AIFNet: Automatic Vascular Function Estimation for Perfusion Analysis Using Deep Learning

Perfusion imaging is crucial in acute ischemic stroke for quantifying the salvageable penumbra and irreversibly damaged core lesions. As such, it helps clinicians to decide on the optimal reperfusion treatment. In perfusion CT imaging, deconvolution methods are used to obtain clinically interpretable perfusion parameters that allow identifying brain tissue abnormalities. Deconvolution methods require the selection of two reference vascular functions as inputs to the model: the arterial input function (AIF) and the venous output function, with the AIF as the most critical model input. When manually performed, the vascular function selection is time demanding, suffers from poor reproducibility and is subject to the professionals' experience. This leads to potentially unreliable quantification of the penumbra and core lesions and, hence, might harm the treatment decision process. In this work we automatize the perfusion analysis with AIFNet, a fully automatic and end-to-end trainable deep learning approach for estimating the vascular functions. Unlike previous methods using clustering or segmentation techniques to select vascular voxels, AIFNet is directly optimized at the vascular function estimation, which allows to better recognise the time-curve profiles. Validation on the public ISLES18 stroke database shows that AIFNet reaches inter-rater performance for the vascular function estimation and, subsequently, for the parameter maps and core lesion quantification obtained through deconvolution. We conclude that AIFNet has potential for clinical transfer and could be incorporated in perfusion deconvolution software.

[1]  R. Meier,et al.  Neural Network-derived Perfusion Maps for the Assessment of Lesions in Patients with Acute Ischemic Stroke. , 2019, Radiology. Artificial intelligence.

[2]  Peter Langhorne,et al.  Organised inpatient (stroke unit) care for stroke. , 2007, The Cochrane database of systematic reviews.

[3]  Mike E. Davies,et al.  Direct Estimation of Pharmacokinetic Parameters from DCE-MRI using Deep CNN with Forward Physical Model Loss , 2018, MICCAI.

[4]  B. Campbell,et al.  Imaging selection for acute stroke intervention , 2018, International journal of stroke : official journal of the International Stroke Society.

[5]  T-Y Lee,et al.  Serial changes in CT cerebral blood volume and flow after 4 hours of middle cerebral occlusion in an animal model of embolic cerebral ischemia. , 2007, AJNR. American journal of neuroradiology.

[6]  F. Calamante Arterial input function in perfusion MRI: a comprehensive review. , 2013, Progress in nuclear magnetic resonance spectroscopy.

[7]  Joanna M. Wardlaw,et al.  Convolutional Neural Networks for Direct Inference of Pharmacokinetic Parameters: Application to Stroke Dynamic Contrast-Enhanced MRI , 2019, Front. Neurol..

[8]  Michael Kistler,et al.  The Virtual Skeleton Database: An Open Access Repository for Biomedical Research and Collaboration , 2013, Journal of medical Internet research.

[9]  R. Bammer,et al.  The Infarct Core is Well Represented by the Acute Diffusion Lesion: Sustained Reversal is Infrequent , 2012, Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism.

[10]  L. K. Hansen,et al.  Defining a local arterial input function for perfusion MRI using independent component analysis , 2004, Magnetic resonance in medicine.

[11]  Lin Shi,et al.  Automatic detection of arterial input function in dynamic contrast enhanced MRI based on affinity propagation clustering , 2014, Journal of magnetic resonance imaging : JMRI.

[12]  M. Guanci,et al.  Acute Ischemic Stroke Review , 2007, The Journal of neuroscience nursing : journal of the American Association of Neuroscience Nurses.

[13]  Lippincott Williams Wilkins,et al.  Stroke--1989. Recommendations on stroke prevention, diagnosis, and therapy. Report of the WHO Task Force on Stroke and other Cerebrovascular Disorders. , 1989, Stroke.

[14]  Leif Østergaard,et al.  Effects of tracer arrival time on flow estimates in MR perfusion‐weighted imaging , 2003, Magnetic resonance in medicine.

[15]  Hiroki Shirato,et al.  Differences in CT perfusion maps generated by different commercial software: quantitative analysis by using identical source data of acute stroke patients. , 2010, Radiology.

[16]  Michael D Hill,et al.  Time to angiographic reperfusion and clinical outcome after acute ischaemic stroke: an analysis of data from the Interventional Management of Stroke (IMS III) phase 3 trial , 2014, The Lancet Neurology.

[17]  Matus Straka,et al.  A benchmarking tool to evaluate computer tomography perfusion infarct core predictions against a DWI standard , 2016, Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism.

[18]  Jitendra Malik,et al.  Normalized cuts and image segmentation , 1997, Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[19]  et al.,et al.  ISLES 2015 ‐ A public evaluation benchmark for ischemic stroke lesion segmentation from multispectral MRI , 2017, Medical Image Anal..

[20]  Søren Christensen,et al.  Automatic selection of arterial input function using cluster analysis , 2006, Magnetic resonance in medicine.

[21]  B. Rosen,et al.  Tracer arrival timing‐insensitive technique for estimating flow in MR perfusion‐weighted imaging using singular value decomposition with a block‐circulant deconvolution matrix , 2003, Magnetic resonance in medicine.

[22]  M. Moseley,et al.  Automated method for generating the arterial input function on perfusion-weighted MR imaging: validation in patients with stroke. , 2005, AJNR. American journal of neuroradiology.

[23]  Paul Suetens,et al.  Prediction of final infarct volume from native CT perfusion and treatment parameters using deep learning , 2018, Medical Image Anal..

