prof. rosemary a. renaut dr. hongbin guo & dr. haewon nam...

41
Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam Department of Mathematics and Statistics, Arizona State Univerisity Dr. Kewei Chen Banner Good Samaritan PET center, Phoenix Supported by: Arizona Alzheimer’s Research Center and NIH BIBIB April 2005 – p.1

Upload: others

Post on 02-Nov-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

� � � � �� � �� � � � � � � � � �� � ��� � ��� � � �

� � � � � ��� �� � � � � � �� �� � � � � ��� � � � � � � �

� � � � � � �� � � � ��

��� !#" $ %'& ! " ( ! ( ( & ! % ( ! %'& ! "

) & � ! � & * ! %,+ &

Prof. Rosemary A. Renaut

Dr. Hongbin Guo & Dr. Haewon Nam

Department of Mathematics and Statistics, Arizona State Univerisity

Dr. Kewei Chen

Banner Good Samaritan PET center, Phoenix

Supported by: Arizona Alzheimer’s Research Center and NIH BIBIB

April 2005– p.1

Page 2: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

OUTLINE

��� � ��� � � �� � � � � � � � � � ��� � � � � � � � ���

� � � � � � � �� � � � � � � � �� � � �� � � � ��� ���

! � � � � � � � �� � � � � � � � � � � �#" � � � � � �

$ � � � � � � � � " % � � � & � � � � � � � � � � �" � �#" � � � � � �•

' �� � � ��

� � � � � � � �� � � � �� % ( � � " � ' �" � � � � � � ��� �

( � � �� � � � � � � $� � � � � � � � � � ��) � � �

$� � � � � � ��� � � �+* � � � � ��� � �

– p.2

Page 3: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

1. Ultimate Goal

��� �� � � �� � �� � �� � �� � � � �� �� � � � � � � � � � � � � � � � � � � � � ��

– p.3

Page 4: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Output Data of One Slice

−5

0

5

x 10−4

−5

0

5

x 10−3

−0.01

0

0.01

−0.02

0

0.02

−0.02

0

0.02

0.04

0.06

−0.0200.020.040.06

−0.0200.020.040.06

−0.0200.020.040.060.08

−0.04−0.0200.020.04

−0.01

0

0.01

0.02

−0.01

0

0.01

0.02

−0.01

0

0.01

0.02

−5051015

x 10−3

−5051015

x 10−3

0

5

10

15

x 10−3

0

5

10

15

x 10−3

0

10

20

x 10−3

0

10

20

x 10−3

0

0.01

0.02

0

0.01

0.02

0

0.01

0.02

0.03

0 10 20−5

0

5

10x 10

−4

Time frame0 20 40

−5

0

5

10x 10

−4

Time in minutes

Averageconcentration

Average concentration

� ��� �� �� � � � � � � � � ��� � � �� � �� � � � ��

�� � � ��� � �� ��� �� � � � � �� � � � �� �� � �� � �

� �� � � � �� � ��

– p.4

Page 5: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

What is the FDG Tracer Model?

� � � � u(t) � � � � � � � � � � ��� �� � � � � � � � � � � �� � � ��� ��

� � � � y(t)= y1(t) + y2(t) � � � � � � � � �� � � � � � �� � � � � � � � �� � ��

� � �

� � � � � � �

� � �

K1−→

k2←−

� � �

� � � � � � �

y1(t)

k3−→

k4←−

� � ��

� � � � �� �

y2(t)

�� � � ��

y1 = K1u(t)− (k2 + k3)y1(t) + k4y2(t)

y2 = k3y1(t)− k4y2(t). (1)

• K1

� � �

k2

� � � � � � �� � � �• k3

� � �

k4

� � � �� � � � �� � � � � � � ��� � � � �� � � � �� � � � � � �

� � � � �� K1k3

k2+k3

Cp

LC= K

Cp

LC, � �� � �� � �� �� � � � �� � � �� �� �� � �

� � �� � u(t) � � �

y(t) � � � �� �

K1, k2, k3, k4

� � �

K.

