Compressed Sensing Real-Time Cine Reduces CMR Arrhythmia-Related Artifacts

Background and objective: Cardiac magnetic resonance (CMR) is a key tool for cardiac work-up. However, arrhythmia can be responsible for arrhythmia-related artifacts (ARA) and increased scan time using segmented sequences. The aim of this study is to evaluate the effect of cardiac arrhythmia on image quality in a comparison of a compressed sensing real-time (CSrt) cine sequence with the reference prospectively gated segmented balanced steady-state free precession (Cineref) technique regarding ARA. Methods: A total of 71 consecutive adult patients (41 males; mean age = 59.5 ± 20.1 years (95% CI: 54.7–64.2 years)) referred for CMR examination with concomitant irregular heart rate (defined by an RR interval coefficient of variation >10%) during scanning were prospectively enrolled. For each patient, two cine sequences were systematically acquired: first, the reference prospectively triggered multi-breath-hold Cineref sequence including a short-axis stack, one four-chamber slice, and a couple of two-chamber slices; second, an additional single breath-hold CSrt sequence providing the same slices as the reference technique. Two radiologists independently assessed ARA and image quality (overall, acquisition, and edge sharpness) for both techniques. Results: The mean heart rate was 71.8 ± 19.0 (SD) beat per minute (bpm) (95% CI: 67.4–76.3 bpm) and its coefficient of variation was 25.0 ± 9.4 (SD) % (95% CI: 22.8–27.2%). Acquisition was significantly faster with CSrt than with Cineref (Cineref: 556.7 ± 145.4 (SD) s (95% CI: 496.7–616.7 s); CSrt: 23.9 ± 7.9 (SD) s (95% CI: 20.6–27.1 s); p < 0.0001). A total of 599 pairs of cine slices were evaluated (median: 8 (range: 6–14) slices per patient). The mean proportion of ARA-impaired slices per patient was 85.9 ± 22.7 (SD) % using Cineref, but this was figure was zero using CSrt (p < 0.0001). The European CMR registry artifact score was lower with CSrt (median: 1 (range: 0–5)) than with Cineref (median: 3 (range: 0–3); p < 0.0001). Subjective image quality was higher in CSrt than in Cineref (median: 3 (range: 1–3) versus 2 (range: 1–4), respectively; p < 0.0001). In line, edge sharpness was higher on CSrt cine than on Cineref images (0.054 ± 0.016 pixel−1 (95% CI: 0.050–0.057 pixel−1) versus 0.042 ± 0.022 pixel−1 (95% CI: 0.037–0.047 pixel−1), respectively; p = 0.0001). Conclusion: Compressed sensing real-time cine drastically reduces arrhythmia-related artifacts and thus improves cine image quality in patients with arrhythmia.

Application of a machine learning contouring tool for the evaluation of left ventricular strain in clinical practice

