Quantitative Analysis of DCE and DSC-MRI: From Kinetic Modeling to Deep Learning

Abstract

Background Perfusion MRI is a well-established imaging modality with a multitude of applications in oncological and cardiovascular imaging. Clinically used processing methods, while stable and robust, have remained largely unchanged in recent years. Despite promising results from novel methods, their relatively minimal improvement compared to established methods did not generally warrant significant changes to clinical perfusion processing.

Results and Conclusion Machine learning in general and deep learning in particular, which are currently revolutionizing computer-aided diagnosis, may carry the potential to change this situation and truly capture the potential of perfusion imaging. Recent advances in the training of recurrent neural networks make it possible to predict and classify time series data with high accuracy. Combining physics-based tissue models and deep learning, using either physics-informed neural networks or universal differential equations, simplifies the training process and increases the interpretability of the resulting models. Due to their versatility, these methods will potentially be useful in bridging the gap between microvascular architecture and perfusion parameters, akin to MR fingerprinting in structural MR imaging. Still, further research is urgently needed before these methods may be used in clinical practice.

Key Points:

• Machine learning offers promising methods for processing of perfusion data.

• Recurrent neural networks can classify time series with high accuracy.

• Data augmentation is essentially especially when using small datasets.

Citation Format

• Rotkopf LT, Zhang KS, Tavakoli AA et al. Quantitative Analysis of DCE and DSC-MRI: From Kinetic Modeling to Deep Learning. Fortschr Röntgenstr 2022; 194: 975 – 982

Zusammenfassung

Hintergrund Das Perfusions-MRT stellt eine etablierte Bildgebungsmodalität mit einer Vielzahl von Anwendungen in der onkologischen, neurovaskulären oder kardiovaskulären Bildgebung dar. Die mittlerweile klinisch etablierten quantitativen Auswertemethoden haben sich in den letzten Jahren jedoch wenig geändert. Dies liegt vorrangig an den geringen Verbesserungen, die durch neue Verfahren erzielt werden konnten.

Ergebnisse und Schlussfolgerung Machine Learning und Deep Learning, die derzeit den state of the art der computergestützten Diagnoseverfahren darstellen, haben das Potenzial die vielfältige Information, die in perfusionsgestützter Bildgebung akquiriert wird, vollständig besser als bisher zu erfassen und zu nutzen. Rekurrente neuronale Netze können nach entsprechendem Training Zeitserien mit hoher Genauigkeit vorhersagen und klassifizieren. Speziell die Kombination aus mikrostrukturellen Gewebsmodellen und Deep Learning mittels physics-informed neural networks oder universal differential equations vereinfacht kann das das Training der Modelle vereinfachen und die Interpretabilität verbessern. Aufgrund ihrer vielseitigen Anwendbarkeit ist es möglich, dass diese Methoden in der Lage sein werden, das Wechselspiel zwischen der mikrovaskulären Architektur und Perfusionsparametern besser zu modellieren. Weitere Forschung in diesem Gebiet ist dringend notwendig, um diese Methoden für die klinische Arbeit einsatzfähig zu machen.

Kernaussagen:

• Machine Learning-Methoden bieten vielversprechende Möglichkeiten zur Auswertung von Perfusionsdaten.

• Rekurrente neuronale Netze können Zeitserien mit hoher Genauigkeit klassifizieren.

• Data augmentation ist besonders bei kleinen Datensätzen essentiell.

Key words

MR-diffusion/perfusion - neural networks - vascular - staging - diagnostic radiology

