Figure 1 | Scientific Reports

Figure 1

From: Dynamic and rapid deep synthesis of chemical exchange saturation transfer and semisolid magnetization transfer MRI signals

Figure 1

Network architectures. (a) A dynamic fully connected NN receives tissue parameters, scanner parameters, and the previous inference cycle signal S\(_i\). The output is the next time-evolution signal element S\(_{i+1}\). The inference cycle continues according to the user’s desired signal acquisition length N. (b) Application Optimized network. The input of the network are tissue and scanner parameters. The output is the entire magnetic signal inferred at once, according to a particular value N used during the training. T\(_{1w}\)—water longitudinal relaxation, T\(_{2w}\)—water transverse relaxation, T\(_{1s}\)—solute longitudinal relaxation, T\(_{2s}\)—solute transverse relaxation, f\(_{si}\)—proton volume fraction for solute i, k\(_{swi}\)—proton exchange rate for solute i, \(\omega _b\)—chemical shift, T\(_p\)—saturation pulse duration, T\(_{rec}\)—recovery time, B\(_0\)—main magnetic field, FA—flip angle, \(\omega _{rf}\)—saturation pulse frequency offset, B\(_1\)—saturation pulse power, n\(_p\)—number of pulses in the saturation pulse train.

Back to article page