Constrained estimation of the arterial input function for myocardial perfusion cardiovascular magnetic resonance

Arterial Input Function (AIF) Correction Using AIF Plus Tissue Inputs with a Bi-LSTM Network

Background: The arterial input function (AIF) is vital for myocardial blood flow quantification in cardiac MRI to indicate the input time-concentration curve of a contrast agent. Inaccurate AIFs can significantly affect perfusion quantification. Purpose: When only saturated and biased AIFs are measured, this work investigates multiple ways of leveraging tissue curve information, including using AIF + tissue curves as inputs and optimizing the loss function for deep neural network training. Methods: Simulated data were generated using a 12-parameter AIF mathematical model for the AIF. Tissue curves were created from true AIFs combined with compartment-model parameters from a random distribution. Using Bloch simulations, a dictionary was constructed for a saturation-recovery 3D radial stack-of-stars sequence, accounting for deviations such as flip angle, T2* effects, and residual longitudinal magnetization after the saturation. A preliminary simulation study established the optimal tissue curve number using a bidirectional long short-term memory (Bi-LSTM) network with just AIF loss. Further optimization of the loss function involves comparing just AIF loss, AIF with compartment-model-based parameter loss, and AIF with compartment-model tissue loss. The optimized network was examined with both simulation and hybrid data, which included in vivo 3D stack-of-star datasets for testing. The AIF peak value accuracy and ktrans results were assessed. Results: Increasing the number of tissue curves can be beneficial when added tissue curves can provide extra information. Using just the AIF loss outperforms the other two proposed losses, including adding either a compartment-model-based tissue loss or a compartment-model parameter loss to the AIF loss. With the simulated data, the Bi-LSTM network reduced the AIF peak error from -23.6 ± 24.4% of the AIF using the dictionary method to 0.2 ± 7.2% (AIF input only) and 0.3 ± 2.5% (AIF + ten tissue curve inputs) of the network AIF. The corresponding ktrans error was reduced from -13.5 ± 8.8% to -0.6 ± 6.6% and 0.3 ± 2.1%. With the hybrid data (simulated data for training; in vivo data for testing), the AIF peak error was 15.0 ± 5.3% and the corresponding ktrans error was 20.7 ± 11.6% for the AIF using the dictionary method. The hybrid data revealed that using the AIF + tissue inputs reduced errors, with peak error (1.3 ± 11.1%) and ktrans error (-2.4 ± 6.7%). Conclusions: Integrating tissue curves with AIF curves into network inputs improves the precision of AI-driven AIF corrections. This result was seen both with simulated data and with applying the network trained only on simulated data to a limited in vivo test dataset.

A simplified method to correct saturation of arterial input function for cardiac magnetic resonance first-pass perfusion imaging: validation with simultaneously acquired PET

BACKGROUND

First-pass perfusion imaging in magnetic resonance imaging (MRI) is an established method to measure myocardial blood flow (MBF). An obstacle for accurate quantification of MBF is the saturation of blood pool signal intensity used for arterial input function (AIF). The objective of this project was to validate a new simplified method for AIF estimation obtained from single-bolus and single sequence perfusion measurements. The reference MBF was measured simultaneously on 13N-ammonia positron emission tomography (PET).

METHODS

Sixteen patients with clinically confirmed myocardial ischemia were imaged in a clinical whole-body PET-MRI system. PET perfusion imaging was performed in a 10-min acquisition after the injection of 10 mCi of 13N-ammonia. The MRI perfusion acquisition started simultaneously with the start of the PET acquisition after the injection of a 0.075 mmol/kg gadolinium contrast agent. Cardiac stress imaging was initiated after the administration of regadenoson 20 s prior to PET-MRI scanning. The saturation part of the MRI AIF data was modeled as a gamma variate curve, which was then estimated for a true AIF by minimizing a cost function according to various boundary conditions. A standard AHA 16-segment model was used for comparative analysis of absolute MBF from PET and MRI.

