Deep learning with attention supervision for automated motion artefact detection in quality control of cardiac T1-mapping

Highlights • A multi-stream CNN ResNet classifier customised for CMR T1-mapping motion artefact detection.• An attention supervision module to guide the training of CNN classifier.• A multiple human observer analysis of the scoring results to adjudicate human and machine performance.

[1]  M. Cerqueira,et al.  Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart. A statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. , 2002, Journal of nuclear cardiology : official publication of the American Society of Nuclear Cardiology.

[2]  Nikos Komodakis,et al.  Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer , 2016, ICLR.

[3]  S. Piechnik,et al.  State-of-the-art review: stress T1 mapping—technical considerations, pitfalls and emerging clinical applications , 2017, Magnetic Resonance Materials in Physics, Biology and Medicine.

[4]  Bolei Zhou,et al.  Learning Deep Features for Discriminative Localization , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[5]  Stefan K. Piechnik,et al.  Measurement of myocardial native T1 in cardiovascular diseases and norm in 1291 subjects , 2017, Journal of Cardiovascular Magnetic Resonance.

[6]  David M Higgins,et al.  Modified Look‐Locker inversion recovery (MOLLI) for high‐resolution T1 mapping of the heart , 2004, Magnetic resonance in medicine.

[7]  Lin Yang,et al.  A New Ensemble Learning Framework for 3D Biomedical Image Segmentation , 2018, AAAI.

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

[9]  Teng-Yi Huang,et al.  Automatic regional analysis of myocardial native T1 values: left ventricle segmentation and AHA parcellations , 2017, The International Journal of Cardiovascular Imaging.

[10]  Thoralf Niendorf,et al.  Myocardial T1 and T2 mapping at 3 T: reference values, influencing factors and implications , 2013, Journal of Cardiovascular Magnetic Resonance.

[11]  M. Cerqueira,et al.  Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart: A statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association , 2002, The international journal of cardiovascular imaging.

[12]  M. Oudkerk,et al.  Caffeine intake inverts the effect of adenosine on myocardial perfusion during stress as measured by T1 mapping , 2016, The International Journal of Cardiovascular Imaging.

[13]  M. Robson,et al.  Noncontrast T1 mapping for the diagnosis of cardiac amyloidosis. , 2013, JACC. Cardiovascular imaging.

[14]  M. Cerqueira,et al.  Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart. A statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. , 2002, Circulation.

[15]  Richard B. Thompson,et al.  Clinical recommendations for cardiovascular magnetic resonance mapping of T1, T2, T2* and extracellular volume: A consensus statement by the Society for Cardiovascular Magnetic Resonance (SCMR) endorsed by the European Association for Cardiovascular Imaging (EACVI) , 2017, Journal of Cardiovascular Magnetic Resonance.

[16]  Stefan Neubauer,et al.  Shortened Modified Look-Locker Inversion recovery (ShMOLLI) for clinical myocardial T1-mapping at 1.5 and 3 T within a 9 heartbeat breathhold , 2010, Journal of cardiovascular magnetic resonance : official journal of the Society for Cardiovascular Magnetic Resonance.

[17]  N. Geller,et al.  Hypertrophic Cardiomyopathy Registry: The rationale and design of an international, observational study of hypertrophic cardiomyopathy. , 2015, American heart journal.

[18]  Abhishek Das,et al.  Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization , 2016, 2017 IEEE International Conference on Computer Vision (ICCV).

[19]  Ben Glocker,et al.  Learning-Based Quality Control for Cardiac MR Images , 2018, IEEE Transactions on Medical Imaging.

[20]  S. Piechnik,et al.  CMR Parametric Mapping as a Tool for Myocardial Tissue Characterization , 2020, Korean circulation journal.

[21]  S. Piechnik,et al.  Myocardial T1 mapping and extracellular volume quantification: an overview of technical and biological confounders , 2017, The International Journal of Cardiovascular Imaging.

[22]  Peter Kellman,et al.  Influence of Off-resonance in myocardial T1-mapping using SSFP based MOLLI method , 2013, Journal of Cardiovascular Magnetic Resonance.

[23]  S. K. White,et al.  Noncontrast myocardial T1 mapping using cardiovascular magnetic resonance for iron overload , 2015, Journal of magnetic resonance imaging : JMRI.

[24]  M. Robson,et al.  Native T1-mapping detects the location, extent and patterns of acute myocarditis without the need for gadolinium contrast agents , 2014, Journal of Cardiovascular Magnetic Resonance.

[25]  Ben Glocker,et al.  Automated cardiovascular magnetic resonance image analysis with fully convolutional networks , 2017, Journal of Cardiovascular Magnetic Resonance.

[26]  S. K. White,et al.  Identification and Assessment of Anderson-Fabry Disease by Cardiovascular Magnetic Resonance Noncontrast Myocardial T1 Mapping , 2013, Circulation. Cardiovascular imaging.

[27]  S. Piechnik,et al.  Standardized image post-processing of cardiovascular magnetic resonance T1-mapping reduces variability and improves accuracy and consistency in myocardial tissue characterization. , 2020, International journal of cardiology.

[28]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[29]  Sergey Ioffe,et al.  Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning , 2016, AAAI.

[30]  David M Higgins,et al.  Human myocardium: single-breath-hold MR T1 mapping with high spatial resolution--reproducibility study. , 2006, Radiology.

[31]  M. Robson,et al.  Non-contrast T1-mapping detects acute myocardial edema with high diagnostic accuracy: a comparison to T2-weighted cardiovascular magnetic resonance , 2012, Journal of Cardiovascular Magnetic Resonance.

[32]  J. Francis,et al.  Adenosine stress native T1 mapping in severe aortic stenosis: evidence for a role of the intravascular compartment on myocardial T1 values , 2014, Journal of Cardiovascular Magnetic Resonance.

[33]  Marie-Pierre Jolly,et al.  Motion correction for myocardial T1 mapping using image registration with synthetic image estimation , 2012, Magnetic resonance in medicine.

[34]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[35]  J. Schulz-Menger,et al.  T1 mapping in patients with acute myocardial infarction. , 2003, Journal of cardiovascular magnetic resonance : official journal of the Society for Cardiovascular Magnetic Resonance.

[36]  Konrad Werys,et al.  Quality Control-Driven Image Segmentation Towards Reliable Automatic Image Analysis in Large-Scale Cardiovascular Magnetic Resonance Aortic Cine Imaging , 2019, MICCAI.

[37]  Philip H. S. Torr,et al.  Learn To Pay Attention , 2018, ICLR.

[38]  Alejandro F. Frangi,et al.  Automatic initialization and quality control of large‐scale cardiac MRI segmentations , 2018, Medical Image Anal..

[39]  Andrew Zisserman,et al.  Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps , 2013, ICLR.

[40]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[41]  Yun Fu,et al.  Tell Me Where to Look: Guided Attention Inference Network , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[42]  P. Kellman,et al.  T1-mapping in the heart: accuracy and precision , 2014, Journal of Cardiovascular Magnetic Resonance.