– p.5

Page 6: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

The Solution of the FDG Tracer Model

� � � � � �

K = (K1, k2, k3, k4) �

y(t) = u(t) ⊗(

c1(K)e−λ1(K)t + c2(K)e−λ2(K)t)

,

� � � � �

� � � � � � �� � �� � &� � � � ��� ���

� � � � � � � � ) � � � � � � � � � � k4

� � � � � �� 0 �� � � � � � � � � � � � � � � � � & � ��) & � �) � � � � � � � � � � �

� � � � � � � " � � � � � 60 � � � � � �� � � �) � � � � � � � � " � � � � " � �� � �� & � � � � � � � � � � � �

�� � � � � � � � �

k4 ��� � � � � � � � � �

y(t) = u(t) ⊗

(

K1k3

k2 + k3+

K1k2

k2 + k3e−(k2+k3)t

)

. (2)

�� � � � � � � ��� � � � � �� � � � � � � � ��� � � � � � � � � � � � � �� � � � � � � �� �

�� � �� � �

u(t) � � � � � �� � �� � �� � �� � �

y(t) ! � � � � � � � � � � � � � � � �

�" � � � � � � � �#– p.6

Page 7: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

What are issues with this parameter estimation problem?

� * � � �� � ��� � �� � � � � � � � � � �� � � � ��� � � � � � � � � � � � �* � � � ��) �

� � � � � � � � � � � � �� � � !� � & � � " � � � �� � � � � � ��� � � � � � � � � � � �� � � � � � � � � �

� � � � � � � �� � � � � � � � � �

$� � � �� � � � � � � � � � � � � � � � ��) � � � � � �* � � � � � & ��•

$ � � � � � � � � " � � � � � � �� � � � ��� � � ��� � � � � � � � � � � � �) � � � �" �� � � � � � � � �� � � � � �

� � � � � � � � ��

� � � � � ! � � � � � � � � � ��� � � �� � � � � � � � � � � � � � � � � � �� � � � � � � � � � �

� � � � � � � � � � � � � � � �� � �� � �� � � � � �

– p.7

Page 8: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Representative input/output

0 5 10 15 20 250

2

4

6

8

10

time in minutes

dens

ity

0 5 10 15 20 25−0.01

0

0.01

0.02

0.03

0.04

time in minutes

freq

uenc

y

� � � � % � � � � � �� � � � ��� � u(t) �

' �" � � % � � �� � � �� � � � ��� � � y(t)

�� � 6 � � * � � � �

� � � � � � � � � � � � � ) � � � � � � � � � � � � � �� � " � � � � � t21 = 25m�� � � �� � � � ��� � � � � � � " � � � � � � � � & � � � � � � � � � � � � � � � � " � � � � � � � � � ∆t22 = 30m�

– p.8

Page 9: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Approximate Input Function–Arterial Blood Samples

0 10 20 30 40 50 600

2

4

6

8

10

12

0 0.5 1 1.5 2 2.5 30

2

4

6

8

10

12

Blood samples for t< 3 minutes

Blood samples for whole 60 minutes

� � � � � � � � � � �� � � ��� � ubs(tj), j = 1, 2, · · · , 34.

– p.9

Page 10: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

2. Some Methods for Identifying the Input

��� �� �� �� � � � �� � � � � � � � � � � � � � � � " � � � � � � � � & � � � & ��

� � � � � � � � � � � � � � � � � � � �� � � � ��� � �� � � � � �

��� � � � � � � � � � � � � � �� � � �

� � � � � � � "

� � � � � � � � � � � � � � � � � � ��� � � & � � � � � � � � � � �� � �� � � � �� � � � � � � � � � � ��� � � � � � � �� %

uPhelps = A1eλ1(t−τ) + A2e

λ2(t−τ) + A3eλ3(t−τ),

� �

uFeng = (A1(t − τ0) − A2 − A3)eλ1(t−τ0) + A2e

λ2(t−τ0) + A3eλ3(t−τ0).

' �� � � � �� � � � � � � � � � � � � � � � � � � � �� � � � � � � � � � �

� ( � � " � � � � � � & � � � � � � � �� � � � ��� � � �� � � � �� � � � � � " � � � � � �� � � � � � � � �) � �

� � � " � � � � � � � � � �� � ��

� � �� � � � �� � � �

– p.10

Page 11: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

� � � � � � � � � � � � �� � � � � � � � � � �

20 40 60 80 100 120

20

40

60

80

100

120

20 40 60 80 100 120

20

40

60

80

100

120

20 40 60 80 100 120

20

40

60

80

100

120

20 40 60 80 100 120

20

40

60

80

100

120

0 0.5 1 1.5 20

1

2

3

4

5

0 0.5 1 1.5 20

1

2

3

4

5

0 0.5 1 1.5 20

1

2

3

4

5

0 0.5 1 1.5 20

1

2

3

4

5

� � � �� � 26 � � �

27 � � � � �� � � 1154� � � � �� � � � �� � � �� � � � � � � � ��

�� � � �� � � �� � �� � � �

��� � � � � �� �� � � � ��� � � �� � � � � � � � � � � � � � � � � � �

� � � � � � � ��� � � �� � � � � � � � � � � � � �� �

��� � � ��

� � ��� � � �� � �� � � � � � � �� � ��� � �� � � � � � � � � � � � � �� � � � �� ��

