Introduction

MRI is a noninvasive modality which allows one to acquire images at high spatial resolution and it produces excellent soft tissue contrasts. A complete MRI exam typically consists of several sequences, designed to obtain images with different contrast This leads to long acquisition time, which can be challenging for children and uncollaborative patients. In this scenario, a method to synthesize MR images with adjustable parameters is desirable.
Current approaches synthesize image contrasts based on measurements of tissue parameters from a single acquisition¹, reducing the scanning time. This technique, known as Synthetic MRI, allows to obtain images by post processing acquired data and radiologists can retrospectively manipulate the image to change the contrasts. MAGnetic resonance image Compilation (MAGiC)² and SyMRI³ are examples of products that implement synthetic MRI. Those tools, capable of generating and visualizing synthetic images, are vendor-specific and we are not aware of open-source tools to achieve the same results using in-house developed multiparametric imaging techniques.
We introduce PySynthMRI, an open-source tool that offers a friendly interface (Figure 1) that uses quantitative maps of T1, T2 and proton density (PD) as input and allows the user to vary scanner settings such as echo time (TE), repetition time (TR), inversion time (TI), that mimic those used in conventional scans. Using the signal equations for different types of acquisitions, the tool dynamically synthesizes and shows to the user the contrast images, such as T1 weighted, T2 weighted, FLAIR . The simplicity of use makes PySynthMRI a valuable tool for both research and teaching.

Design and features

The PySynthMRI tool is written in Python 3.6, an open programming language available in all computing platforms. The tool makes use of the open-source numpy, open-cv and pyqtgraph packages, chosen with the aim of maximizing the efficiency in manipulating image matrices and therefore to make the interaction with the graphical interface fluid for the user. The code is organized in a Model-View-Controller (MVC) pattern and its modularity simplifies the tests and makes it possible to add new features without impacting the codebase.
The interface of PySynthMRI allows the user to load the needed quantitative maps, including but not limited to T1, T2 and proton density quantitative maps. It provides a simple way to modify sequence parameters in order to choose values that generate the synthetic image, which is immediately shown to the user using the preferred interpolation for best rendering.
New custom sequence parameters, including their weights in the signal equation, can be added using configuration files or dynamically in order to obtain unconventional contrast weighted images. PySynthMRI generates the images S as a function of chosen scanner parameters and provided quantitative maps, according to the model represented by
$$abs(\prod_{p\ \in\ params} o_p+s_p*e^{\frac{-p}{Q_p}})$$
where each scanner parameter p is associated to a different quantitative map Q, an offset o and a scale factor s.
Since scanner parameters are independent, any contrast image can be obtained. To facilitate the user, the tool contains a set of default parameters in its configuration file used to synthesize conventional contrast images. T1-weighted MP2RAGE (Magnetization-Prepared 2 RApid Gradient Echoes) and GRE (Gradient Echo) are obtained respectively using TI and TR.
$$MP2RAGE =abs(1 - 2 * exp(-TI /T1)))$$
$$GRE =abs(1 - 2 * exp(-TR /T1))$$
The T2-weighted FSE (Fast Spin Echo) image is synthesized using the echo time, scaled on proton density as follows
$$FSE =abs(PD *exp(-TE/T2))$$
Furthermore, FLAIR (Fluid Attenuated Inversion Recovery) is obtained according to the following equation
$$FLAIR =abs(PD *exp(-TE/T2)*exp(-T1SAT/T1)*(1 - 2 * exp(-TI /T1)))$$
Once generated, the synthetic images can be exported in the preferred format, obtaining a new file with updated metadata.
The tool is in continuous extension. The source code as well as documentation is released on GitHub under free MIT software license and can be accessed under: https://github.com/FiRMLAB-Pisa/pySynthMRI.git

Discussion and Conclusion

The development and dissemination of quantitative MRI techniques such as Magnetic Resonance Fingerprinting makes the PySynthMRI tool beneficial for both MRI research and teaching. In the research field the graphical interface of our tool allows specialists to easily synthesize all needed contrast images at post-processing time, granting the possibility to reach the optimal contrast weighting for that particular subject. In teaching it can be used to show how sequence parameters impact the generation of images with different tissue contrast.
The modular architecture allows developers and contributors to easily extend and customize the tool to incorporate additional parameter maps such as T2* or MT or different predefined contrast images. The tool functionality could be further improved including a segmentation software module to immediately show how different generated images impact on the computed results.

Acknowledgements

No acknowledgement found.