Impact of late gadolinium enhancement image acquisition resolution on neural network based automatic scar segmentation

BACKGROUND

Automatic myocardial scar segmentation from late gadolinium enhancement (LGE) images using neural networks promises an alternative to time-consuming and observer-dependent semi-automatic approaches. However, alterations in data acquisition, reconstruction as well as post-processing may compromise network performance. The objective of the present work was to systematically assess network performance degradation due to a mismatch of point-spread function between training and testing data.

METHODS

Thirty-six high-resolution (0.7×0.7×2.0 mm3) LGE k-space datasets were acquired post-mortem in porcine models of myocardial infarction. The in-plane point-spread function and hence in-plane resolution Δx was retrospectively degraded using k-space lowpass filtering, while field-of-view and matrix size were kept constant. Manual segmentation of the left ventricle (LV) and healthy remote myocardium was performed to quantify location and area (% of myocardium) of scar by thresholding (≥ SD5 above remote). Three standard U-Nets were trained on training resolutions Δxtrain = 0.7, 1.2 and 1.7 mm to predict endo- and epicardial borders of LV myocardium and scar. The scar prediction of the three networks for varying test resolutions (Δxtest = 0.7 to 1.7 mm) was compared against the reference SD5 thresholding at 0.7 mm. Finally, a fourth network trained on a combination of resolutions (Δxtrain = 0.7 to 1.7 mm) was tested.

RESULTS

The prediction of relative scar areas showed the highest precision when the resolution of the test data was identical to or close to the resolution used during training. The median fractional scar errors and precisions (IQR) from networks trained and tested on the same resolution were 0.0 percentage points (p.p.) (1.24 - 1.45), and - 0.5 - 0.0 p.p. (2.00 - 3.25) for networks trained and tested on the most differing resolutions, respectively. Deploying the network trained on multiple resolutions resulted in reduced resolution dependency with median scar errors and IQRs of 0.0 p.p. (1.24 - 1.69) for all investigated test resolutions.

CONCLUSION

A mismatch of the imaging point-spread function between training and test data can lead to degradation of scar segmentation when using current U-Net architectures as demonstrated on LGE porcine myocardial infarction data. Training networks on multi-resolution data can alleviate the resolution dependency.

Considerations for hyperpolarized 13 C MR at reduced field: Comparing 1.5T versus 3T

PURPOSE

In contrast to conventional MR, signal-to-noise ratio (SNR) is not linearly dependent on field strength in hyperpolarized MR, as polarization is generated outside the MR system. Moreover, field inhomogeneity-induced artifacts and other practical limitations associated with field strengths ≥ $$ \ge $$ 3T are alleviated at lower fields. The potential of hyperpolarized   13 $$ {}^{13} $$ C spectroscopy and imaging at 1.5T versus 3T is demonstrated in silico, in vitro, and in vivo for applications on clinical MR systems.

THEORY AND METHODS

Theoretical noise and SNR behavior at different field strengths are investigated based on simulations. A thorough field comparison between 1.5T and 3T is performed using thermal and hyperpolarized   13 $$ {}^{13} $$ C spectroscopy and imaging. Cardiac in vivo data is obtained in pigs using hyperpolarized [1-   13 $$ {}^{13} $$ C]pyruvate spectroscopy and imaging at 1.5T and 3T.

RESULTS

Based on theoretical considerations and simulations, the SNR of hyperpolarized MR at identical acquisition bandwidths is independent of the field strength for typical coil setups, while adaptively changing the acquisition bandwidth proportional to the static magnetic field allows for net SNR gains of up to 40% at 1.5T compared to 3T. In vitro   13 $$ {}^{13} $$ C data verified these considerations with less than 7% deviation. In vivo feasibility of hyperpolarized [1-   13 $$ {}^{13} $$ C]pyruvate dynamic metabolic spectroscopy and imaging at 1.5T is demonstrated in the pig heart with comparable SNR between 1.5T and 3T while B   0 $$ {}_0 $$ artifacts are noticeably reduced at 1.5T.