RESULTS

Overall, there were 256 segments in 16 patients, mean resting perfusion for PET was 1.06 ± 0.34 ml/min/g and 1.04 ± 0.30 ml/min/g for MRI (P = 0.05), whereas mean stress perfusion for PET was 2.00 ± 0.74 ml/min/g and 2.12 ± 0.76 ml/min/g for MRI (P < 0.01). Linear regression analysis in MBF revealed strong correlation (r = 0.91, slope = 0.96, P < 0.001) between PET and MRI. Myocardial perfusion reserve, calculated from the ratio of stress MBF over resting MBF, also showed a strong correlation between MRI and PET measurements (r = 0.82, slope = 0.81, P < 0.001).

CONCLUSION

The results demonstrated the feasibility of the simplified AIF estimation method for the accurate quantification of MBF by MRI with single sequence and single contrast injection. The MRI MBF correlated strongly with PET MBF obtained simultaneously. This post-processing technique will allow easy transformation of clinical perfusion imaging data into quantitative information.

Whole-heart, ungated, free-breathing, cardiac-phase-resolved myocardial perfusion MRI by using Continuous Radial Interleaved simultaneous Multi-slice acquisitions at sPoiled steady-state (CRIMP)

PURPOSE

To develop a whole-heart, free-breathing, non-electrocardiograph (ECG)-gated, cardiac-phase-resolved myocardial perfusion MRI framework (CRIMP; Continuous Radial Interleaved simultaneous Multi-slice acquisitions at sPoiled steady-state) and test its quantification feasibility.

METHODS

CRIMP used interleaved radial simultaneous multi-slice (SMS) slice groups to cover the whole heart in 9 or 12 short-axis slices. The sequence continuously acquired data without magnetization preparation, ECG gating or breath-holding, and captured multiple cardiac phases. Images were reconstructed by a motion-compensated patch-based locally low-rank reconstruction. Bloch simulations were performed to study the signal-to-noise ratio/contrast-to-noise ratio (SNR/CNR) for CRIMP and to study the steady-state signal under motion. Seven patients were scanned with CRIMP at stress and rest to develop the sequence. One human and two dogs were scanned at rest with a dual-bolus method to test the quantification feasibility of CRIMP. The dual-bolus scans were performed using both CRIMP and an ungated radial SMS saturation recovery (SMS-SR) sequence with injection dose = 0.075 mmol/kg to compare the sequences in terms of SNR, cardiac phase resolution and quantitative myocardial blood flow (MBF).

RESULTS

Perfusion images with multiple cardiac phases in all image slices with a temporal resolution of 72 ms/frame were obtained. Simulations and in-vivo acquisitions showed CRIMP kept the inner slices in steady-state regardless of motion. CRIMP outperformed SMS-SR in slice coverage (9 over 6), SNR (mean 20% improvement), and provided cardiac phase resolution. CRIMP and SMS-SR sequences provided comparable MBF values (rest systolic CRIMP = 0.58 ± 0.07, SMS-SR = 0.61 ± 0.16).

CONCLUSION

CRIMP allows for whole-heart, cardiac-phase-resolved myocardial perfusion images without ECG-gating or breath-holding. The sequence can provide MBF if an accurate arterial input function is obtained separately.

Simultaneous acquisition sequence for improved hepatic pharmacokinetics quantification accuracy (SAHA) for dynamic contrast-enhanced MRI of liver

PURPOSE

To propose a simultaneous acquisition sequence for improved hepatic pharmacokinetics quantification accuracy (SAHA) method for liver dynamic contrast-enhanced MRI.

METHODS

The proposed SAHA simultaneously acquired high temporal-resolution 2D images for vascular input function extraction using Cartesian sampling and 3D large-coverage high spatial-resolution liver dynamic contrast-enhanced images using golden angle stack-of-stars acquisition in an interleaved way. Simulations were conducted to investigate the accuracy of SAHA in pharmacokinetic analysis. A healthy volunteer and three patients with cirrhosis or hepatocellular carcinoma were included in the study to investigate the feasibility of SAHA in vivo.

