Convolutional Neural Networks for Direct Inference of Pharmacokinetic Parameters

Application to Stroke Dynamic Contrast-Enhanced MRI

Cagdas Ulas (Corresponding Author), Dhritiman Das, Michael J Thrippleton, Maria del C. Valdes Hernandez, Paul A Armitage, Stephen D Makin, Joanna M Wardlaw, Bjoern H Menze

Research output: Contribution to journalArticle

Abstract

Background and Purpose: The T1-weighted dynamic contrast enhanced (DCE)-MRI is an imaging technique that provides a quantitative measure of pharmacokinetic (PK) parameters characterizing microvasculature of tissues. For the present study, we propose a new machine learning (ML) based approach to directly estimate the PK parameters from the acquired DCE-MRI image-time series that is both more robust and faster than conventional model fitting. Materials and Methods: We specifically utilize deep convolutional neural networks (CNNs) to learn the mapping between the image-time series and corresponding PK parameters. DCE-MRI datasets acquired from 15 patients with clinically evident mild ischaemic stroke were used in the experiments. Training and testing were carried out based on leave-one-patient-out cross- validation. The parameter estimates obtained by the proposed CNN model were compared against the two tracer kinetic models: (1) Patlak model, (2) Extended Tofts model, where the estimation of model parameters is done via voxelwise linear and nonlinear least squares fitting respectively. Results: The trained CNN model is able to yield PK parameters which can better discriminate different brain tissues, including stroke regions. The results also demonstrate that the model generalizes well to new cases even if a subject specific arterial input function (AIF) is not available for the new data. Conclusion: A ML-based model can be used for direct inference of the PK parameters from DCE image series. This method may allow fast and robust parameter inference in population DCE studies. Parameter inference on a 3D volume-time series takes only a few seconds on a GPU machine, which is significantly faster compared to conventional non-linear least squares fitting.
Original languageEnglish
Article number1147
JournalFrontiers in Neurology
Volume9
DOIs
Publication statusPublished - 8 Jan 2019

Fingerprint

Pharmacokinetics
Stroke
Neural Networks (Computer)
Least-Squares Analysis
Population Dynamics
Microvessels
Brain
Machine Learning

Keywords

  • contrast agent concentration convolutional neural networks dynamic contrast enhanced MRI ischaemic stroke loss function pharmacokinetic parameter inference tracer kinetic modeling
  • Loss function
  • Dynamic contrast enhanced MRI
  • Tracer kinetic modeling
  • Pharmacokinetic parameter inference
  • Convolutional neural networks
  • Contrast agent concentration
  • Ischaemic stroke
  • SMALL VESSEL DISEASE
  • dynamic contrast enhanced MRI
  • contrast agent concentration
  • loss function
  • ischaemic stroke
  • pharmacokinetic parameter inference
  • MODELS
  • TRACER
  • convolutional neural networks
  • BLOOD
  • tracer kinetic modeling

ASJC Scopus subject areas

  • Clinical Neurology
  • Neurology

Cite this

Convolutional Neural Networks for Direct Inference of Pharmacokinetic Parameters : Application to Stroke Dynamic Contrast-Enhanced MRI. / Ulas, Cagdas (Corresponding Author); Das, Dhritiman; Thrippleton, Michael J; Valdes Hernandez, Maria del C. ; Armitage, Paul A; Makin, Stephen D; Wardlaw, Joanna M; Menze, Bjoern H.

In: Frontiers in Neurology, Vol. 9, 1147, 08.01.2019.