[24]  Roland Bammer,et al.  Ischemic core and hypoperfusion volumes predict infarct size in SWIFT PRIME , 2016, Annals of neurology.

[25]  Paul Suetens,et al.  Perfusion parameter estimation using neural networks and data augmentation , 2018, BrainLes@MICCAI.

[26]  Konstantinos Kamnitsas,et al.  Efficient multi‐scale 3D CNN with fully connected CRF for accurate brain lesion segmentation , 2016, Medical Image Anal..

[27]  K Scheffler,et al.  Analysis of input functions from different arterial branches with gamma variate functions and cluster analysis for quantitative blood volume measurements. , 2000, Magnetic resonance imaging.

[28]  B. Rosen,et al.  High resolution measurement of cerebral blood flow using intravascular tracer bolus passages. Part I: Mathematical approach and statistical analysis , 1996, Magnetic resonance in medicine.

[29]  M. Reiser,et al.  Deconvolution of bolus-tracking data: a comparison of discretization methods , 2007, Physics in medicine and biology.

[30]  D. DeLong,et al.  User-defined vascular input function curves: influence on mean perfusion parameter values and signal-to-noise ratio. , 2004, Radiology.

[31]  W. J. Lorenz,et al.  Quantification of regional cerebral blood flow and volume with dynamic susceptibility contrast-enhanced MR imaging. , 1994, Radiology.

[32]  Arnau Oliver,et al.  Acute ischemic stroke lesion core segmentation in CT perfusion images using fully convolutional neural networks , 2019, Comput. Biol. Medicine.

[33]  D. Gadian,et al.  Delay and dispersion effects in dynamic susceptibility contrast MRI: Simulations using singular value decomposition , 2000, Magnetic resonance in medicine.

[34]  Alessandra Bertoldo,et al.  Automatic selection of arterial input function on dynamic contrast-enhanced MR images , 2011, Comput. Methods Programs Biomed..

[35]  R. Fahed,et al.  Reader response: Automated CT perfusion imaging for acute ischemic stroke: Pearls and pitfalls for real-world use , 2020 .

[36]  Qiyong Guo,et al.  Automated detection of the arterial input function using normalized cut clustering to determine cerebral perfusion by dynamic susceptibility contrast‐magnetic resonance imaging , 2015, Journal of magnetic resonance imaging : JMRI.

[37]  Paul Suetens,et al.  Contra-Lateral Information CNN for Core Lesion Segmentation Based on Native CTP in Acute Stroke , 2018, BrainLes@MICCAI.

[38]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[39]  T-Y Lee,et al.  Theoretic Basis and Technical Implementations of CT Perfusion in Acute Ischemic Stroke, Part 1: Theoretic Basis , 2009, American Journal of Neuroradiology.

[40]  Jonathan Rubin,et al.  Ischemic Stroke Lesion Segmentation in CT Perfusion Scans using Pyramid Pooling and Focal Loss , 2018, BrainLes@MICCAI.

[41]  Tao Song,et al.  Integrated Extractor, Generator and Segmentor for Ischemic Stroke Lesion Segmentation , 2018, BrainLes@MICCAI.

[42]  E Klotz,et al.  Perfusion measurements of the brain: using dynamic CT for the quantitative assessment of cerebral ischemia in acute stroke. , 1999, European journal of radiology.

[43]  K. Murase,et al.  Determination of arterial input function using fuzzy clustering for quantification of cerebral blood flow with dynamic susceptibility contrast‐enhanced MR imaging , 2001, Journal of magnetic resonance imaging : JMRI.

[44]  David Robben Image-based Quantification of Cerebral Vascular Connectivity , 2016 .

[45]  Shaoting Zhang,et al.  Automatic Ischemic Stroke Lesion Segmentation from Computed Tomography Perfusion Images by Image Synthesis and Attention-Based Deep Neural Networks , 2020, Medical Image Anal..

[46]  Rebecca Fahrig,et al.  Deconvolution-Based CT and MR Brain Perfusion Measurement: Theoretical Model Revisited and Practical Implementation Details , 2011, Int. J. Biomed. Imaging.

[47]  Mark W Parsons,et al.  Whole-Brain CT Perfusion to Quantify Acute Ischemic Penumbra and Core. , 2016, Radiology.

[48]  B. Rosen,et al.  High resolution measurement of cerebral blood flow using intravascular tracer bolus passages. Part II: Experimental comparison and preliminary results , 1996, Magnetic resonance in medicine.

[49]  Roland Bammer,et al.  Influence of Arterial Input Function on Hypoperfusion Volumes Measured With Perfusion-Weighted Imaging , 2003, Stroke.

[50]  Fabien Scalzo,et al.  A Machine Learning Approach to Perfusion Imaging With Dynamic Susceptibility Contrast MR , 2018, Front. Neurol..

[51]  Qi Yang,et al.  An Automatic Estimation of Arterial Input Function Based on Multi-Stream 3D CNN , 2019, Front. Neuroinform..

[52]  Richard Frayne,et al.  PerfTool: A software platform for investigating bolus‐tracking perfusion imaging quantification strategies , 2007, Journal of magnetic resonance imaging : JMRI.

[53]  Jens Fiehler,et al.  Automatic arterial input function selection in CT and MR perfusion datasets using deep convolutional neural networks. , 2020, Medical physics.

[54]  A. Sorensen,et al.  Automated perfusion‐weighted MRI using localized arterial input functions , 2006, Journal of magnetic resonance imaging : JMRI.