Feliks Kogan, Postdoctoral Faculty Sponsor
A deep learning approach for magnetic resonance fingerprinting: Scaling capabilities and good training practices investigated by simulations.
Physica medica : PM : an international journal devoted to the applications of physics to medicine and biology : official journal of the Italian Association of Biomedical Physics (AIFB)
2021; 89: 80-92
MR fingerprinting (MRF) is an innovative approach to quantitative MRI. A typical disadvantage of dictionary-based MRF is the explosive growth of the dictionary as a function of the number of reconstructed parameters, an instance of the curse of dimensionality, which determines an explosion of resource requirements. In this work, we describe a deep learning approach for MRF parameter map reconstruction using a fully connected architecture. Employing simulations, we have investigated how the performance of the Neural Networks (NN) approach scales with the number of parameters to be retrieved, compared to the standard dictionary approach. We have also studied optimal training procedures by comparing different strategies for noise addition and parameter space sampling, to achieve better accuracy and robustness to noise. Four MRF sequences were considered: IR-FISP, bSSFP, IR-FISP-B1, and IR-bSSFP-B1. A comparison between NN and the dictionary approaches in reconstructing parameter maps as a function of the number of parameters to be retrieved was performed using a numerical brain phantom. Results demonstrated that training with random sampling and different levels of noise variance yielded the best performance. NN performance was at least as good as the dictionary-based approach in reconstructing parameter maps using Gaussian noise as a source of artifacts: the difference in performance increased with the number of estimated parameters because the dictionary method suffers from the coarse resolution of the parameter space sampling. The NN proved to be more efficient in memory usage and computational burden, and has great potential for solving large-scale MRF problems.
View details for DOI 10.1016/j.ejmp.2021.07.013
View details for PubMedID 34352679
Characterization of Structural Bone Properties through Portable Single-Sided NMR Devices: State of the Art and Future Perspectives.
International journal of molecular sciences
2021; 22 (14)
Nuclear Magnetic Resonance (NMR) is a well-suited methodology to study bone composition and structural properties. This is because the NMR parameters, such as the T2 relaxation time, are sensitive to the chemical and physical environment of the 1H nuclei. Although magnetic resonance imaging (MRI) allows bone structure assessment in vivo, its cost limits the suitability of conventional MRI for routine bone screening. With difficulty accessing clinically suitable exams, the diagnosis of bone diseases, such as osteoporosis, and the associated fracture risk estimation is based on the assessment of bone mineral density (BMD), obtained by the dual-energy X-ray absorptiometry (DXA). However, integrating the information about the structure of the bone with the bone mineral density has been shown to improve fracture risk estimation related to osteoporosis. Portable NMR, based on low-field single-sided NMR devices, is a promising and appealing approach to assess NMR properties of biological tissues with the aim of medical applications. Since these scanners detect the signal from a sensitive volume external to the magnet, they can be used to perform NMR measurement without the need to fit a sample inside a bore of a magnet, allowing, in principle, in vivo application. Techniques based on NMR single-sided devices have the potential to provide a high impact on the clinical routine because of low purchasing and running costs and low maintenance of such scanners. In this review, the development of new methodologies to investigate structural properties of trabecular bone exploiting single-sided NMR devices is reviewed, and current limitations and future perspectives are discussed.
View details for DOI 10.3390/ijms22147318
View details for PubMedID 34298936
Single-sided NMR to estimate morphological parameters of the trabecular bone structure.
Magnetic resonance in medicine
PURPOSE: Single-sided 1 H-NMR is proposed for the estimation of morphological parameters of trabecular bone, and potentially the detection of pathophysiological alterations of bone structure. In this study, a new methodology was used to estimate such parameters without using an external reference signal, and to study intratrabecular and intertrabecular porosities, with a view to eventually scanning patients.METHODS: Animal trabecular bone samples were analyzed by a single-sided device. The Carr-Purcell-Meiboom-Gill sequence of 1 H nuclei of fluids, including marrow, confined inside the bone, was analyzed by quasi-continuous T2 distributions and separated into two 1 H pools: short and long T2 components. The NMR parameters were estimated using models of trabecular bone structure, and compared with the corresponding micro-CT.RESULTS: Without any further assumptions, the internal reference parameter (short T2 signal intensity fraction) enabled prediction of the micro-CT parameters BV/TV (volume of the trabeculae/total sample volume) and BS/TV (external surface of the trabeculae/total sample volume) with linear correlation coefficient >0.80. The assignment of the two pools to intratrabecular and intertrabecular components yielded an estimate of average intratrabecular porosity (33±5)%. Using the proposed models, the NMR-estimated BV/TV and BS/TV were found to be linearly related to the corresponding micro-CT values with high correlation (>0.90 for BV/TV; >0.80 for BS/TV) and agreement coefficients.CONCLUSION: Low-field, low-cost portable devices that rely on intrinsic magnetic field gradients and do not use ionizing radiation are viable tools for in vitro preclinical studies of pathophysiological structural alterations of trabecular bone.
View details for DOI 10.1002/mrm.28648
View details for PubMedID 33349979
- Single-sided NMR for the diagnosis of osteoporosis: Diffusion weighted pulse sequences for the estimation of trabecular bone volume fraction in the presence of muscle tissue ELSEVIER SCIENCE BV. 2018: 166–70
Bone volume-to-total volume ratio measured in trabecular bone by single-sided NMR devices
MAGNETIC RESONANCE IN MEDICINE
2018; 79 (1): 501–10
Reduced bone strength is associated with a loss of bone mass, usually evaluated by dual-energy X-ray absorptiometry, although it is known that the bone microstructure also affects the bone strength. Here, a method is proposed to measure (in laboratory) the bone volume-to-total volume ratio by single-sided NMR scanners, which is related to the microstructure of the trabecular bone.Three single-sided scanners were used on animal bone samples. These low-field, mobile, low-cost devices are able to detect the NMR signal, regardless of the sample sizes, without the use of ionizing radiations, with the further advantage of signal localization offered by their intrinsic magnetic field gradients.The performance of the different single-sided scanners have been discussed. The results have been compared with bone volume-to-total volume ratio by micro CT and MRI, obtaining consistent values.Our results demonstrate the feasibility of the method for laboratory analyses, which are useful for measurements like porosity on bone specimens. This can be considered as the first step to develop an NMR method based on the use of a mobile single-sided device, for the diagnosis of osteoporosis, through the acquisition of the signal from the appendicular skeleton, allowing for low-cost, wide screening campaigns. Magn Reson Med 79:501-510, 2018. © 2017 International Society for Magnetic Resonance in Medicine.
View details for DOI 10.1002/mrm.26697
View details for Web of Science ID 000417926300049
View details for PubMedID 28394083