– p.11

Page 12: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Time Activity Curve of Blood Region

10−2

10−1

100

101

102

0

1

2

3

4

5

6

7

8

9

10Blood samplesAverage CA−ROI TACEstimated input functionLinear interpolation points

(τ, θ v(τ))$� � � � � � � � � � � � � � � � � � � � � � � � � � �

� � � � � �� � � � � � � � � � � � � � � � �� � � �� � �

�� � � � � � � � � & � � � " � � � � � � � � � � � � � � &

� �) �� � & � � � � � � � �� � � � � �� � � � � � � � � �

� � � �� � � � ��� ��� � � � � � � � � � � � � � � � � � � �

� � � � � � � � � � � � � � � & � � � " � � � � � �

� � � � � � � � � � & � �) � � � � � � � � � � � � � � � �

' � (� � � � � � � � � � � � � �� � � � � � � �

& � � � � � � � � � � � � � � � � � � � � � � � �� �

� � � �� � � � � � � � � � � � � � � � � � � �� � � �

� � � � � � � � � � � � � � � �� � � � � �� � � � �

� �� � � � � � � � � " � � � � � � � � � � ��

– p.12

Page 13: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

� �� ��� � � � � � � � � � � � � �

�� ��� � � � � � �

� � � � ) � � � � & � �) & � �) � � � � � � � � �� � � � � � � � � � � � � � � � � � � �•

� � � � � � � � � & � �) � � � " � � � � � � � � � �)ue(t, θ, λ, δ) =

0 t ≤ τ0,

θv(τp)((t−τ0)(τp−τ0)

) τ0 < t ≤ τp

θ(v(τ)(t−τp)+v(τp)(τ−t)

(τ−τp) ) τp < t ≤ τ

θv(τ)e−λ(t−τ)δ

t > τ

.

� � � � � � � � � � � � �� � � � � � � � � � � � � � �

θ � λ � � �

δ � � � � � � � � � � � � � � �� τ � τ0 � τp�

θ

� � �� � � �� �� � � � � � � � � � � � &� � � � � � �� � & � � � � � � ��) � � � � � � � � � �� � � � � ��� �

– p.13

Page 14: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

� � � � �� � � � � � � � � � � � �� �� � � � � � �� � � �

� � � � � � � � � � � � � � � � � � � � & � � � ��� � � � � � � �� � � � � � � � � � � � � � � �� � �� � � � � � � � � � �

� � � � � � � � � � �� � � � � � � � � � " � & � � � � � � � �� � � �� �� � � � � � � � � � � �� � � � � � � � � � �

�� � � � � � � �

� �� � � � � � � � � � � � � � � �τ uPhelps uFeng uExp uPExp

1 1.53(−2)

8.0(−3)

5.23(−2)

1.0(−2)

3.65(−1)

1.1(−1)

1.79(−2)

1.1(−2)

2 1.44(−2)

6.2(−3)

1.59(−2)

7.3(−3)

2.75(−1)

9.7(−2)

1.69(−2)

9.9(−3)

3 1.48(−2)

7.4(−3)

1.51(−2)

7.7(−3)�

2.42(−1)

8.3(−2)

1.67(−2)

1.0(−2)

4 1.49(−2)

6.5(−3)

1.44(−2)

7.4(−3)�

2.50(−1)

8.7(−2)

1.68(−2)

1.0(−2)

– p.14

Page 15: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Do numbers tell everything?

10−2

100

102

0

2

4

6

8

10−2

100

102

0

2

4

6

8

10−2

100

102

0

2

4

6

8

10−2

100

102

0

2

4

6

8

blood samplenew−IF Feng−IFTri−exp−IF

� � � � � � �� � � � � �* � � � � � � � � � � � � � � � � � � " � � � �* � �� � �� � � � �" � � t � � � � � � � � �

� � � � " � � � � � � �� � � � � � � ��� � % � � � � � � �� � � �� � � �� � � � � � � � � � � � � � � � � � � � � � � � � � � �

�� � � � � � � �) τ �– p.15

Page 16: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

3.Simultaneous Estimate Algorithm(SIME)

$� � � �� � � � ��� � � � � � � � � � � ��� � � %

minx,α

Φ(x, α) =3

i=1

n∑

j=1

wj

[

yTACi (tj) − αi · yi(tj) − (1 − αi) · ue(tj , θ)

]2.

