Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Feb 9;15(1):4825.
doi: 10.1038/s41598-025-88579-z.

Flexible and cost-effective deep learning for accelerated multi-parametric relaxometry using phase-cycled bSSFP

Affiliations

Flexible and cost-effective deep learning for accelerated multi-parametric relaxometry using phase-cycled bSSFP

Florian Birk et al. Sci Rep. .

Abstract

To accelerate the clinical adoption of quantitative magnetic resonance imaging (qMRI), frameworks are needed that not only allow for rapid acquisition, but also flexibility, cost efficiency, and high accuracy in parameter mapping. In this study, feed-forward deep neural network (DNN)- and iterative fitting-based frameworks are compared for multi-parametric (MP) relaxometry based on phase-cycled balanced steady-state free precession (pc-bSSFP) imaging. The performance of supervised DNNs (SVNN), self-supervised physics-informed DNNs (PINN), and an iterative fitting framework termed motion-insensitive rapid configuration relaxometry (MIRACLE) was evaluated in silico and in vivo in brain tissue of healthy subjects, including Monte Carlo sampling to simulate noise. DNNs were trained on three distinct in silico parameter distributions and at different signal-to-noise-ratios. The PINN framework, which incorporates physical knowledge into the training process, ensured more consistent inference and increased robustness to training data distribution compared to the SVNN. Furthermore, DNNs utilizing the full information of the underlying complex-valued MR data demonstrated ability to accelerate the data acquisition by a factor of 3. Whole-brain relaxometry using DNNs proved to be effective and adaptive, suggesting the potential for low-cost DNN retraining. This work emphasizes the advantages of in silico DNN MP-qMRI pipelines for rapid data generation and DNN training without extensive dictionary generation, long parameter inference times, or prolonged data acquisition, highlighting the flexible and rapid nature of lightweight machine learning applications for MP-qMRI.

Keywords: Deep Neural Networks; MIRACLE; Multi-parametric Quantitative MRI; Phase-Cycled bSSFP; Relaxometry.

PubMed Disclaimer

Conflict of interest statement

Declarations. Competing Interests: The authors declare no competing interests.

