Imperial College London


Faculty of MedicineNational Heart & Lung Institute

Professor of Cardiology



+44 (0)20 7351 8810d.pennell




CMR UnitRoyal BromptonRoyal Brompton Campus






BibTex format

author = {Ferreira, PF and Martin, RR and Scott, AD and Khalique, Z and Yang, G and Nielles-Vallespin, S and Pennell, DJ and Firmin, DN},
doi = {10.1002/mrm.28294},
journal = {Magnetic Resonance in Medicine},
pages = {2801--2814},
title = {Automating in vivo cardiac diffusion tensor postprocessing with deep learning-based segmentation},
url = {},
volume = {84},
year = {2020}

RIS format (EndNote, RefMan)

AB - PurposeIn this work we develop and validate a fully automated postprocessing framework for in vivo diffusion tensor cardiac magnetic resonance (DTCMR) data powered by deep learning.MethodsA UNet based convolutional neural network was developed and trained to segment the heart in shortaxis DTCMR images. This was used as the basis to automate and enhance several stages of the DTCMR tensor calculation workflow, including image registration and removal of data corrupted with artifacts, and to segment the left ventricle. Previously collected and analyzed scans (348 healthy scans and 144 cardiomyopathy patient scans) were used to train and validate the UNet. All data were acquired at 3 T with a STEAMEPI sequence. The DTCMR postprocessing and UNet training/testing were performed with MATLAB and Python TensorFlow, respectively.ResultsThe UNet achieved a median Dice coefficient of 0.93 [0.92, 0.94] for the segmentation of the leftventricular myocardial region. The image registration of diffusion images improved with the UNet segmentation (P < .0001), and the identification of corrupted images achieved an F1 score of 0.70 when compared with an experienced user. Finally, the resulting tensor measures showed good agreement between an experienced user and the fully automated method.ConclusionThe trained UNet successfully automated the DTCMR postprocessing, supporting realtime results and reducing human workload. The automatic segmentation of the heart improved image registration, resulting in improvements of the calculated DT parameters.
AU - Ferreira,PF
AU - Martin,RR
AU - Scott,AD
AU - Khalique,Z
AU - Yang,G
AU - Nielles-Vallespin,S
AU - Pennell,DJ
AU - Firmin,DN
DO - 10.1002/mrm.28294
EP - 2814
PY - 2020///
SN - 0740-3194
SP - 2801
TI - Automating in vivo cardiac diffusion tensor postprocessing with deep learning-based segmentation
T2 - Magnetic Resonance in Medicine
UR -
UR -
UR -
UR -
VL - 84
ER -