Research output: Contribution to journalArticle

Ulas, Cagdas ; Das, Dhritiman ; Thrippleton, Michael J ; Valdes Hernandez, Maria del C. ; Armitage, Paul A ; Makin, Stephen D ; Wardlaw, Joanna M ; Menze, Bjoern H. / Convolutional Neural Networks for Direct Inference of Pharmacokinetic Parameters : Application to Stroke Dynamic Contrast-Enhanced MRI. In: Frontiers in Neurology. 2019 ; Vol. 9.
@article{fcc7b1bdd1f44553814b636abb1b4641,
title = "Convolutional Neural Networks for Direct Inference of Pharmacokinetic Parameters: Application to Stroke Dynamic Contrast-Enhanced MRI",
abstract = "Background and Purpose: The T1-weighted dynamic contrast enhanced (DCE)-MRI is an imaging technique that provides a quantitative measure of pharmacokinetic (PK) parameters characterizing microvasculature of tissues. For the present study, we propose a new machine learning (ML) based approach to directly estimate the PK parameters from the acquired DCE-MRI image-time series that is both more robust and faster than conventional model fitting. Materials and Methods: We specifically utilize deep convolutional neural networks (CNNs) to learn the mapping between the image-time series and corresponding PK parameters. DCE-MRI datasets acquired from 15 patients with clinically evident mild ischaemic stroke were used in the experiments. Training and testing were carried out based on leave-one-patient-out cross- validation. The parameter estimates obtained by the proposed CNN model were compared against the two tracer kinetic models: (1) Patlak model, (2) Extended Tofts model, where the estimation of model parameters is done via voxelwise linear and nonlinear least squares fitting respectively. Results: The trained CNN model is able to yield PK parameters which can better discriminate different brain tissues, including stroke regions. The results also demonstrate that the model generalizes well to new cases even if a subject specific arterial input function (AIF) is not available for the new data. Conclusion: A ML-based model can be used for direct inference of the PK parameters from DCE image series. This method may allow fast and robust parameter inference in population DCE studies. Parameter inference on a 3D volume-time series takes only a few seconds on a GPU machine, which is significantly faster compared to conventional non-linear least squares fitting.",
keywords = "contrast agent concentration convolutional neural networks dynamic contrast enhanced MRI ischaemic stroke loss function pharmacokinetic parameter inference tracer kinetic modeling, Loss function, Dynamic contrast enhanced MRI, Tracer kinetic modeling, Pharmacokinetic parameter inference, Convolutional neural networks, Contrast agent concentration, Ischaemic stroke, SMALL VESSEL DISEASE, dynamic contrast enhanced MRI, contrast agent concentration, loss function, ischaemic stroke, pharmacokinetic parameter inference, MODELS, TRACER, convolutional neural networks, BLOOD, tracer kinetic modeling",
author = "Cagdas Ulas and Dhritiman Das and Thrippleton, {Michael J} and {Valdes Hernandez}, {Maria del C.} and Armitage, {Paul A} and Makin, {Stephen D} and Wardlaw, {Joanna M} and Menze, {Bjoern H}",
note = "Ulas, Cagdas Das, Dhritiman Thrippleton, Michael J Valdes Hernandez, Maria Del C Armitage, Paul A Makin, Stephen D Wardlaw, Joanna M Menze, Bjoern H eng Switzerland Front Neurol. 2019 Jan 8;9:1147. doi: 10.3389/fneur.2018.01147. eCollection 2018.",
year = "2019",
month = "1",
day = "8",
doi = "10.3389/fneur.2018.01147",
language = "English",
volume = "9",
journal = "Frontiers in Neurology",
issn = "1664-2295",
publisher = "Frontiers Media S.A.",

}

TY - JOUR

T1 - Convolutional Neural Networks for Direct Inference of Pharmacokinetic Parameters

T2 - Application to Stroke Dynamic Contrast-Enhanced MRI

AU - Ulas, Cagdas

AU - Das, Dhritiman

AU - Thrippleton, Michael J

AU - Valdes Hernandez, Maria del C.

AU - Armitage, Paul A

AU - Makin, Stephen D

AU - Wardlaw, Joanna M

AU - Menze, Bjoern H

N1 - Ulas, Cagdas Das, Dhritiman Thrippleton, Michael J Valdes Hernandez, Maria Del C Armitage, Paul A Makin, Stephen D Wardlaw, Joanna M Menze, Bjoern H eng Switzerland Front Neurol. 2019 Jan 8;9:1147. doi: 10.3389/fneur.2018.01147. eCollection 2018.