Publication History

Received: 11 June 2021

Accepted: 25 January 2022

Article published online:
24 February 2022

• References

• 1 Payne SJ, El-Bouri WK. Modelling dynamic changes in blood flow and volume in the cerebral vasculature. NeuroImage 2018; 176: 124-137
• 2 Hirsch S, Reichold J, Schneider M. et al. Topology and Hemodynamics of the Cortical Cerebrovascular System. J Cereb Blood Flow Metab 2012; 32: 952-967
• 3 Peyrounette M, Davit Y, Quintard M. et al. Multiscale modelling of blood flow in cerebral microcirculation: Details at capillary scale control accuracy at the level of the cortex. PLOS ONE 2018; 13: e0189474
• 4 Lauwers F, Cassot F, Lauwers-Cances V. et al. Morphometry of the human cerebral cortex microcirculation: general characteristics and space-related profiles. Neuroimage 2008; 39: 936-948
• 5 Smith AF, Doyeux V, Berg M. et al. Brain Capillary Networks Across Species: A few Simple Organizational Requirements Are Sufficient to Reproduce Both Structure and Function. Front Physiol 2019; 10: 233
• 6 Park CS, Payne SJ. A generalized mathematical framework for estimating the residue function for arbitrary vascular networks. Interface Focus 2013; 3: 20120078
• 7 Rothenburg L, Matyas EL, Ambrus SZ. Statistical aspects of flow in a random network of channels. Stoch Hydrol Hydraul 1987; 1: 217-240
• 8 Hanahan D, Weinberg RA. The Hallmarks of Cancer. Cell 2000; 100: 57-70
• 9 Bergers G, Benjamin LE. Tumorigenesis and the angiogenic switch. Nat Rev Cancer 2003; 3: 401-410
• 10 Heye AK, Thrippleton MJ, Armitage PA. et al. Tracer kinetic modelling for DCE-MRI quantification of subtle blood–brain barrier permeability. NeuroImage 2016; 125: 446-455
• 11 Bammer R. MR and CT perfusion and pharmacokinetic imaging: clinical applications and theory. 2016
• 12 Boxerman JL, Schmainda KM, Weisskoff RM. for Contrast Agent Extravasation Significantly Correlate with Glioma Tumor Grade, Whereas. Ajnr Am J Neuroradiol 2006; 859-867
• 13 Troprès I, Grimault S, Vaeth A. et al. Vessel size imaging. Magn Reson Med 2001; 45: 397-408
• 14 Kiselev VG, Strecker R, Ziyeh S. et al. Vessel size imaging in humans. Magn Reson Med 2005; 53: 553-563
• 15 Zhang K, Yun SD, Triphan SMF. et al. Vessel architecture imaging using multiband gradient-echo/spin-echo EPI. PLoS ONE 2019; 14: 1-19
• 16 Maximov II, Farrher E, Grinberg F. et al. Spatially variable Rician noise in magnetic resonance imaging. Med Image Anal 2012; 16: 536-548
• 17 Stewart GN. Researches on the Circulation Time in Organs and on the Influences which affect it: Parts I.-III. J Physiol 1893; 15: 1-89
• 18 Meier P, Zierler KL. On the Theory of the Indicator-Dilution Method for Measurement of Blood Flow and Volume. J Appl Physiol 1954; 6: 731-744
• 19 Calamante F, Mørup M, Hansen LK. Defining a local arterial input function for perfusion MRI using independent component analysis: Defining a Local AIF for Perfusion MRI. Magn Reson Med 2004; 52: 789-797
• 20 Mouridsen K, Christensen S, Gyldensted L. et al. Automatic selection of arterial input function using cluster analysis. Magn Reson Med 2006; 55: 524-531
• 21 Peruzzo D, Bertoldo A, Zanderigo F. et al. Automatic selection of arterial input function on dynamic contrast-enhanced MR images. Comput Methods Programs Biomed 2011; 104: e148-e157
• 22 Huber T, Rotkopf L, Wiestler B. et al. Wavelet-based reconstruction of dynamic susceptibility MR-perfusion: a new method to visualize hypervascular brain tumors. Eur Radiol 2019; 29: 2669-2676
• 23 Rotkopf LT, Wiestler B, Preibisch C. et al. The wavelet power spectrum of perfusion weighted MRI correlates with tumor vascularity in biopsy-proven glioblastoma samples. PLOS ONE 2020; 15: e0228030
• 24 Mouridsen K, Friston K, Hjort N. et al. Bayesian estimation of cerebral perfusion using a physiological model of microvasculature. NeuroImage 2006; 33: 570-579
• 25 Mehndiratta A, MacIntosh BJ, Crane DE. et al. A control point interpolation method for the non-parametric quantification of cerebral haemodynamics from dynamic susceptibility contrast MRI. NeuroImage 2013; 64: 560-570
• 26 Chappell MA, Groves AR, Whitcher B. et al. Variational Bayesian Inference for a Nonlinear Forward Model. IEEE Trans Signal Process 2009; 57: 223-236
• 27 Tietze A, Nielsen A, Klærke Mikkelsen I. et al. Bayesian modeling of Dynamic Contrast Enhanced MRI data in cerebral glioma patients improves the diagnostic quality of hemodynamic parameter maps. PLOS ONE 2018; 13: e0202906
• 28 Tofts PS, Kermode AG. Measurement of the blood-brain barrier permeability and leakage space using dynamic MR imaging. 1. Fundamental concepts. Magn Reson Med 1991; 17: 357-367
• 29 Brix G, Kiessling F, Lucht R. et al. Microcirculation and microvasculature in breast tumors: Pharmacokinetic analysis of dynamic MR image series. Magn Reson Med 2004; 52: 420-429
• 30 Patlak CS, Blasberg RG. Graphical evaluation of blood-to-brain transfer constants from multiple-time uptake data. Generalizations. J Cereb Blood Flow Metab Off J Int Soc Cereb Blood Flow Metab 1985; 5: 584-590
