-test-AH: cumulateddtrotationmetric
3
4
Rep 1
Rep 2
Rep 3
Rep 4
Rep 5
3.5
2
Rep 1
Rep 2
Rep 3
Rep 4
Rep 5
DWI rotationmetric: degree
DWI tanslationy: mm
3
1
0
-1
-2
2.5
2
1.5
1
0.5
-3
20
40
60
80
100
# of DWI
120
140
160
0
20
40
60
80
100
# of DWI
120
140
160
Eddy current monitoring
Importance of tensor fitting quality
Images that CAN’T be registered
absolute fitting error before registration:
4
3.5

4.5
A
3
2.5
2
1.5
1
0.5
0
0
0.5
1
1.5
2
apparent translation: mm
2.5
3
histogram eddyzmetrics
0.02
4.5
4
absolute fitting error after registration:

C
3.5
B
3
2.5
2
1.5
1
0.5
0
0
0.5
1
1.5
2
apparent translation: mm
histogram eddyB0metrics
2.5
2.5
3
Outlier rejection
80
50
60
100
a
b
c
150
40
20
d
0
50
100
150
80
50
60
100
e
f
g
150
40
20
h
0
50
100
150
50
1.5
1
100
i
j
k
150
2
l
0.5
50
100
150
0
Population report Matlab code
Original image
Select subject to
show results
After registration
Artifact masked
MriStudio Pipeline
DtiStudio
DWIs
DiffeoMap
RoiEditor
254 structures
Example of multi-modal analysis
Automated pipeline
Read DICOM data
Tensor calculation
Scalar map calculation
QC report
Save images
Skull-strip
Linear registration
Save images
Atlas-based parcellation
Save tables
Database solution
Initiate image analysis pipeline
Specify the data for analysis
Results become available
Calculation starts
Fiber tracking: path-generation
Automated parcellation of the
brain
Fiber tracking: path-generation
Overall direction of DTI research
Smaller b (<1,200), fewer directions (<30), 2mm, 5 min
DWIs
•
•
•
•
•
Tensor calculation
Automated and quantitative quality control pipeline
Fully automated cloud-based web-based pipeline
Integration of image quantification schemes
Dynamic programming for fiber tracking
Automated fiber tracking
Acknowledgment
• Program development
– Hangyi Jiang
– Xin Li
• Server implementation
– Anthony Kolasny
– Can Ceritoglu
– Bill Schneider
• Quantification module
– Michael Miller
– Yue Li
– Xiaoying Tang
• Atlas generation
– Kenichi Oishi
– Anderia Faria