Generation of hybrid color images from T1 and T2 acquired simultaneously with MRF
Katherine L. Wright1, Peter Schmitt2, Dan Ma1, Anagha Deshmane3, Vikas Gulani1, and Mark Griswold1

1Radiology, Case Western Reserve University, Cleveland, OH, United States, 2Siemens Healthcare, Erlangen, Germany, 3Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States


This work proposes a method for the calculation of a single color image using quantitative T1 and T2 measurements acquired with Magnetic Resonance Fingerprinting. Quantitative MRF parameters are transformed and scaled with the goal of making normal tissues appear in grayscale and tissues with different T1 and T2 values (lesions) appear in color.


Magnetic Resonance Fingerprinting (MRF) is a platform for obtaining multiple simultaneous quantitative measurements of MR properties (1), and has been used for direct calculation of T1 and T2 (1,2,3) maps as well as tissue fraction maps for gray matter (GM), white matter (WM), and CSF (4,5). Another possibility is that these coregistered maps can be used to produce single images in which color is used to encode the multidimensional quantitative information, to ease visual interpretation. In this study, a color conversion is derived, in which normal T1 and T2 values in GM, WM and CSF are mapped to grayscale values, and areas where the relaxation parameters vary significantly from normal appear colored. This algorithm was used to generate colored images in both normal volunteers and patients with brain tumors, and the method was found to be reproducible and sensitive to tissue changes. The combination of the MRF sequence with the color display shown here has the potential to simplify both the acquisition and interpretation of clinical brain exams.


Experiments were performed on a 3T Siemens Verio and Skyra (Siemens, Erlangen, Germany) on a healthy volunteers and brain tumor patients (n=6). Using previously published MRF acquisition and reconstruction methods (1,3), T1 and T2 maps were generated by matching MRF time courses to a dictionary of signals that encompasses a large range of parameter values. Tissue fractions of GM, WM, and CSF were computed using a 3-component decomposition as previously described in (4,5). For generation of the hybrid color images, each property map was transformed to values between 0 and 1. The respective parameter values found in WM, GM and CSF were then mapped to predefined target intensities of 0.7, 0.5 and 0.1, respectively, using a cubic interpolation, thus creating an image with a T1-weighted appearance. While multiple combinations of parameters can be visualized with this method, two initial results are shown here. For Figure 1, an inversion recovery T1-weighted intensity map was used for the green color component, T2-weighted intensity map was used for blue, while the average intensity of all the IR and T2 parameters was assigned to the red color channel. For Figure 2, green and blue channels were similar to Figure 1, but a weighted average of the WM and GM fraction maps was assigned to the red color channel.


Example quantitative T1 and T2 maps and a hybrid color image from a normal volunteer are shown in Figure 1. It can be seen that normal brain tissue appears similar to a traditional T1- weighted image. However, blood vessels appear red and the putamen appears slightly blue, demonstrating the desired color behavior in tissues that do not have the reference GM, WM or CSF relaxation rates. Figure 2 shows a color image of a patient with a glioblastoma (GBM) using the second combination of factors. In this example, the lesion appears in bright purple and red shades, and is obvious in comparison to healthy tissues.

Discussion and Conclusion

This work demonstrates a potential method for the calculation of color images based on T1, T2 and tissue fraction maps. The goal is to obtain images in which brain tissue with normal relaxation appears nearly in grayscale, while any deviation from normal results in a residual color. These maps have the potential to simplify the interpretation of clinical brain MR, since information about three of the most relevant parameters are presented in a single, easy to interpret view. This type of display is particularly useful in combination with MRF, since the impact of misregistration due to interscan motion and differences in slice profiles between maps is eliminated. Further optimization and extensive testing in patients will be necessary to determine robustness in a wide variety of clinical settings. Other schemes to map normal tissue are possible, but the concept of mapping normal tissue to greyscale while channeling pathology to color could provide potentially clinically relevant information and easy interpretation of quantitative maps.


The authors would like to acknowledge funding from Siemens Healthcare and NIH grants NIH 1R01EB016728-01A1 and NIH 5R01EB017219-02.


(1) Ma D. et. al. Nature (2013) 495, 187–192.
(2) Badve C, et al. ISMRM 2015, pg 2254.
(3) Jiang Y. et. al. Magn. Reson. Med. (2014). doi: 10.1002/mrm.25559.
(4) Deshmane, A, et. al. ISMRM 2014, pg94.
(5) Deshmane, A et al. ISMRM 2015, pg 71.


(Top) Hybrid color visualization image. (Center and Bottom) Quantitative T1 and T2 Maps acquired using MRF.

Hybrid color visualization of a tumor (Glioblastoma) using quantitative T1, T2, and tissue fraction percentages.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)