• αi

� � �� � � � �� � � � � � � � & � � � �� � � � � � � �� � � � � � � � � � � & � � � � � � � � � � � �� � �" �

� � � � � �

� � � � �� � � � � � � � � � � � �� � � � � � � � � � � � � � � �� � � �) � yi

� � � � � �� � � �� � & ��

� � � � � � � � � �� � � � � � � � � � � " �•

� ��� � � & � � � � � � � � � Φ

� �� � � � � � � � � � � � � ��) � � � � � � � � � � �� � �

ue

minλ,δ

3∑

i=1

[

θv(τ)e−λ(ti−τ)δ

− ubs(ti)]2

.

– p.16

Page 17: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Intelligent Clustering of Voxel Data to Remove Noise

−1 0 1 20

5000

10000

15000

Density

Fre

quen

cy

Subject:4248, Slice:16, t=23 Seconds

� −20 −10 0 10 20 30 40 50 600

500

1000

1500

Density

Fre

quen

cy

Subject: 4248, Slice:16, t=43 seconds

−10 −5 0 5 10 15 20 25 300

500

1000

1500

2000

2500

3000

Density

Fre

quen

cy

Subject:4248, Slice:16, t=5.75 minutes

� −10 0 10 20 30 40 500

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

Density

Fre

quen

cy

Subject: 4248, Slice : 16, t=45min

� � � � � � � � � ��� � �� � � � � � � �� � � �� � � � � � � �� � � � � � � �� � �� � � � �� � �� � � �� � � � �

t = 23 �� �� � � � 43 �� �� � � � 5.75 � � � � � � � � � � � �� � � � �� � � �� 45 � � �

� ��� �� � ��� � � �� � � �� � �

– p.17

Page 18: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Cluster curves: An Example

0 5 10 15 20 25 30 35 40 450

0.005

0.01

0.015

0.02

0.025

0.03

0.035

0.04

0.045Averaged curves using clustering methods

time in minutes

conc

entr

atio

n va

lues

curve1curve2curve3curve4curve5

– p.18

Page 19: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Validating Clustering of Voxel Data

� � � � � �

$ � � � � � � � � " � � � � � � � � � � � � � � � � � � � � �

$ � � � � � � � � " � �� � � �� � � ��) � � � � �� � � � � � � � � � � � � � � � � � � �� � ��� � � � � �

� � � � � � � � � � � � " � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �� % � �� � � � � ��� � �� � � � � � � � � � � � � � � � � � � � � �

$ � � � � � � � � " �#" � � � � � � � �� � � � � � � � �� � � � � � ��� � � � � � � � � � � � � � � � � � � � � � � ��

� �

� � � � � � � � � � � � � � � � � � � � � � � � � � �" � � � � � � � � � � � � � ��

$� � � � � � ��� � �

!�� � � � � � � �� � � � � � � � � � � � � � � � � �#" � � � � � � �

� � � � � � � � � � � � � � � � � � � � � � �� � � � � � � � � � � "

– p.19

Page 20: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Comparison of Recovered Input and Measured Data

0 0.5 1 1.50

2

4

6

8

10

12

0<t<1.5

time (min.)

estimated input function

blood samples

� 10−2

10−1

100

101

102

0

2

4

6

8

10

12

log of time (min.)

estimated input function

blood samples

� 10−2

10−1

100

101

102

0

2

4

6

8

10

12

Shifted to alignment

log of time (min.)

estimated input function

blood samples

� � � � �� � � �� � � � ��� � �� � � � � � � � �) � �� � � � � � � � � � � � �� � � � � � � � � � � � � � � � � � �

� � � � � � � � � � �� � ( � � � � � � � � � � � � � � � � � � � � �� �� � � � � � � � � � � � � � � � � � � � � � � �� � ��

� � � � � � � � � � � � � � � � � � " � & � � � � � � " � � � � � � � � � � � � � � � � � � �� � � � � � � � � � � � � � �

� � � � � � � �� � � � � � � � � � � � � � � � � � � � �� � � �� � � � �� � � � � � � � � � � � �) � � � � � � & � �� � � �

� � � � � � � � � � � � � � � � � � �� � ��� � �� � ��

– p.20

Page 21: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Representative Shifted Comparisons