• 31 Flouri D, Lesnic D, Sourbron SP. Fitting the two-compartment model in DCE-MRI by linear inversion: Fitting the Two-Compartment Model in DCE-MRI by Linear Inversion. Magn Reson Med 2016; 76: 998-1006
• 32 Kallehauge JF, Sourbron S, Irving B. et al. Comparison of linear and nonlinear implementation of the compartmental tissue uptake model for dynamic contrast-enhanced MRI: Linear Implementation of the Compartmental Tissue Uptake Model. Magn Reson Med 2017; 77: 2414-2423
• 33 St Lawrence KS, Lee TY. An adiabatic approximation to the tissue homogeneity model for water exchange in the brain: I. Theoretical derivation. J Cereb Blood Flow Metab Off J Int Soc Cereb Blood Flow Metab 1998; 18: 1365-1377
• 34 Cybenko G. Approximation by superpositions of a sigmoidal function. Math Control Signals Syst 1989; 2: 303-314
• 35 Kratsios A. The Universal Approximation Property: Characterization, Construction, Representation, and Existence. Ann. Math. Artif. Intell. 2021
• 36 Ulas C, Tetteh G, Thrippleton MJ. et al Direct Estimation of Pharmacokinetic Parameters from DCE-MRI using Deep CNN with Forward Physical Model Loss. ArXiv180402745 Cs 2018
• 37 Ulas C, Das D, Thrippleton MJ. et al. Convolutional Neural Networks for Direct Inference of Pharmacokinetic Parameters: Application to Stroke Dynamic Contrast-Enhanced MRI. Front Neurol 2019; 9: 1147
• 38 Nalepa J, Ribalta Lorenzo P, Marcinkiewicz M. et al. Fully-automated deep learning-powered system for DCE-MRI analysis of brain tumors. Artif Intell Med 2020; 102: 101769
• 39 Aggarwal CC. Neural Networks and Deep Learning: A Textbook. Cham: Springer International Publishing; 2018
• 40 Greff K, Srivastava RK, Koutník J. et al. LSTM: A Search Space Odyssey. IEEE Trans Neural Netw Learn Syst 2017; 28: 2222-2232
• 41 Sherstinsky A. Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) Network. Phys Nonlinear Phenom 2020; 404: 132306
• 42 Choi KS, Choi SH, Jeong B. Prediction of IDH genotype in gliomas with dynamic susceptibility contrast perfusion MR imaging using an explainable recurrent neural network. Neuro-Oncol 2019; 21: 1197-1209
• 43 Huynh B, Antropova N, Giger ML. Recurrent neural networks for breast lesion classification based on DCE-MRIs. In: Mori K, Petrick N. (eds.) Med. Imaging 2018 Comput.-Aided Diagn. Houston, United States: SPIE; 2018: 93
• 44 Iqbal S, Ghani Khan MU, Saba T. et al. Deep learning model integrating features and novel classifiers fusion for brain tumor segmentation. Microsc Res Tech 2019; 82: 1302-1315
• 45 Kang J, Samarasinghe G, Senanayake U. et al Deep Learning for Volumetric Segmentation in Spatio-Temporal Data: Application to Segmentation of Prostate in DCE-MRI. In: 2019 IEEE 16th Int. Symp. Biomed. Imaging ISBI 2019. Venice, Italy: IEEE, 2019: 61–65
• 46 Lang N, Zhang Y, Zhang E. et al. Differentiation of spinal metastases originated from lung and other cancers using radiomics and deep learning based on DCE-MRI. Magn Reson Imaging 2019; 64: 4-12
• 47 Zou J, Balter JM, Cao Y. Estimation of pharmacokinetic parameters from DCE‐MRI by extracting long and short time‐dependent features using an LSTM network. Med Phys 2020; 47: 3447-3457
• 48 Efron B, Tibshirani R. Bootstrap Methods for Standard Errors, Confidence Intervals, and Other Measures of Statistical Accuracy. Stat Sci 1986; 1: 54-75
• 49 Gal Y, Ghahramani Z. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. ArXiv150602142 Cs Stat 2016
• 50 McDermott P, Wikle C. Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data. Entropy 2019; 21: 184
• 51 Mirikitani DT, Nikolaev N. Recursive Bayesian Recurrent Neural Networks for Time-Series Modeling. IEEE Trans Neural Netw 2010; 21: 262-274
• 52 Kenesei T, Abonyi J. Interpretability of Computational Intelligence-Based Regression Models. Cham: Springer International Publishing; 2015
• 53 Ancona M, Öztireli C, Ceolini E. et al. Towards better understanding of gradient-based attribution methods for Deep Neural Networks. 2018: 16
• 54 Guo T, Lin T, Antulov-Fantulin N. Exploring Interpretable LSTM Neural Networks over Multi-Variable Data. ArXiv190512034 Cs Stat 2019
• 55 Weinreb JC, Barentsz JO, Choyke PL. et al. PI-RADS Prostate Imaging – Reporting and Data System: 2015, Version 2. Eur Urol 2016; 69: 16-40
• 56 Spak DA, Plaxco JS, Santiago L. et al. BI-RADS ® fifth edition: A summary of changes. Diagn Interv Imaging 2017; 98: 179-190
• 57 Bonekamp D, Macura KJ. Dynamic Contrast-Enhanced Magnetic Resonance Imaging in the Evaluation of the Prostate. Top Magn Reson Imaging 2008; 19: 273-284
• 58 Raissi M, Perdikaris P, Karniadakis GE. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J Comput Phys 2019; 378: 686-707
• 59 Chen RTQ, Rubanova Y, Bettencourt J. et al Neural Ordinary Differential Equations. ArXiv180607366 Cs Stat 2019
• 60 Rackauckas C, Ma Y, Martensen J. et al Universal Differential Equations for Scientific Machine Learning. ArXiv200104385 Cs Math Q-Bio Stat 2020