Funding Acknowledgements Type of funding sources: Private grant(s) and/or Sponsorship. Main funding source(s): AK has been funded by the Egyptian cultural centre and educational bureau of the Egyptian embassy in London and the Ministry of higher education in Egypt. SEP acknowledges support from the “SmartHeart” EPSRC programme grant (www.nihr.ac.uk; EP/P001009/1) and the London Medical Imaging and AI Centre for Value-Based Healthcare. This new centre is one of the UK Centres supported by a £50m investment from the Data to Early Diagnosis and Precision Medicine strand of the government’s Industrial Strategy Challenge Fund, managed and delivered by UK Research and Innovation (UKRI). SEP acknowledges support from the CAP-AI programme, London’s first AI enabling programme focused on stimulating growth in the capital’s AI Sector. CAP-AI is led by Capital Enterprise in partnership with Barts Health NHS Trust and Digital Catapult and is funded by the European Regional Development Fund and Barts Charity. SEP also acts as a paid consultant to Circle Cardiovascular Imaging Inc., Calgary, Canada and Servier onbehalf Barts Heart Centre, St Bartholomew’s Hospital, Barts Health NHS Trust, West Smithfield, London, UK Background Manual contouring of cardiovascular magnetic resonance (CMR) cine images remains common practice and the reference standard for left ventricular (LV) volumes and mass evaluation. However, it is time-consuming and machine learning (ML) may significantly reduce the time required for contouring. Accurate LV contours are the basis for reliable LV strain analysis using tissue tracking. Purpose To assess the impact of a ML contouring tool alone versus expert adjusted contours on LV strain. Methods We retrospectively selected 402 CMR studies with diagnoses of myocardial infarction (n = 108), myocarditis (n = 130) and healthy controls (n = 164) from the Barts BioResource between January 2015 to June 2018. CMR examinations were obtained using 1.5T and 3T scanners (Siemens Healthineers, Germany). We excluded 32 cases due to phase inconsistency between short (SAX) and long axes (LAX) cine images or suboptimal cine image quality. For the remaining 370 cases, steady state free precession cine images for LAX and SAX were analysed by the ML contouring tool (using CVI42 research prototype software 5.11). Manual expert adjustment for the contours was done for each case if considered suboptimal for strain analysis in the reference end-diastolic phase. Strain results from ML and expert adjusted ML methods were compared for strain agreement. Times taken by these methods were recorded and compared against the time taken for standard manual contouring. Results SAX and LAX derived strains by ML and expert adjusted ML methods showed good agreement by Bland-Altman analysis (Figure 1) with excellent coefficient of concordance using Kendall W which is 0.98 for global SAX, radial and circumferential strains (mean difference(MD) = -1.7% (lower and upper limits of agreement (UL,LL) -6.6,3.2), MD = 0.5% (-1.0,2.1)) and is 0.95 for global LAX derived strain (radial and longitudinal, MD = 0.7% (UL,LL -8.7 ,7.4),MD= 0.2% (-1.9,2.5), respectively). Time taken for adjustment of ML contours was significantly shorter than manual contouring (1.35 minutes vs 8.0 minutes, around 590% time saving in ML adjusted method). Conclusions ML contouring compared to expert manual adjustment has a clinically reasonable agreement when used for measuring LV strain. Also, using the ML tool with expert adjustment shows significant time saving for analysis and reporting time compared to entirely manual analysis, favouring its application in routine clinical practice. Abstract Figure.

A Deep Learning Based Cardiac Cine Segmentation Framework for Clinicians – Transfer Learning Application to 7T

BackgroundArtificial neural networks have shown promising performance in automatic segmentation of cardiac magnetic resonance imaging. However, initial training of such networks requires large amounts of annotated data and generalization to different vendors, field strengths, sequence parameters, and pathologies is often limited. Transfer learning has been proposed to address this challenge, but specific recommendations on the type and amount of data required is lacking. In this study we aim to assess data requirements for transfer learning to cardiac 7T in humans where the segmentation task can be challenging. In addition, we provide guidelines, tools, and annotated data to enable transfer learning approaches of other researchers and clinicians.MethodsA publicly available model for bi-ventricular segmentation is used to annotate a publicly available data set. This labelled data set is subsequently used to train a neural network for segmentation of left ventricular and myocardial contours in cardiac cine MRI. The network is used as starting point for transfer learning to the segmentation task on 7T cine data of healthy volunteers (n=22, 7873 images). Structured and random data subsets of different sizes were used to systematically assess data requirements for successful transfer learning.ResultsInconsistencies in the publically available data set were corrected, labels created, and a neural network trained. On 7T cardiac cine images the initial model achieved DICELV=0.835 and DICEMY=0.670. Transfer learning using 7T cine data and ImageNet weight initialization significantly (p<10−3) improved model performance to DICELV=0.900 and DICEMY=0.791. Using only end-systolic and end-diastolic images reduced training data by 90%, with no negative impact on segmentation performance (DICELV=0.908, DICEMY=0.805).ConclusionsThis work demonstrates the benefits of transfer learning for cardiac cine image segmentation on a quantitative basis. We also make data, models and code publicly available, while providing practical guidelines for researchers planning transfer learning projects in cardiac MRI.

Low-Rank and Sparse Decomposition Model for Accelerating Dynamic MRI Reconstruction

The reconstruction of dynamic magnetic resonance imaging (dMRI) from partially sampled k-space data has to deal with a trade-off between the spatial resolution and temporal resolution. In this paper, a low-rank and sparse decomposition model is introduced to resolve this issue, which is formulated as an inverse problem regularized by robust principal component analysis (RPCA). The inverse problem can be solved by convex optimization method. We propose a scalable and fast algorithm based on the inexact augmented Lagrange multipliers (IALM) to carry out the convex optimization. The experimental results demonstrate that our proposed algorithm can achieve superior reconstruction quality and faster reconstruction speed in cardiac cine image compared to existing state-of-art reconstruction methods.