10−2

10−1

100

101

102

0

2

4

6

8

10

121206

10−2

10−1

100

101

102

0

2

4

6

8

10

121227

10−2

10−1

100

101

102

0

2

4

6

8

10

12

140817

10−2

10−1

100

101

102

0

1

2

3

4

5

6

7

8

91154

� � � � � � �� � � � � � � �� � � � � � � ue �� � � � � � � � � � �� � � � � � � � � � � � � � � �� � � ��� � � � � � ubs

�� � � � � � �� � � � � � � �� � �� � � � �� � � � �� � � �� � � ��

– p.21

Page 22: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

� � � � �� � �� � �� � � � � � � � � � � � �� � � � � ��

� � �� � � θ λ δ α1 α2 α3

1206 2.049 0.923 0.233 0.930 0.931 0.926

1227 4.000 1.438 0.158 0.952 0.956 0.953

817 2.843 0.691 0.241 0.935 0.935 0.919

1154 2.730 0.872 0.237 0.958 0.957 0.960

1208 1.790 0.518 0.318 0.912 0.915 0.912

1231 2.294 0.828 0.265 0.938 0.932 0.929

1245 2.287 0.350 0.364 0.928 0.929 0.922

827 3.190 0.797 0.257 0.935 0.934 0.928

1264 3.000 1.558 0.152 0.939 0.952 0.934

1078 4.000 0.889 0.260 0.961 0.963 0.963

1188 2.246 0.628 0.247 0.930 0.925 0.914

1234 3.543 0.781 0.277 0.941 0.941 0.943

1086 4.000 1.121 0.189 0.961 0.960 0.958

1191 3.781 0.792 0.257 0.942 0.940 0.941

� �� � � 3.100 0.89 0.239 0.941 0.939 0.936– p.22

Page 23: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Validation by Quantification

0 0.05 0.1 0.150

0.05

0.1

0.15

K1 , y=0.61x+0.049 ( r=0.64774, p=0.0019)

0 0.05 0.1 0.15 0.2 0.250

0.05

0.1

0.15

0.2

0.25

k2 , y=0.68x+0.03 (r=0.94, p=3.25e−9)

0 0.05 0.10

0.02

0.04

0.06

0.08

0.1

0.12

k3 , y=0.92x+0.005 (r=0.87, p=3.48e−7)

0 0.02 0.04 0.060

0.01

0.02

0.03

0.04

0.05

0.06

0.07

K , y=1.02x+5.5e−05 (r=0.996, p<1.2e−16)

– p.23

Page 24: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

4. Observations from Results

� � � �� � � � � � � � � � K

� � � � � � �� � � � � � � � � � �� � � � � � � �� � � � � � � � � � � � �

� � � �� � � ��

� � � �� � � � � � � � � �� � � � � � � � � � � � � � � � �� � � � � � � � � � � � � � � � � � � � � � � � � � � � �

� � � � � �� � �) � � � � � � � � � �� � � � � � � � � � � � � � � �� � � � � � � � � � � � � � � � � � � � ��

�� � � � � � � � �) � � � � � � � � �) � � � � � � � � � � � � �� � � � � �� � � �� �� � � � � � � � � � � �

0.25 � � � & � � � " ��

� � � � � � & � � � � �� � � � � � � � � � � �� � � � � � � � � �� � � �� �� �� � � � � � � �� � � � � � � � � � � �� � � ��)

� � � � � � � �– p.24

Page 25: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Additional Comments on the Optimization

$� � � � � � � � �� ( � � � �� � � � � � � � �� � � � �� � � � � � � � �� �� � � � � � � � � � ��� � � ��� ���

$ � � � � � � �� � & �� $ � � � � � � � � " � �� & � � �� � � �� � � � � �� � � � � � � �� � � � � � � � ��

0

0.005

0.01

0.015

0.02

0.025

0.03

k4 − no bounds imposed FLS

20 40 60 80

10

20

30

40

50

60

700

0.01

0.02

0.03

0.04

k4 − using bounds from Feng, 1995

20 40 60 80

10

20

30

40

50

60

70

0

0.005

0.01

0.015

0.02

0.025

k4 − using bounds from Piert, 1996

20 40 60 80

10

20

30

40

50

60

700

0.01

0.02

0.03

0.04

k4 − using bounds derived from clustering

20 40 60 80

10

20

30

40

50

60

70 �

0

0.02

0.04

0.06

0.08

0.1K − no bounds imposed FLS

20 40 60 80

10

20

30

40

50

60

700

0.01

0.02

0.03

0.04

0.05

0.06

0.07

K − using bounds from Feng, 1995

20 40 60 80

10

20

30

40

50

60

70

0

0.01

0.02

0.03

0.04

0.05

0.06

K − using bounds from Piert, 1996

20 40 60 80

10

20

30

40

50

60

70

0.02

0.04

0.06

0.08

K − using bounds derived from clustering

20 40 60 80

10

20

30

40

50

60

70

– p.25

Page 26: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Some Remaining Issues