CONCLUSION

Hyperpolarized   13 $$ {}^{13} $$ C MR at lower field strengths is favorable in terms of SNR and off-resonance effects, which makes 1.5T a promising alternative to 3T, especially for clinical cardiac metabolic imaging.

Free-breathing motion-informed locally low-rank quantitative 3D myocardial perfusion imaging

Purpose

To propose respiratory motion‐informed locally low‐rank reconstruction (MI‐LLR) for robust free‐breathing single‐bolus quantitative 3D myocardial perfusion CMR imaging. Simulation and in‐vivo results are compared to locally low‐rank (LLR) and compressed sensing reconstructions (CS) for reference.

Methods

Data were acquired using a 3D Cartesian pseudo‐spiral in‐out k‐t undersampling scheme (R = 10) and reconstructed using MI‐LLR, which encompasses two stages. In the first stage, approximate displacement fields are derived from an initial LLR reconstruction to feed a motion‐compensated reference system to a second reconstruction stage, which reduces the rank of the inverse problem. For comparison, data were also reconstructed with LLR and frame‐by‐frame CS using wavelets as sparsifying transform (ℓ1‐wavelet).

Reconstruction accuracy relative to ground truth was assessed using synthetic data for realistic ranges of breathing motion, heart rates, and SNRs. In‐vivo experiments were conducted in healthy subjects at rest and during adenosine stress. Myocardial blood flow (MBF) maps were derived using a Fermi model.

Results

Improved uniformity of MBF maps with reduced local variations was achieved with MI‐LLR. For rest and stress, intra‐volunteer variation of absolute and relative MBF was lower in MI‐LLR (±0.17 mL/g/min [26%] and ±1.07 mL/g/min [33%]) versus LLR (±0.19 mL/g/min [28%] and ±1.22 mL/g/min [36%]) and versus ℓ1‐wavelet (±1.17 mL/g/min [113%] and ±6.87 mL/g/min [115%]). At rest, intra‐subject MBF variation was reduced significantly with MI‐LLR.

Conclusion

The combination of pseudo‐spiral Cartesian undersampling and dual‐stage MI‐LLR reconstruction improves free‐breathing quantitative 3D myocardial perfusion CMR imaging under rest and stress condition.

Click here for author‐reader discussions

Self-assembly and spectroscopic fingerprints of photoactive pyrenyl tectons on hBN/Cu(111)

The controlled modification of electronic and photophysical properties of polycyclic aromatic hydrocarbons by chemical functionalization, adsorption on solid supports, and supramolecular organization is the key to optimize the application of these compounds in (opto)electronic devices. Here, we present a multimethod study comprehensively characterizing a family of pyridin-4-ylethynyl-functionalized pyrene derivatives in different environments. UV-vis measurements in toluene solutions revealed absorption at wavelengths consistent with density functional theory (DFT) calculations, while emission experiments showed a high fluorescence quantum yield. Scanning tunneling microscopy (STM) and spectroscopy (STS) measurements of the pyrene derivatives adsorbed on a Cu(111)-supported hexagonal boron nitride (hBN) decoupling layer provided access to spatially and energetically resolved molecular electronic states. We demonstrate that the pyrene electronic gap is reduced with an increasing number of substituents. Furthermore, we discuss the influence of template-induced gating and supramolecular organization on the energies of distinct molecular orbitals. The selection of the number and positioning of the pyridyl termini in tetrasubstituted, trans- and cis-like-disubstituted derivatives governed the self-assembly of the pyrenyl core on the nanostructured hBN support, affording dense-packed arrays and intricate porous networks featuring a kagome lattice.