Figures

Fig. 1
Fig. 1
The workflow proposed in this work (purple cubes represent the extended input and output in case of complex-based DNNs). (a) Data Simulation: The input parameters formula image entering the analytical bSSFP signal model (see Eq. 1) were sampled from three different distributions (in vivo, uniform, and uniform extended) for formula image and formula image, and from a single uniform distribution for formula image and formula image. The sequence parameters from the in vivo acquisition protocol (TR, TE, formula image, formula image) were used to draw 400,000 signal samples formula image from each formula image and formula image distribution. (b) Multi-Parametric-Fitting Frameworks: The input to each of the three frameworks, which means the physics-informed neural network (PINN or formula image 1), the supervised neural network (SVNN or formula image 2), and the iterative golden section search (GSS) fitting (MIRACLE, 3), consisted of the amplitudes (magnitude-based) or real and imaginary parts (complex-based, without imaginary part of formula image) of the three lowest-order SSFP configurations computed from a Fourier transform (FT) of the phase-cycled bSSFP signal with the option to add noise and in addition of formula image 1) and 2) use the same multilayer perceptron architecture (magnitude-based: 64 neurons per hidden layer, complex-based: 256 neurons per hidden layer) to estimate the inverse signal model and predict the parameters formula image. 1) uses the predicted formula image, formula image, and formula image (with the addition of the formula image input) to generate an estimated signal formula image and compare it to the input signal formula image in the formula image loss, while 2) compares the predicted formula image, formula image, and formula image directly to the respective ground truth target parameters formula image in the formula image loss. The off-resonance formula image was only utilized for the complex-based DNNs (purple cubes).
Fig. 2
Fig. 2
Influence of training data SNR and training data distribution on accuracy of investigated DNNs in silico. The relative error in percent formula image with formula image, between the mean of the MC simulation formula image and the ground truth formula image, is quantified for formula image and formula image parameter estimates of the SVNNs (left) and PINNs (right) trained on noise-free (a, SNR = inf) and noise-corrupted (b, SNR = 25) data with different training data distributions. The MC estimation is performed on a noise-corrupted in silico linear test grid with formula image matched to in vivo conditions as well as a formula image and formula image range corresponding to brain tissues (consistent with the parameter range of the in vivo and uniform distribution employed for DNN training). Parameter over- and underestimation with respect to the ground truth are shown in red and blue, respectively.
Fig. 3
Fig. 3
Coefficient of determination versus test data SNR of investigated DNNs relative to MIRACLE in silico. The CoD between the mean MC relaxation parameter predictions formula image and the ground truth formula image with formula image (formula image in red and formula image in blue) is shown for the linear test grid (left column) and the in vivo distribution test data (right column). (a) CoD versus test data SNR for MIRACLE (formula image). (b) and (c) The absolute CoD difference between the DNNs and MIRACLE (formula image) versus test data SNR for the PINN (b) and the SVNN (c). For both SVNN and PINN, three models trained on noise-free data (SNR = inf) with different data distributions are evaluated: in vivo (formula image), uniform (formula image), and uniform extended distribution (formula image). Note that positive/negative values in (b) and (c) are referring to higher/lower CoD values of the DNNs relative to MIRACLE.
Fig. 4
Fig. 4
In vivo analysis of the effect of the DNN training data distribution relative to MIRACLE. A representative axial slice of the in vivo whole-brain formula image (first row) and formula image (third row) predictions of an unseen test subject is shown for the MIRACLE framework (first column), and both DNNs, each trained on in silico data without additional noise (SNR = inf) and three different distributions (in vivo, uniform, and uniform extended). The absolute differences (formula image with i = 1, 2) between the DNN predictions and the MIRACLE prediction are shown in the second and fourth row for formula image and formula image, respectively. Red and blue refer to an over- and underestimation of the DNN framework predictions relative to the MIRACLE framework predictions, respectively.
Fig. 5
Fig. 5
Influence of off-resonance-related phase accumulation within TR (formula image) on DNN and MIRACLE relaxometry in silico depending on the number of bSSFP phase cycles. DNNs were trained on noise-free data from a uniform distribution of formula image and formula image and for a varying number of phase cycles (formula image). The noise-free test data were simulated based on reference formula image (62 ms) and formula image (939 ms) white matter relaxation values at 3 T. The relative error between the parameter predictions formula image and the simulated ground truth value formula image, formula image with i = 1, 2, is shown for MIRACLE and the standard magnitude-based DNNs (left column) as well as the complex-based DNNs (right column). Dashed lines indicate the 0 formula image error for reference.
Fig. 6
Fig. 6
Performance of complex-based DNN versus MIRACLE and magnitude-based DNN in vivo formula image estimation in case of accelerated pc-bSSFP acquisitions with only 6 and 4 phase cycles in comparison to the standard protocol with 12 phase cycles. A representative axial slice of the in vivo whole-brain formula image predictions of an unseen test subject obtained with 12, 6, and 4 phase cycles is shown in the first, second, and third column for each framework (MIRACLE, SVNN, formula image, PINN, formula image, from top to bottom), respectively. The absolute differences formula image between the parameter predictions with 6 and 4 phase cycles relative to the reference with 12 phase cycles are shown in the fourth and fifth column, respectively.
Fig. 7
Fig. 7
Performance of complex-based DNN versus MIRACLE and magnitude-based DNN in vivo formula image estimation in case of accelerated pc-bSSFP acquisitions with only 6 and 4 phase cycles in comparison to the standard protocol with 12 phase cycles. A representative axial slice of the in vivo whole-brain formula image predictions of an unseen test subject obtained with 12, 6, and 4 phase cycles is shown in the first, second, and third column for each framework (MIRACLE, SVNN, formula image, PINN, formula image, from top to bottom), respectively. The absolute differences formula image between the parameter predictions with 6 and 4 phase cycles relative to the reference with 12 phase cycles are shown in the fourth and fifth column, respectively.
Fig. 8
Fig. 8
Efficiency of standard magnitude-based DNN inverse signal model learning versus epochs, corroborated by representative relaxation time maps of single-epoch in vivo whole-brain inference. The CoD during SVNN (a) and PINN (b) training is calculated for each epoch with respect to the final-epoch model and plotted versus epochs for in vivo formula image (red) and formula image (blue) predictions in whole-brain WM, GM, and WM+GM tissue masks of an unseen test subject. Additionally, the validation loss for both DNN frameworks is shown in black on a logarithmic scale. The employed DNNs were trained on the in silico uniform noise-free data distribution. Note that the final validation loss of the SVNN framework is on the order of one magnitude lower than the one of the PINN framework due to the different definitions of the loss functions and embedding of physical constraints for the PINN. Corresponding representative axial, coronal, and sagittal slices of in vivo whole-brain formula image and formula image single-echo versus final-epoch predictions of an unseen test subject are shown for SVNN (c) and PINN (d). See Supplementary Figure S8 for the corresponding results with complex-based DNNs.

References

    1. Van Beek, E. J. et al. Value of MRI in medicine: More than just another test?. J. Magn. Reson. Imaging49, e14–e25. 10.1002/jmri.26211 (2019). - PMC - PubMed
    1. Gracien, R.-M. et al. How stable is quantitative MRI?–Assessment of intra- and inter-scanner-model reproducibility using identical acquisition sequences and data analysis programs. Neuroimage207, 116364. 10.1016/j.neuroimage.2019.116364 (2020). - PubMed
    1. Granziera, C. et al. Quantitative magnetic resonance imaging towards clinical application in multiple sclerosis. Brain144, 1296–1311. 10.1093/brain/awab029 (2021). - PMC - PubMed
    1. Straub, S. et al. Quantitative magnetic resonance imaging biomarkers for cortical pathology in multiple sclerosis at 7 T. NMR Biomed.36, e4847. 10.1002/nbm.4847 (2023). - PubMed
    1. Sarkar, P., Sherwani, P., Dev, R. & Tiwari, A. Role of T2 relaxometry in localization of mesial temporal sclerosis and the degree of hippocampal atrophy in patients with intractable temporal lobe epilepsy: A cross sectional study. Hippocampus33, 1189–1196. 10.1002/hipo.23572 (2023). - PubMed

LinkOut - more resources