� � � �� � � � � � ��� � � � � � � � � � �� & � � � � � � � � � �� � � � � � � � � � � � �� � � � � � ��� �

�� � � � � � � � � �

' � � � & � � � � � � � � � � � � � " % � � � � � � � � � � � � �� & � � � � � � � � � � ��) � � � � � � � � � � � �

� � � � � �� � � � � � ��� � � � ( � � " � � � ��) � � � � � �� � � � � �� � � � � � � ��) � � � � � � �) � � " $

� �" ��� ��� � ' �� � � � �" � � � � � � ��� � � � � � ' � � �� � �� � � �� � � � � � � �) $ � �� � � �

� � � �) � � � �

( � � �� � � � � � � $� � � � � � � � � � ��) � � � �� � � � � � � �) $ � � � ��

– p.26

Page 27: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

5. MR -PET Image Registration: Initial Work

� � � � � � � � � � � �" � � � � � � ��� � � � � � � � ' � � � � � � � � � � � � � � � � � � � �� � � � � �

� � � � � � � � � � � � � � �� � ��� � � � � � � � " � � � � � � � � � � ��

� � �� � � ( � �� � � � � ��� � � � * � � ��� � � � � � ( � � � � � � � � � � " �� �� � � � � � � � � � � � � � � �

� � � � ��

T1−MR

20 40 60 80 100 120

20

40

60

80

100

120

200

400

600

Clustered PET

20 40 60 80 100 120

20

40

60

80

100

1201

2

3

4

Registered Image using Bspline

20 40 60 80 100 120

20

40

60

80

100

120

200

400

600

Difference of PET and registered MR

20 40 60 80 100 120

20

40

60

80

100

120 −0.5

0

0.5

� * � � � � � � � � � � � ' � � � � � � � " � � � � " � ( � � � � � � � � � �

� � � � � �� � � � � � �� � �� � � � � � � � �

– p.27

Page 28: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

6. Independent Component Analysis for the ROI

� � � � � � � � � � � �� � � � � � � ��� � � � � � � � � � � � � � �� � �ABTAC(t) = α · u(t) + β · y(t)

�� � � & � � � " � � � � � � � �" ��� � � $�

� � $ � � � � � � � � � �" � ��� � � � � � � � � � � $�

y(t) �� $ � � � � � �� � ��

� � � � � � � �� α

� � �

β

� � �� � � � �� � � � � � � � � &� � � � � � � � � � � � � � & � � � � � � ��

� �� � � � � � & � ��) �

� � � � � � � � � � � � � � � � � � �� � �u(t)

�� � � � & � �� � α � � �

β

�) � � � � � � � � � � �� �� $ � � � � � �� � ��

( � � $ � � � � � � $ � � � � � � � � � � � � � � � �� � � � � � � � ��) � � � � � � � � �� � � � � � � � �) �

� � � � �� " � �� � �( � � �� � � � � � � $� � � � � � � � � � ��) � � � � ( $ � � � � � � � $ � �� � � �

� � � �) �

– p.28

Page 29: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

ICA Theory

! � � � � � � � � � � � � � � � � � � � � � � � � � � � � �� �

� � � � � � � � � �� � � � � � � � � � �� �� � � � � * � �

� � � � � � �" � � � � �

� � � � � � � � � � � � � �� �� � � � � � � � � � �� � � � � � � & �� � � � � � � � � � � � � � � � � � � �

� � � � � �� �

� � � � � � � � � �� �� � � � �� � � � � � � � � � � � � � � � �� � � � � � � � � � � � � � �" � � � � � �� � � � �

� � � � � � �" ��� � � �

� � � � � & � � � � � � � �� � � � � � � �� � �" � � � � � � � � � � �� � � � � �� � � � �� � � � � � � � � � � � � �

� � � � � & �� � � � � �" ��� ��

– p.29

Page 30: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

ICA Procedure

!� � � � � � � � � � � � � � % � � ' � � � � �� � � � ��� � �� � � � � � � � � � � � � � � � � � � � � � � �� �� � � � �

� � � � � � �� � � � � � � � � � � � � � &� � � � � � � � � �� � � � � � � � � � � � �� � � � � � � � �) �

� � � � � � � �� � � $ � �� � ( $ � � � ��) � � � �•

!� � � � � � � � � � � � � � ( $ � & � � � � �� � � � � � �� �

20 � � �� � � � �� � � � � � � �)

� � � � �� � � � � � � & �� � � ��

$� � � � � � � � � � � � � � � � &� � � � � � � � � � � � � � & � � � � � � �� �

– p.30

Page 31: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

ICA Setting Determination and Validation

� & � � � � � � �� �� � � � �� � � � � � � � � � � � � � � &� � � � � � ��� � � � � � � � � � � � � � � � ��� ���

� � � � � � � � � � � � � � � � �� � � � � � � � �� � � � �� � � � �� � � � � � � � �) � �" � � � � � � � � � ��� ���

� � � � � � � � � � � � � � � � �� � � � � � �� � �� � � � �� � � � � � � � � � � � � � � � � � � � �" ��� �

� � � � � � ��� ���

$� � � � � � � � � � � � " � � � � � & � � ( � � � � �� � � � � � � & � � ( $ �� � � � � �) � � � � �

� � � � � � � " �

$� � � � � � � � � � � � � �� � � � � � � � $ � '" � �

K

� � � � � ( $ � � � � � �� � � � ��� � ��

� � � � � � � � � � � � � � � � � � � � � � � � � �� � � � � � ���

– p.31

Page 32: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

7. Results of ICA defined ROI : Subvolume

– p.32

Page 33: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

ICA Determined Carotid Artery Region: Whole Brain

– p.33

Page 34: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Blood-sampled and ICA image-derived input functions

� � � � � � � " � � � � � � � � � � � � � � � � � � � �� �� � & ��

– p.34

Page 35: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Voxel-by-voxel CMRgl (K) comparison for one subject

� �� � � �� � � � � � �� � � ��� �� � �� � � �� � � � � � � � ��� � �� ��

�� � � � �� � � � � � �� � �� � � � � � � � � �� � � �

– p.35

Page 36: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Global CMRgl comparison (24 subjects)

� �� � � �� � � � � � � � � � � � � �� � � ��� �� � �� � � �� � � � � � � � � � � � �� � � ��� � �� �

u(t)�

– p.36

Page 37: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Voxel-by-voxel CMRgl comparison (24 subjects)

– p.37

Page 38: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

Conclusion

( � � " � � � � � & � � � � � � � �� � � � ��� � � & � � ��� � � � � � � � � � � � � & � � � � � � � � � � �� � �

�� � � � � � � � � " � � � � � � � � � � � � � � � � � � � � � � � � ( $ � � � � � � ' � (� � � � �* � � � � � � �

� � � � � � � � � & �� �� � � � � � � � � � � � � � � � � � � � � �� � � � ��� � � �•

� � � � � & � � � � � � � � � � �� � � �� � � � � � � � �#" � � � � � � � � � � � � � " � � � � � � � �� �� � �

� � � � �� � � � � � �� � K1, k2

� � �

k3�

� � �� � � � � � � � � � � � � � � � � �#� �� � � � � � � � � � " � �� � � � � � � �� � ) � � � � � � �

� � � � � � � � � � � � � � � � � � � � � �� � � �� � � � � � � " � � � �) � �� �

��� � � � � � � � �� �* � � � � � � � ( $ � � � �� � � � �� � � � � �� � � � � � � � � � �� � � � � � ��� ���

$� � � � � � � � � � � � � � � � � � �� � � �� � � � � � � � � � � � � �� � � �) �

– p.38

Page 39: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

References

� � �� � �� � ��

� �� � � � � ��

��

��� � � � � � � ��

��

� � � �� � � � � ��

� � ��� � � � � � ���

�� � � � � � � ��

��

� � � �

�� � � � � �� �� �� � � � � � �� � � � � � � � �� � �� � � � �

� � � � � � � � � � � � � � �� �� � � �� � �

� � � ��� � ��� � vol. 238 E � � �

� � � � �

� � ���

� ��� � � ��

��

��� � � � � � � ��

��

� �� � � � � ��

��

� � � �� � � � � ��

��

�� � � � � � �

��

��

� � � �

� �� � � � � � � � �� � � � � �� � � � � � �� � �� � � � � � � �� ��

� � � � � � � � � � � � �� � � � � �

18

� �� � � � ��� �� � � � �� �� � �� � � �� � � � � �

� � � � � �

��� � ��� !� � vol. 6 � � �

"# � � " � �

� �# ��

� �� �� � ��

�� �� �� � � � � � ��

� �� � � � � � � �

��� � �� � � � � � � $�

�� �� �� � � � � ��

� �� � � � � � ��

�� � � � � � ��

� � � � ��� � � � � ��

� � ��� � �� �� � � � � �� �� ��

� �� � � � � �� � � � � ��� �� � � � �� �� � � � � � � � � � � � � � � � � �� � � � � � �

