Generating Synthetic MR Spectroscopic Imaging Data with Generative Adversarial Networks to Train Machine Learning Models.

PURPOSE: To develop a new method to generate synthetic MR spectroscopic imaging (MRSI) data for training machine learning models. METHODS: This study targeted routine MRI examination protocols with single voxel spectroscopy (SVS). A novel model derived from pix2pix generative adversarial networks was proposed to generate synthetic MRSI data using MRI and SVS data as inputs. T1- and T2-weighted, SVS, and reference MRSI data were acquired from healthy brains with clinically available sequences. The proposed model was trained to generate synthetic MRSI data. Quantitative evaluation involved the calculation of the mean squared error (MSE) against the reference and metabolite ratio value. The effect of the location of and the number of the SVS data on the quality of the synthetic MRSI data was investigated using the MSE. RESULTS: The synthetic MRSI data generated from the proposed model were visually closer to the reference. The 95% confidence interval (CI) of the metabolite ratio value of synthetic MRSI data overlapped with the reference for seven of eight metabolite ratios. The MSEs tended to be lower in the same location than in different locations. The MSEs among groups of numbers of SVS data were not significantly different. CONCLUSION: A new method was developed to generate MRSI data by integrating MRI and SVS data. Our method can potentially increase the volume of MRSI data training for other machine learning models by adding SVS acquisition to routine MRI examinations.

Cognitive control affects motor learning through local variations in GABA within the primary motor cortex.

The primary motor cortex (M1) is crucial for motor learning; however, its interaction with other brain areas during motor learning remains unclear. We hypothesized that the fronto-parietal execution network (FPN) provides learning-related information critical for the flexible cognitive control that is required for practice. We assessed network-level changes during sequential finger tapping learning under speed pressure by combining magnetic resonance spectroscopy and task and resting-state functional magnetic resonance imaging. There was a motor learning-related increase in preparatory activity in the fronto-parietal regions, including the right M1, overlapping the FPN and sensorimotor network (SMN). Learning-related increases in M1-seeded functional connectivity with the FPN, but not the SMN, were associated with decreased GABA/glutamate ratio in the M1, which were more prominent in the parietal than the frontal region. A decrease in the GABA/glutamate ratio in the right M1 was positively correlated with improvements in task performance (p = 0.042). Our findings indicate that motor learning driven by cognitive control is associated with local variations in the GABA/glutamate ratio in the M1 that reflects remote connectivity with the FPN, representing network-level motor sequence learning formations.

Comparison of 3T and 7T MRI for the visualization of globus pallidus sub-segments.

The success of deep brain stimulation (DBS) targeting the internal globus pallidus (GPi) depends on the accuracy of electrode localization inside the GPi. In this study, we sought to compare visualization of the medial medullary lamina (MML) and accessory medullary lamina (AML) between proton density-weighted (PDW) and T2-weighted (T2W) sequences on 3T and 7T MRI scanners. Eleven healthy participants (five men and six women; age, 19-28 years; mean, 21.5) and one 61-year-old man were scanned using two-dimensional turbo spin-echo PDW and T2W sequences on 3T and 7T MRI scanners with a 32-channel receiver head coil and a single-channel transmission coil. Profiles of signal intensity were obtained from the pixel values of straight lines over the GP regions crossing the MML and AML. Contrast ratios (CRs) for GPe/MML, GPie/MML, GPie/AML, and GPii/AML were calculated. Qualitatively, 7T visualized both the MML and AML, whereas 3T visualized the MML less clearly and hardly depicted the AML. The T2W sequence at 7T yielded significantly higher CRs for GPie/MML, GPie/AML, and GPii/AML than the PDW sequence at 7T or 3T. The T2W sequence at 7T allows visualization of the internal structures of GPi segments with high signal intensity and contrast.

Usefulness of model-based iterative reconstruction in semi-automatic volumetry for ground-glass nodules at ultra-low-dose CT: a phantom study.

This study aimed to investigate the effects of parameter presets of the forward projected model-based iterative reconstruction solution (FIRST) on the accuracy of pulmonary nodule volume measurement. A torso phantom with simulated nodules [diameter: 5, 8, 10, and 12 mm; computed tomography (CT) density: - 630 HU] was scanned with a multi-detector CT at tube currents of 10 mA (ultra-low-dose: UL-dose) and 270 mA (standard-dose: Std-dose). Images were reconstructed with filtered back projection [FBP; standard (Std-FBP), ultra-low-dose (UL-FBP)], FIRST Lung (UL-Lung), and FIRST Body (UL-Body), and analyzed with a semi-automatic software. The error in the volume measurement was determined. The errors with UL-Lung and UL-Body were smaller than that with UL-FBP. The smallest error was 5.8% ± 0.3 for the 12-mm nodule with UL-Body (middle lung). Our results indicated that FIRST Body would be superior to FIRST Lung in terms of accuracy of nodule measurement with UL-dose CT.

Development and Characterization of Novel LipoCEST Agents Based on Thermosensitive Liposomes.

PURPOSE: To develop a novel probe for chemical exchange saturation transfer magnetic resonance imaging (CEST MRI) based on thermosensitive liposomes (lipoCEST) for theranostics, in which diagnostics and therapy are integrated into a single platform. METHODS: We developed two kinds of lipoCEST agents. The first kind encapsulated dysprosium (Dy)-1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA)-Na·3NaCl, terbium-DOTA-Na·3NaCl, or thulium-DOTA-Na·3NaCl into the inner cavity of thermosensitive liposomes, while the second kind encapsulated Dy-DOTA-Na and incorporated amphiphilic metal complex [thulium-diethylenetriamine pentaacetic acid-bis (stearylamide) (Tm-DTPA-BSA)] as a membrane constituent. The nuclear magnetic resonance (NMR)- and Z-spectra of these lipoCEST agents were acquired at various temperatures on a 9.4T MRI scanner. To investigate their applicability to the drug release induced by hyperthermia, we also encapsulated a fluorescent dye (calcein) into the inner cavity of liposomes and measured calcein release after warming them. RESULTS: The intra- and extraliposomal water signals could be differentiated in all agents from their NMR- and Z-spectra. The agent incorporating Tm-DTPA-BSA showed the largest chemical shift (approximately 15 ppm) derived from the intraliposomal water protons. The calcein retained in this agent was successfully released at 44°C. The agent incorporating 30 mol% of Tm-DTPA-BSA in its membrane released more calcein at 42-44ºC than that of the agent incorporating 10 mol%. CONCLUSION: We developed novel thermosensitive lipoCEST agents and characterized them. Our preliminary results suggest that they are useful and can be applied to theranostics.