PY - 2019/1/8

Y1 - 2019/1/8

N2 - Background and Purpose: The T1-weighted dynamic contrast enhanced (DCE)-MRI is an imaging technique that provides a quantitative measure of pharmacokinetic (PK) parameters characterizing microvasculature of tissues. For the present study, we propose a new machine learning (ML) based approach to directly estimate the PK parameters from the acquired DCE-MRI image-time series that is both more robust and faster than conventional model fitting. Materials and Methods: We specifically utilize deep convolutional neural networks (CNNs) to learn the mapping between the image-time series and corresponding PK parameters. DCE-MRI datasets acquired from 15 patients with clinically evident mild ischaemic stroke were used in the experiments. Training and testing were carried out based on leave-one-patient-out cross- validation. The parameter estimates obtained by the proposed CNN model were compared against the two tracer kinetic models: (1) Patlak model, (2) Extended Tofts model, where the estimation of model parameters is done via voxelwise linear and nonlinear least squares fitting respectively. Results: The trained CNN model is able to yield PK parameters which can better discriminate different brain tissues, including stroke regions. The results also demonstrate that the model generalizes well to new cases even if a subject specific arterial input function (AIF) is not available for the new data. Conclusion: A ML-based model can be used for direct inference of the PK parameters from DCE image series. This method may allow fast and robust parameter inference in population DCE studies. Parameter inference on a 3D volume-time series takes only a few seconds on a GPU machine, which is significantly faster compared to conventional non-linear least squares fitting.

AB - Background and Purpose: The T1-weighted dynamic contrast enhanced (DCE)-MRI is an imaging technique that provides a quantitative measure of pharmacokinetic (PK) parameters characterizing microvasculature of tissues. For the present study, we propose a new machine learning (ML) based approach to directly estimate the PK parameters from the acquired DCE-MRI image-time series that is both more robust and faster than conventional model fitting. Materials and Methods: We specifically utilize deep convolutional neural networks (CNNs) to learn the mapping between the image-time series and corresponding PK parameters. DCE-MRI datasets acquired from 15 patients with clinically evident mild ischaemic stroke were used in the experiments. Training and testing were carried out based on leave-one-patient-out cross- validation. The parameter estimates obtained by the proposed CNN model were compared against the two tracer kinetic models: (1) Patlak model, (2) Extended Tofts model, where the estimation of model parameters is done via voxelwise linear and nonlinear least squares fitting respectively. Results: The trained CNN model is able to yield PK parameters which can better discriminate different brain tissues, including stroke regions. The results also demonstrate that the model generalizes well to new cases even if a subject specific arterial input function (AIF) is not available for the new data. Conclusion: A ML-based model can be used for direct inference of the PK parameters from DCE image series. This method may allow fast and robust parameter inference in population DCE studies. Parameter inference on a 3D volume-time series takes only a few seconds on a GPU machine, which is significantly faster compared to conventional non-linear least squares fitting.

KW - contrast agent concentration convolutional neural networks dynamic contrast enhanced MRI ischaemic stroke loss function pharmacokinetic parameter inference tracer kinetic modeling

KW - Loss function

KW - Dynamic contrast enhanced MRI

KW - Tracer kinetic modeling

KW - Pharmacokinetic parameter inference

KW - Convolutional neural networks

KW - Contrast agent concentration

KW - Ischaemic stroke

KW - SMALL VESSEL DISEASE

KW - dynamic contrast enhanced MRI

KW - contrast agent concentration

KW - loss function

KW - ischaemic stroke

KW - pharmacokinetic parameter inference

KW - MODELS

KW - TRACER

KW - convolutional neural networks

KW - BLOOD

KW - tracer kinetic modeling

UR - http://www.scopus.com/inward/record.url?scp=85065397393&partnerID=8YFLogxK

UR - http://www.mendeley.com/research/parameters-application-stroke-dynamic-contrastenhanced-mri

U2 - 10.3389/fneur.2018.01147

DO - 10.3389/fneur.2018.01147

M3 - Article

VL - 9

JO - Frontiers in Neurology

JF - Frontiers in Neurology

SN - 1664-2295

M1 - 1147

ER -