� � � � � � � ��� � �� � � � � � � � � � �� � � � � �� �� � � � � � � � � � � � � � �� �%�& ' ��� �(�) � vol. 188 � � �

� " � � � "�

� � � "�

�� �� � � ��

� �� � � � � ��

��

� �� � � � � � ��

��

� � � � � � � � �

��

� � � � � � � � ���

� � � ��� � �� �� � �� � � � � � � � � � � � �� � � � �� �� � � � � � � � � � � � � � ��

� �� � � � � �� �� � � � � � �� � � � � � � � � � � � � � * ! � � + � ,� ' vol.24 � � �

� � � � " �- � � �# �

– p.39

Page 40: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

References

� � � �� �� �

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �� � � � � ��� � �� �

� � � � � � � �� � � � � �� � vol. 21, no. 6 � � �

�� # � � � �

� � �# �

� � � � ��

� � � � � � ��

��

� �� � � � � � �

� � � � � � � � �

��

� � � � � � � ���

� � � �� ��� � � � ��

��

� � �� � � � � ��

� � � � � � � ��

� �� � � � � ��

��� � � � �� �

� � � � � � � �� � � �� � � � �� � � � � � � ��� � � � � � � ��� � � � � � �

� � � � � �� �� �� � � � �� � �� � � � � � � �� � � � � � � ��� !� � &) � vol. 21, no. 2

� � � - � " �- � "

� � � - �

� � ��� � � ��

� ��� � � � � ��

�� � � � � � � ��

�� � �� � � � � �� � ��

� �� � � � � ��

� � �� �

� � � ��

�� � � � � � � ��

� � � � � � ��

� �� � �� � � � � �� �� �� � �� � � � �� � � �

� � � � �� � � � � � � � � � � � � � � �� � � �� �� � �� � � � �� � � � � � � �� � �

� � � � � � � 18

� �� � � � ��� �� � � � �� �� ��� � �� � �� � � � � � � � �

� �� � � ��� � �� � � � � � � � � � � � � �

� �� !� � � �(� � ' � �(� � ,� & � vol. 18 � � �

# �� � # � "

� � � ��

� �� � � ��

��

�� � � � � �� � �

$� � � � � �� � ��

$ � � � � $� � ��

� � � � �

� � � � � � �� �� � � � � � � � � � �� � � � � �� � � � � �� � � � � � ��� � � � � � � � � � �

� � � � � � � �� � � � � �� � � � � � � � � � � � �� � � � � � � � � � � �� � ��� � � � � �

�� � � �� � � � � � � � �

* * * � !& � � � �� ! �� + �� � � � ��� � ' vol. 1, no. 4

� � �

�- " � �� - � � �# �

– p.40

Page 41: Prof. Rosemary A. Renaut Dr. Hongbin Guo & Dr. Haewon Nam ...rosie/mypresentations/asucsetalk.pdf · Intelligent Clustering of Voxel Data to Remove Noise −1 0 1 2 0 5000 10000 15000

References

� �� � � � � ��

��

�� �� � � � � � � � � � �� � � � � ��

� � � � � � � �

� � � � � � � � ��

� � �� � �

� � � ��

� � � � � � � � ��

� � � � � � � � � $�

� � � � � ��

� � � � � � � � ��

�� � � � ��� � � � � � ��

�� � �� ��� � � � ��

��

�� � � � � ��

��

� � � � � � �� � � ��� � �� � � � � � � � � � � � � �� �

11

� �� � �� �� � � � �� �� � � � �� � � �� � � �

� � �� � � � � � �

, � � &) � ��� � vol. 5, no. 2 � � � # � � # �

� � � "�

� $� � � �� � �

$� � � � � ��

�� � � � � ��

��

� � � � ��� � � � ��

��

� � � �� � � � � � � � � �� � � � � � � �� � � � � � � � � �� � � � � �� � � � � �� � � � � � ��� � � � �

� � � � � � � � � � �� �� � � ��� � �

* * * � !& � � � �� ! � �� + �� � � � ��� � '

vol. 5, no. 1 � � �� # � # �

� � � ��

– p.41