Biomedical Data Analysis

(BML738)

Dr. Anup Singh,

Centre for Biomedical Engineering
IIT-Delhi

Semester-II, 2024-25 Lecture-16: Dynamic data analysis

 PL model fitting examples
2
Concentration (mM/L)

a b c

+b2
f
Concentration (mM/L)

d e A

B E 0
C

Time points (unit time point = 5.25 s)

D
-b2
Figure 6.2: shows example graphs(a-e) of PL-fitting to concentration
time curves at different voxels marked on the slope-2 colored image (f).
Generalized Piecewise Linear Fitting Model

Singh A, et al. JMRI 2009

 Given a random sample of n observations S(t) (or C(t)) for t
=1, 2, 3,…n, at a given pixel the generalized PL model has the
form:

c; t 
c + b (t −  );  t 

f (t ) = 
1
(6.8)
c + b1 (  −  ) + b2 (t −  );  t 
c + b1 (  −  ) + b2 ( −  ) + b3 (t −  ); t 

where the parameter α is the BAT, β (TTP) is the point of

intersection of 2nd and 3rd line segment, γ is the point of
intersection of 3rd and 4th line segment, c is the constant (noise
level). Parameters b1 (slope-1), b2 (slope-2) and b3 (slope-3)
are the slopes of 2nd, 3rd and 4th line segments respectively, of
fitting model function. Units of α, β and γ are s-1 and units of
slopes are mmol L-1 s-1.
 0.80

Concentration (m Mol/L)
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31

Time Points (unit time point = 5.25 s)

Figure 6.5 show the example of fitting a generalized

PL model (bold line) to a C(t) curve measured in a
normal brain tissue (arterial voxel).
 BML738

Mathematical Modeling of C(t)

6

 Compartmental modeling
o Tissue is assumed to be composed of multiple compartments
o Well mixing of contrast agent is assumed in a particular
compartment
o Change in the concentration of contrast in one of the tissue
compartments is assumed to be linear combination of
concentration in all other compartments.

o Cp(t) or AIF or C0(t) can be measured or modeled separately. So

it is assumed to be a known quantity.

Ref: [Link]
 BML738

Mathematical Modeling of C(t)

7

 One-Tissue model or 2-compartment model. (Model-1)

K1
Cp C1

𝑑𝐶1(𝑡)
= 𝐾1𝐶𝑝 𝑡 , 𝑤ℎ𝑒𝑟𝑒 𝐾1 𝑖𝑠 𝑟𝑎𝑡𝑒 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡
𝑑𝑡

𝑡
 C1(t) = 𝐾1 ‫׬‬0 𝐶𝑝 г 𝑑г

 C (t) = [Link](t) + C1(t)

The term, Vp Cp , is included as vasculature is also part of the tissue

 BML738

Dynamic data or Perfusion MRI data

8
6 1.8
Concentration (mM/L)

1.6
5 AIF or Cp(t)
1.4

4 1.2

1
3
0.8

2 0.6

0.4
1
0.2

0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5

Time (minutes)
6
Concentration (mM/L)

0
0 0.5 1 1.5 2 2.5

Time (minutes)
 BML738

Mathematical Modeling of C(t)

9

 One-Tissue model or 2-compartment model. (Model-2)

K1
Cp C1

K2

𝑑𝐶1(𝑡)
⇒ = 𝐾1𝐶𝑝 𝑡 − 𝐾2𝐶1 𝑡
𝑑𝑡
𝑡
 C1(t) = 𝐾1 ‫׬‬0 𝐶𝑝 𝑡 − г . 𝑒 −𝐾2.г 𝑑г

𝑡
 C (t) = Vp. Cp(t) + C1(t) = [Link](t) +𝐾1 ‫׬‬0 𝐶𝑝 𝑡 − г . 𝑒 −𝐾2.г 𝑑г
 Solution of differential equ.
10

𝒅𝑪𝟏(𝒕)
= 𝑲𝟏𝑪𝒑 𝒕 − 𝑲𝟐𝑪𝟏 𝒕 [𝟏]
𝒅𝒕
𝒕
 C1(t) = 𝑲𝟏 ‫ 𝒕 𝒑𝑪 𝟎׬‬− г . 𝒆−𝑲𝟐.г 𝒅г

Source: [Link]
 BML737

C(t) using Model-1 and Model-2

11

Model-2
Model-1
1.8 1.8

1.6 1.6

1.4 1.4

1.2 1.2

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
 BML737

12

[Link]
 BML737

Dynamics of FDG-PET
13

Figure. A dynamic 18 F-FDG-PET

image series in the coronal plane
showing the heart and the
implanted bilateral human breast
cancer xenografts MAS98.12
xenografts. The corresponding
uptake kinetics are shown in graphs
A and B, respectively, where the
latter constitutes the arterial input
function (AIF). Also, the results
from the pharmacokinetic model fi t
are shown.

