235.273 267.182 Novel Non-invasive Estimation of Coronary ... · Background TAFE Formulation TAFE...

1
Novel Non-invasive Estimation of Coronary Blood Flow using Contrast Advection in Computed Tomography Angiography Parastou Eslami 1 , Jung-Hee Seo 1 , Amir Ali Rahsepar 2 , Richard George 2 , Albert Lardo 2,3 , Rajat Mittal 1 1. Department of Mechanical Engineering, Johns Hopkins University 2. Department of Medicine, Division of Cardiology, Johns Hopkins University 3. Biomedical Engineering, Johns Hopkins University, Baltimore, MD Background TAFE Formulation TAFE Validation Coronary computed tomography angiography (CTA) is a promising tool for assessment of coronary stenosis and plaque burden. Recent studies have shown the presence of axial contrast concentration gradients in obstructed arteries, but the mechanism responsible for this phenomenon is not well understood. We use computational fluid dynamics to study intracoronary contrast dispersion and the correlation of concentration gradients with intracoronary blood flow and stenotic severity. Data from our CFD patient-specific simulations reveals that contrast dispersions are generated by intracoronary advection effects, and therefore, encode the coronary flow velocity. This novel method- Transluminal Attenuation Flow Encoding (TAFE) - is used to estimate the flowrate in phantom studies as well as preclinical experiments. Methods Patient-specific models were extracted from CT images using Mimics © Stenosis geometry was manually created from the normal vessel to mimic 70% stenosis. Number of tetrahedral elements ~ 4.55 x 10 5 (Figure 1) Used ANSYS for CFD solutions: 2nd order backward Euler scheme, ∆t= 0.01 s Used Arterial Input Function (AIF) and Transluminal Attenuation Gradient (TAG) from CT images (Figure 2) to solve for the flow rate analytically :TAFE (Figure 3 and EQ 1). Figure 3 : AIF (left) and Transluminal (right) profile of the contrast agent. TAG is the slope of concentration of contrast agent vs the distance along the vessel plot. Figure 1 : (A) 3-D model of a patient specific coronary artery for the normal (unstenosed) case. (B) Model of the artery with 70% stenosis. (C) Computational meshes employed in the various segments of the model. A B C t t s T d HU max s Temporal profile at the ostium Transluminal profile at t=t s +T d max TAG ~ CF d HU VT HU V CF T d V CF,n 1 2 3 3' 2' 1' V CF,s 3" 2" 1" TAG Normal Stenosed Figure 2: Example of transluminal contrast gradient for a stenosed artery. Luminal cross sections are sampled and plotted over the vessel length to obtain an axial variation of cross-sectional averaged attenuation (HU) (top figure). Bottom figure shows the axial and cross- sectional visualizations of lumen area by contrast agent. Transport of Contrast : + . = 2 1D Equation Approximation 1 () + 1 () ≈0 = == Q = −A s . (1) TAFE Formula for predicting CFV from TAG Computational Results Concentration (B) Figure 4 : Patient specific computational results: Pressure (A), normalized contrast agent concentration (C/C max ) for the stenosed case (B) and no stenosis case (C) Concentration (C) Pressure (A) Figure 5 : Patient specific normalized transluminal attenuation profiles along the axial direction of main arterial segment in LAD (steady flow) with P A =3 and P B =0 mmHg. (A) Correlation between CFD calculation of flow rate and TAFE calculation of the flow rate in the no-stenosis (normal) and 70% area stenosis cases. (B) and (C) Steigner et. al. (A) (B) (C) Experimental Setup Phantom Study Pre-clinical Study Contrast is following a parabolic velocity profile CT imaging artifacts, ex: PVA and filtering Conclusions CFD analysis provides insights into the hemodynamic mechanism for TAG generation Insights and mathematical analysis lead to o Identification of key control variables o Formula for quantification of CFV from TAG (TAFE) TAFE does NOT require: o Anesthesia o Catheter placement o Sophisticated CFD modeling and hardware o CT data to be sent outside of the hospital for analysis Phantom studies for a straight tube and pre-clinical studies demonstrates a promising validation of the TAFE formulation with the assumption of linear AIF Figure 7 : Correlation between coronary blood flow (TAFE estimated), indexed to myocardial mass and microsphere myocardial blood flow (measured data)in the stenosis model (A) and MI models added to the stenosis model (B). Figure 6 : Correlation between TAFE estimate velocity and the true pump velocity. The correlation is far from the 45 degree angle because of radial variation of contrast because of specific gravity and imaging artifacts such as partial volume averaging Acknowledgement & Conflict of Interest RM & ACL would like to acknowledge support from the Coulter Foundation for this work. PE is funded by the Graduate Partnership Program fellowship at NIH. RM and ACL have other significant financial interests in technologies associated with contrast agent gradient based flow assessment. y = 0.9225x - 6.0123 R² = 0.9288 10 15 20 25 30 35 40 10 20 30 40 50 TAFE Estimated Flowrate (ml/min True Pump Flowrate (ml/min) x y 200 250 300 200 250 300 C 331 299.091 267.182 235.273 203.364 171.455 139.545 107.636 75.7273 43.8182 11.9091 -20 Radial Variation of Contrast