RESULTS

Simulation studies showed that SAHA can provide closer results to the true values and lower root mean square error of estimated pharmacokinetic parameters in all of the tested scenarios. The in vivo scans of subjects provided fair image quality of both 2D images for arterial input function and portal venous input function and 3D whole liver images. The in vivo fitting results showed that the perfusion parameters of healthy liver were significantly different from those of cirrhotic liver and HCC.

CONCLUSIONS

The proposed SAHA can provide improved accuracy in pharmacokinetic modeling and is feasible in human liver dynamic contrast-enhanced MRI, suggesting that SAHA is a potential tool for liver dynamic contrast-enhanced MRI. Magn Reson Med 79:2629-2641, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

An empirical method for reducing variability and complexity of myocardial perfusion quantification by dual bolus cardiac MRI

PURPOSE

Myocardial perfusion can be quantified using the "dual bolus" technique, which uses two separate contrast boluses to avoid signal nonlinearity in the blood pool. This technique relies on knowing the precise ratio of contrast concentrations between the two boluses. In this study, we investigated the variability found in these ratios, as well as the error it introduces, and developed a method for correction.

METHODS

Five dogs received dual bolus myocardial perfusion MRI scans. Perfusion was calculated separately using assumed contrast dilution ratios and empirically determined contrast ratios. Perfusion was compared with reference standard fluorescent microspheres. The same technique was then applied to a cohort of six patients with no significant coronary artery stenosis by cardiac catheterization.

RESULTS

Assumed contrast dilution ratios were 10:1 for all animal and patient scans. Empirically derived contrast ratios were significantly different for animal (8.51:1 ± 1.53:1, P < 0.001) and patient scans (7.32:1 ± 2.27:1, P < 0.01). Incorporating empirically derived ratios for animal scans improved correlation with microspheres from 0.84 to 0.90 (P < 0.05).

CONCLUSION

Variability in dual bolus contrast concentration ratios is an important source of experimental error, especially outside of a carefully controlled laboratory setting. Empirically deriving the correct ratio is feasible and improves the accuracy of quantitative perfusion measurements. Magn Reson Med 77:2347-2355, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

A Comparison of Theory-Based and Experimentally Determined Myocardial Signal Intensity Correction Methods in First-Pass Perfusion Magnetic Resonance Imaging

OBJECTIVES

To evaluate the impact of correcting myocardial signal saturation on the accuracy of absolute myocardial blood flow (MBF) measurements.

MATERIALS AND METHODS

We performed 15 dual bolus first-pass perfusion studies in 7 dogs during global coronary vasodilation and variable degrees of coronary artery stenosis. We compared microsphere MBF to MBF calculated from uncorrected and corrected MRI signal. Four correction methods were tested, two theoretical methods (Th1 and Th2) and two empirical methods (Em1 and Em2).

RESULTS

The correlations with microsphere MBF (n = 90 segments) were: uncorrected (y = 0.47x + 1.1, r = 0.70), Th1 (y = 0.53x + 1.0, r = 0.71), Th2 (y = 0.62x + 0.86, r = 0.73), Em1 (y = 0.82x + 0.86, r = 0.77), and Em2 (y = 0.72x + 0.84, r = 0.75). All corrected methods were not significantly different from microspheres, while uncorrected MBF values were significantly lower. For the top 50% of microsphere MBF values, flows were significantly underestimated by uncorrected SI (31%), Th1 (25%), and Th2 (19%), while Em1 (1%), and Em2 (9%) were similar to microsphere MBF.

CONCLUSIONS

Myocardial signal saturation should be corrected prior to flow modeling to avoid underestimation of MBF by MR perfusion imaging.