Supramolecular Spangling, Crocheting, and Knitting of Functionalized Pyrene Molecules on a Silver Surface

Pyrenes, as photoactive polycyclic aromatic hydrocarbons (PAHs), represent promising modules for the bottom-up assembly of functional nanostructures. Here, we introduce the synthesis of a family of pyrene derivatives peripherally functionalized with pyridin-4-ylethynyl termini and comprehensively characterize their self-assembly abilities on a smooth Ag(111) support by scanning tunneling microscopy. By deliberate selection of number and geometric positioning of the pyridyl-terminated substituents, two-dimensional arrays, one-dimensional coordination chains, and chiral, porous kagomé-type networks can be tailored. A comparison to phenyl-functionalized reference pyrenes, not supporting the self-assembly of ordered structures at low coverage, highlights the role of the pyridyl moieties for supramolecular crocheting and knitting. Furthermore, we demonstrate the selective spangling of pores in the two-dimensional pyrene assemblies by a distinct number of iodine atoms as guests by atomically resolved imaging and complementary X-ray photoelectron spectroscopy.

Two-level spatial modulation of vibronic conductance in conjugated oligophenylenes on boron nitride

Intramolecular current-induced vibronic excitations are reported in highly ordered monolayers of quaterphenylene dicarbonitriles at an electronically patterned boron nitride on copper platform (BN/Cu(111)). A first level of spatially modulated conductance at the nanometer-scale is induced by the substrate. Moreover, a second level of conductance variations at the molecular level is found. Low temperature scanning tunneling microscopy studies in conjunction with molecular dynamics calculations reveal collective amplification of the molecule's interphenylene torsion angles in the monolayer. Librational modes influencing these torsion angles are identified as initial excitations during vibronic conductance. Density functional theory is used to map phenylene breathing modes and other vibrational excitations that are suggested to be at the origin of the submolecular features during vibronic conductance.

Complex behavioral strategy and reversal learning in the water maze without NMDA receptor-dependent long-term potentiation

Successful performance of the water maze task requires that rats learn complex behavioral strategies for swimming in a pool of water, searching for and interacting with a hidden platform before its spatial location can be learned. To evaluate whether NMDA receptor-dependent long-term potentiation (NMDA-LTP) is required for learning the required behavioral strategies, rats with NMDA-LTP blocked by systemic pharmacological treatment were trained in the behavioral strategies using simplified and stepwise training methods. Despite the blockade of NMDA-LTP in the dentate gyrus and hippocampal area CA1, rats learned the required behavioral strategies and used them to learn both initial and reversed platform locations. This is the first evaluation of the role of NMDA-LTP specifically in behavioral strategy learning. Although hippocampal NMDA-LTP might contribute to the water maze task, this form of LTP is not essential for learning complex behavioral strategies or multiple hidden platform locations.

Behavioural, physiological and morphological analysis of a line of apolipoprotein E knockout mouse

Using apolipoprotein E knockout mice derived from the Maeda source [Piedrahita J. A. et al. (1992) Proc. natn. Acad Sci. US.A. 89, 4471 4475], we have studied the influence of apolipoprotein E gene deletion on normal CNS function by neurological tests and water maze learning, hippocampal ultrastructure assessed by quantitative immunocytochemistry and electron microscopy, CNS plasticity, i.e. hippocampal long-term potentiation and amygdaloid kindling, and CNS repair, i.e. synaptic recovery in the hippocampus following deafferentation. In each study there was little difference between the apolipoprotein E knockout mice and wild-type controls of similar age and genetic background. Apolipoprotein E knockout mice aged eight months demonstrated accurate spatial learning and normal neurological function. Synaptophysin and microtubule-associated protein 2 immunohistochemistry and electron microscopic analysis of these animals revealed that the hippocampal synaptic and dendritic densities were similar between genotypes. The induction and maintenance of kindled seizures and hippocampal long-term potentiation were indistinguishable between groups. Finally, unilateral entorhinal cortex lesions produced a marked loss of hippocampal synaptophysin immunoreactivity in both groups and a marked up-regulation of apolipoprotein E in the wild-type group. Both apolipoprotein E knockout and wild-type groups showed immunohistochemical evidence of reactive synaptogenesis, although the apolipoprotein E knockout group may have initially shown greater synaptic loss. It is suggested that either apolipoprotein E is of no importance in the maintenance of synaptic integrity and in processes of CNS plasticity and repair, or more likely, alternative (apolipo)proteins may compensate for the loss of apolipoprotein E in the knockout animals.