Transcript of 235.273 267.182 Novel Non-invasive Estimation of Coronary ... · Background TAFE Formulation TAFE...

Page 1: 235.273 267.182 Novel Non-invasive Estimation of Coronary ... · Background TAFE Formulation TAFE Validation Coronary computed tomography angiography (CTA) is a promising tool for

Novel Non-invasive Estimation of Coronary Blood Flow using

Contrast Advection in Computed Tomography Angiography

Parastou Eslami1, Jung-Hee Seo1, Amir Ali Rahsepar2, Richard George2, Albert Lardo2,3, Rajat

Mittal1

1. Department of Mechanical Engineering, Johns Hopkins University 2. Department of Medicine, Division of

Cardiology, Johns Hopkins University 3. Biomedical Engineering, Johns Hopkins University, Baltimore, MD

Background TAFE Formulation TAFE ValidationCoronary computed tomography angiography (CTA) is a promising tool for

assessment of coronary stenosis and plaque burden. Recent studies have shown

the presence of axial contrast concentration gradients in obstructed arteries, but

the mechanism responsible for this phenomenon is not well understood. We use

computational fluid dynamics to study intracoronary contrast dispersion and the

correlation of concentration gradients with intracoronary blood flow and stenotic

severity. Data from our CFD patient-specific simulations reveals that contrast

dispersions are generated by intracoronary advection effects, and therefore,

encode the coronary flow velocity. This novel method- Transluminal Attenuation

Flow Encoding (TAFE) - is used to estimate the flowrate in phantom studies as

well as preclinical experiments.

Methods• Patient-specific models were extracted from CT images using Mimics©

• Stenosis geometry was manually created from the normal vessel to mimic 70%

stenosis. Number of tetrahedral elements ~ 4.55 x 105 (Figure 1)

• Used ANSYS for CFD solutions: 2nd order backward Euler scheme, ∆t= 0.01 s

• Used Arterial Input Function (AIF) and Transluminal Attenuation Gradient (TAG)

from CT images (Figure 2) to solve for the flow rate analytically :TAFE (Figure 3

and EQ 1).

Figure 3: AIF (left) and

Transluminal (right) profile of

the contrast agent. TAG is the

slope of concentration of

contrast agent vs the distance

along the vessel plot.

Figure 1: (A) 3-D model of a

patient specific coronary artery for

the normal (unstenosed) case. (B)

Model of the artery with 70%

stenosis. (C) Computational

meshes employed in the various

segments of the model.