Source: Kristian A, et al. Dynamic 18 F-FDG-PET for monitoring treatment effect following
anti-angiogenic therapy in triple-negative breast cancer xenografts. Acta Oncologica, 2013; 52:
1566–1572
 BML737
Dynamics of FDG-PET (TAC)
15

Figure: (a)The standard fully

compartmental FDG-PET tracer
kinetic model was employed in
this study. It consists of the(Cp)
input function compartment, as well
as the tissue compartments for
free-tracer (C1) and metabolized
FDG (C2) concentration over post-
injection time.

(b) the k-parameter table

(c) the TACs for multiple normal and

tumor regions.

Source: Nicolas A. Karakatsanis, et al. Generalized 3d and 4d motion compensated whole-body pet image
reconstruction employing nested EM deconvolution. IEEE. DOI: 10.1109/IST.2014.6958485
 BML737

16

Source:[Link]
net/petanalysis/model_compartm
[Link]
 BML738

Arterial Input Function(AIF)-PET data

21

Arterial input function or Concentration

of tracer in blood plasma (Cp(t))
 BML738

AIF or Cp(t) vs Time curve

22

3.5

2.5

1.5

0.5

0
0 2 4 6 8 10

Time (min)
 BML738

AIF or Plasma curve modeling

23

 Concentration of tracer in blood plasma (Cp(t)) can be

modeled using following empirical model function:

Cp(t)
 BML738

Parameters from Fitting AIF

24

 Results of fitting AIF using Eq.[2] for different scans

 BML738
Summary of Dynamic Data based upon
contrast agent injection
25

 Dynamic Data can be obtained using different imaging/signal

modalities and different types of contrast agents
 Qualitative Analysis of Dynamic data
o Based upon observations. Low vs high, slow-enhancement, etc.
 Semi-quantitative Analysis of Dynamic data
o A number of parameters can be computed by simple calculations:
◼ BAT, Peak value, time to peak, AUC, wash-in-slope, etc.
o Dynamic data can be analyzed using Empirical Models
◼ Piece-wise linear model, polynomials, exponential functions, etc.
o Semi-quantitative parameters are in-directly related to physiology
 Quantitative Analysis of Dynamic data
o Based upon mathematical modeling, such as compartmental modeling
or pharmacokinetic modeling
o Parameters are directly related to physiological information
 BML738

Dynamic Biomedical data using some Modulation

26

 Modulation of concentration of a molecules, examples

o Creatine (Cr) Or Phosphocreatine(PCr) concentration
◼ modulation of concentration using exercise
o Change in concentration of neurotransmitters
◼ Glutamate, GABA, etc.
◼ Modulation of their concentration using coffee, alcohol, Movie, etc.
o Change in concentration of a molecule using some enzymes
injection
 Change in heart rate
o Exercise, stress, excitement, yoga, drugs, etc.
 Change in blood flow due to exercise
o fMRI data example
 Insulin injection role in maintaining glucose level
 Think for more examples
 BML738

CrCEST in Calf Muscle @ 7T

27

The creatine kinase reaction plays a vital role in the muscle energetics.

Cr + ATP ADP + PCr + H+

➢ [Cr] = 15mM; [PCr] = 33mM

Feliks Kogan, Mohammad Haris, Anup Singh, et al. Mag Reson. Med. 2012
 BML738

28

MR spectra of the same volunteer pre- and post-exercise, and at the end of the recovery period a) at 3T and
b) at 7T. We have normalized the amplitude of all post-exercise spectra to the pre-exercise ones. We
observed an almost threefold increase of SNR at 7T relative to 3T. Evolution of PCr MRS signal intensity
during the execution of the exercise (I) and the recovery period (II) from the same volunteer. PCr depletion
rates were estimated by fitting a linear function to phase I, while the PCr recovery kinetics were characterized
by fitting a mono-exponential growth function to phase II of the exercise. (c) At 3T, the PCr depletion rate is
0.35 s−1 (estimated from the slope of the fitted line, r = 0.997) and the recovery rate constant is 22.4 s (r =
0.981). (d) At 7T the PCr depletion rate is 0.49 s−1 (r = 0.945) and the recovery rate constant is 23.89 s (r =
0.998).

Adapted from Ref: [Link]

 BML738

Dynamic data of PCr modulation

29

 The rate of phosphocreatine (PCr) resynthesis following

physical exercise, has been extensively studied with 31P-
MRS.

Adapted from Ref: [Link]

 BML738

31P-MRI at 3T and 7T MRI scanner

30

Adapted from Ref: [Link]

 Dynamic data of Cr modulation BML738

Plantar Flexion Exercise Data @7T

31

Feliks Kogan, Mohammad Haris, Anup Singh, et al. Mag Reson. Med. 2012