DeepCEST: 9.4 T Chemical exchange saturation transfer MRI contrast predicted from 3 T data – a proof of concept study

Abstract

Purpose To determine the feasibility of employing the prior knowledge of well‐separated chemical exchange saturation transfer (CEST) signals in the 9.4 T Z‐spectrum to separate overlapping CEST signals acquired at 3 T, using a deep learning approach trained with 3 T and 9.4 T CEST spectral data from brains of the same subjects. Methods Highly spectrally resolved Z‐spectra from the same volunteer were acquired by 3D‐snapshot CEST MRI at 3 T and 9.4 T at low saturation power of B1 = 0.6 µT. The volume‐registered 3 T Z‐spectra‐stack was then used as input data for a three layer deep neural network with the volume‐registered 9.4 T fitted parameter stack as target data. Results An optimized neural net architecture could be found and verified in healthy volunteers. The gray‐/white‐matter contrast of the different CEST effects was predicted with only small deviations (Pearson R = 0.89). The 9.4 T prediction was less noisy compared to the directly measured CEST maps, although at the cost of slightly lower tissue contrast. Application to an unseen tumor patient measured at 3 T and 9.4 T revealed that tumorous tissue Z‐spectra and corresponding hyper‐/hypointensities of different CEST effects can also be predicted (Pearson R = 0.84). Conclusion The 9.4 T CEST signals acquired at low saturation power can be accurately estimated from CEST imaging at 3 T using a neural network trained with coregistered 3 T and 9.4 T data of healthy subjects. The deepCEST approach generalizes to Z‐spectra of tumor areas and might indicate whether additional ultrahigh‐field (UHF) scans will be beneficial.

DeepCEST 3T: Robust MRI parameter determination and uncertainty quantification with neural networks—application to CEST imaging of the human brain at 3T

“…NN1 shows de‐noising capability (Figure C), which was also observed in previous work and is demonstrated in more detail in Supporting Information Figure .…”

“…Because of that, networks operating only in spectral, but not spatial, domain were argued to generalize also to tumor spectra as long as they can be expressed as non-linear combinations of healthy spectra. NN1 shows de-noising capability ( Figure 4C), which was also observed in previous work 18 and is demonstrated in more detail in Supporting Information Figure S10…”

“…This probabilistic output layer can be adapted to any other neural network used for regression problems with only small changes to the network architecture. Hereby, it constitutes a simple way of enabling uncertainty estimation for similar deep learning approaches like the prediction of 9.4T contrasts from 3T Z‐spectra or reconstruction of MR fingerprinting data, a technique that has also found applications in CEST …”

“…Deep learning was recently applied to circumvent the mentioned problems of complex non‐linear models used in oxygen extraction fraction and diffusion parameter mapping . In CEST imaging, a first neural network approach was shown to successfully map 3T data to 9.4T CEST contrasts . However, because deep neural networks (NN) are often considered as “black boxes” with limited insight into the mapping from input to prediction, a question that arises is how accurate and reliable such predictions are.…”

“…17 In CEST imaging, a first neural network approach was shown to successfully map 3T data to 9.4T CEST contrasts. 18 However, because deep neural networks (NN) are often considered as "black boxes" with limited insight into the mapping from input to prediction, a question that arises is how accurate and reliable such predictions are. A neural network that gives seemingly confident predictions for all possible inputs, regardless if reasonably supported by learning or not, may lead to critical failure cases and result in mistrust by potential users.…”