A

B

C

s

Av(s)

stenosis

Normal

Flow Limiting,

Stenosed

Q

Qn

ostium

Q

Qs

Qn>Qs, VCF,n>VCF,s

VCF,n

VCF,s

t

ts Td

HUmax

s

Temporal profile at the ostium Transluminal profile at t=ts+Td

maxTAG ~CF d

HU

V T

HU

VCFTd

VCF,n

1

2

3 3'

2'

1'VCF,s

3"

2"

1"

TAG

Normal

Stenosed

a

b

Figure 2: Example of transluminal

contrast gradient for a stenosed artery.

Luminal cross sections are sampled and

plotted over the vessel length to obtain

an axial variation of cross-sectional

averaged attenuation (HU) (top figure).

Bottom figure shows the axial and cross-

sectional visualizations of lumen area by

contrast agent.

Transport of Contrast:

𝜕𝐶

𝜕𝑡+ 𝑼. 𝛁 𝐶 = 𝐷𝛻2𝐶

1D Equation Approximation 1

𝑄(𝑡)

𝜕𝐶

𝜕𝑡+1

𝐴(𝑠)

𝜕𝐶

𝜕𝑠≈ 0

𝑇𝐴𝐺 =𝜕𝐶

𝜕𝑠==𝐴 𝑠

𝑄

𝜕𝐶

𝜕𝑡 Q = −A s

𝜕𝐶𝜕𝑡𝜕𝐶𝜕𝑠

𝐸𝑄. (1)

TAFE Formula for predictingCFV from TAG

Computational ResultsConcentration (B)

Figure 4: Patient specific computational results: Pressure (A), normalized contrast

agent concentration (C/Cmax) for the stenosed case (B) and no stenosis case (C)

Concentration (C)Pressure (A)

Figure 5: Patient specific

normalized transluminal

attenuation profiles along the

axial direction of main arterial

segment in LAD (steady flow)

with PA=3 and PB=0 mmHg. (A)

Correlation between CFD

calculation of flow rate and

TAFE calculation of the flow

rate in the no-stenosis (normal)

and 70% area stenosis cases.

(B) and (C) Steigner et. al.

(A) (B)

(C)

Experimental Setup

Phantom Study

Pre-clinical Study

• Contrast is following a parabolic velocity profile

• CT imaging artifacts, ex: PVA and filtering

Conclusions• CFD analysis provides insights into the hemodynamic mechanism for

TAG generation

• Insights and mathematical analysis lead to

o Identification of key control variables

o Formula for quantification of CFV from TAG (TAFE)

• TAFE does NOT require:

o Anesthesia

o Catheter placement

o Sophisticated CFD modeling and hardware

o CT data to be sent outside of the hospital for analysis

• Phantom studies for a straight tube and pre-clinical studies demonstrates

a promising validation of the TAFE formulation with the assumption of

linear AIF

Figure 7: Correlation between coronary blood flow (TAFE estimated), indexed to

myocardial mass and microsphere myocardial blood flow (measured data)in the stenosis

model (A) and MI models added to the stenosis model (B).

Figure 6: Correlation between TAFE estimate velocity and the true pump velocity. The

correlation is far from the 45 degree angle because of radial variation of contrast because

of specific gravity and imaging artifacts such as partial volume averaging

Acknowledgement & Conflict of Interest RM & ACL would like to acknowledge support from the Coulter Foundation for this work. PE is funded by the

Graduate Partnership Program fellowship at NIH. RM and ACL have other significant financial interests in

technologies associated with contrast agent gradient based flow assessment.

y = 0.9225x - 6.0123R² = 0.9288

10

15

20

25

30

35

40

10 20 30 40 50

TAFE

Est

imat

ed F

low

rate

(m

l/m

in

True Pump Flowrate (ml/min)

x

y

200 250 300

200

250

300C

331

299.091

267.182

235.273

203.364

171.455

139.545

107.636

75.7273

43.8182

11.9091

-20

Radial Variation of

Contrast