Assessment of myocardial perfusion with MRI using a modified dual bolus method

Quantification of regional myocardial blood flow (rMBF) with first-pass magnetic resonance imaging (FP-MRI) requires two contrast agent injections (dual bolus technique), inducing error in the determined rMBF if the injections differ. We hypothesize that using input and residue curves of the same injection would be more reliable. We aim to introduce and evaluate a novel method to correct the high concentration arterial input function (AIF) for determination of rMBF. Sixteen patients with non-Hodgkin's lymphoma were examined before and after chemotherapy. The input function was solved by correcting initial high concentration AIF using the ratio of low and high contrast AIF areas, normalized by corresponding heart rates (modified dual bolus method). For comparison, the scaled low contrast AIF was used as an input function (dual bolus method). Unidirectional transfer coefficient K(trans) was calculated using both methods. K(trans)-values determined with the dual bolus (0.81 ± 0.32 ml g(-1) min(-1)) and modified dual bolus (0.77 ± 0.42 ml g(-1) min(-1)) methods were in agreement (p = 0.21). Mean K(trans)-values increased from 0.76 ± 0.43 to 0.89 ± 0.55 ml g(-1) min(-1) after chemotherapy (p = 0.17). The modified dual bolus technique agrees with the dual bolus technique (R2 = 0.899) when the low and high contrast injections are similar. However, when this is not the case, the modified dual bolus technique may be more reliable.

Absolute quantification of myocardial blood flow with constrained estimation of the arterial input function

PURPOSE

To evaluate the performance of the constrained alternating minimization with model (CAMM) method for estimating the input function from the myocardial tissue curves.

MATERIALS AND METHODS

Myocardial perfusion imaging was performed on seven canine models of coronary artery disease in 15 imaging sessions. In each session, stress was induced with intravenous infusion of adenosine and a variable occluder created coronary artery stenosis. A dual bolus protocol was used for each acquisition, and input functions were then estimated using the CAMM method with data acquired from the high dose scan following each imaging session. For each acquisition, myocardial blood flow was measured by injected microspheres.

RESULTS

The dual bolus and CAMM-derived flows were not significantly different (P = 0.18), and the correlation between the two methods was high (r = 0.97). The correlation between the dual bolus and CAMM methods and microsphere measurements was lower than that for the two MR methods (r = 0.53; r = 0.43, respectively).

CONCLUSION

The CAMM method presented here shows promise in estimating myocardial blood flow in patients with coronary artery disease at stress with a single injection and without any specialized acquisitions. Further work is needed to validate the approach in a clinical setting.

Single-channel blind estimation of arterial input function and tissue impulse response in DCE-MRI

Multipass dynamic MRI and pharmacokinetic modeling are used to estimate perfusion parameters of leaky capillaries. Curve fitting and nonblind deconvolution are the established methods to derive the perfusion estimates from the observed arterial input function (AIF) and tissue tracer concentration function. These nonblind methods are sensitive to errors in the AIF, measured in some nearby artery or estimated by multichannel blind deconvolution. Here, a single-channel blind deconvolution algorithm is presented, which only uses a single tissue tracer concentration function to estimate the corresponding AIF and tissue impulse response function. That way, many errors affecting these functions are reduced. The validity of the algorithm is supported by simulations and tests on real data from mouse. The corresponding nonblind and multichannel methods are also presented.

Myocardial perfusion acquisition without magnetization preparation or gating

Current myocardial perfusion MRI acquisitions are performed with a saturation recovery sequence, in large part to minimize sensitivity to arrhythmia. A new approach is proposed here where the images are acquired ungated at steady state without use of a saturation pulse. The data are acquired continuously and reach steady state after the first few images. A confluence of advances has made this new paradigm of an ungated steady-state acquisition possible-very rapid undersampled readouts with new reconstruction technologies permit enough measurements that continuous acquisition becomes a feasible approach. Gating can be applied retrospectively from a logged electrocardiogram (ECG) or with self-gating methods. In this work, simulations and measurements in a concentration phantom are used to demonstrate that similar contrast and signal can be obtained with the standard saturation recovery and the proposed spoiled gradient echo (SPGR) acquisition. Specifically, for a flip angle of 14° and a saturation recovery time of 80 ms, similar signals are acquired over a range of T(1) s that reflect realistic myocardial tissue concentrations. Preliminary results in one subject are presented to show the potential of this new approach. The method may allow for cine cardiac perfusion and more signal-to-noise ratio-efficient acquisitions.