Computed FLAIR-DWI Technique combined with DWI, PDW, T2W and T1W Imagings
Yuki Takai1 and Tokunori Kimura2

1MRI development department, Toshiba Medical Systems, Otawara, Japan, 2Clinical Application Research and Development Dept., Toshiba Medical Systems, Otawara, Japan

Synopsis

We proposed a new computed DWI technique allowing to provide quantitative maps of ADC, T2, T1 and DWI images with arbitrary contrast of PDW, T2W, and T1W. We clarified that our techniques enabled to provide FLAIR-DWI images with higher SNR than actually measured FLAIR-DWI, and furthermore enhanced by using SE-based ADC map and SE- or IR-based T1 map with optimal TI for brain tissue. Although further optimization is required, it is expected to be especially useful for clinical brain diagnosis.

Purpose

A computed diffusion imaging (cDWI), which allows to provide high b-value (b) equivalent DWI images using relatively low b DWI images, was proposed and applied to improve contrast between tumors and the background tissue especially in body diffusion imaging [1-3]. More recently a short-TE cDWI technique was proposed to reduce T2 shine-through effects or to further enhance tissues with short T2 and lower ADC, where DWI images with arbitrary combination of TE (including zero) and b can be generated [4]. On the other hand, a synthetic MRI technique has proposed and gathering attention where spin echo (SE)-or inversion recovery (IR)-based images with arbitral imaging parameters with shorter acquisition time than actual imaging [5]. The purpose of this study was to propose a new computed DWI technique to generate arbitral combination of imaging parameters of TR (or TI), TE and b in combination of 4 kinds of acquired images, and to assess for volunteer brain imaging. Here Fluid-Attenuated IR (FLAIR)-DWI imaging was particularly focused on as an application of this technique.

Methods

Theory

Spin echo (SE)-based measured DWI (mDWI) signal intensity at TR, TE and b for tissue of T1, T2 and ADC is modeled by using arbitrary coefficient k as:

S(TR,TE,b)=k*(1-exp[-TR/T1])*exp[-TE/T2]*exp[-b*ADC] ---- (1),

or IR-based mDWI signal intensity at TI, TE, and b when TR is infinity is modeled as:

S(TI,TE,b)=k*(1-2 exp[-TI/T1])*exp[-TE/T2]*exp[-b*ADC] ---- (2).

Algorithm for 4-point cDWI method based on these models was shown in Fig. 1. Here following methods for computed FLAIR-DWI were compared including measured FLAIR-DWI.

a) FLAIR-mDWI: 1-point directly acquired IR signal at T2W-FLAIR DWI condition was assumed.

b) ADCbyFLAIR-cDWI: ADC was calculated with 2 points FLAIR signals setting at TI1=T1 of CSF.

c) ADCbySE&T1bySE FLAIR-cDWI: ADC was calculated with 2 points T2W-SE signals. T1 was calculated using 2 points signals T1W-SE setting at TR1=T1 of brain tissue and PDW-SE.

d) ADCbySE&T1byIR FLAIR-cDWI: ADC was calculated with the same way as method-c. T1 was calculated using 2 points signals of T1W-IR setting at TI1=T1 and PDW-SE.

Monte-Carlo simulation

Assuming Rician distribution of noise on the magnitude DWI signals, the mean and the SD of quantitative parameters and cDWI signals were measured after 5000 times each trial for 4 methods using the parameters shown in Table 1. The k in Eq.(1-2) was 1 and the added Gaussian noise SD for mDWI was 0.02. Here TE was commonly set at TE2 for T2W condition. The mean, SD and SNR assuming brain tissue were compared.

Volunteer study

Imaging was performed on 3T MR imager of Toshiba Vantage Titan 3T with a single-shot SE-EPI with the same parameters in Table 1 except for TI1=2000ms and TE2=80ms. Motion probing gradient (MPG) was applied to a direction perpendicular to the running direction of targeted fibers (corpus callosum (CC)). The cDWI images of PDW and T2W each with bc=0, 1000 and 2000 s/mm2 were calculated, where TIc was experimentally decided at 2000ms.

Results

For simulations, the background noise SDs were increased in the order of method d <c <b <a (Fig. 2) reflecting noise propagation effects of quantitative parameters. Table 2 shows the summary of the simulation. As the SDs in ADC and in R1 were smaller, SDs in cDWI signals became smaller; i.e., the SDs in ADC obtained by SE (b, c, and d) were smaller than by the T2W-FLAIR (a), and the SDs in R1 obtained by IR(d) were less than those by SE(c). In addition, noise bias effects on DWI signals due to Rican noise were significant in high b for FLAIR-mDWI. Regarding MRI experiments (Fig. 3), FLAIR image of PDW (TEc=28ms) and T2W (TEc=80ms) were shown in addition to measured images and quantitative maps. Brain tissue provided sufficient SNR even using T1bySE method despite of poor T1W image. The SNR for the R1 map with T1byIR was better than with T1bySE, and therefore those effects reflected on the corresponding cDWI images. It was good news that the CSF signals in FLAIR-cDWIs became negative without using phase correction even using T1bySE.

Conclusion

We proposed a new computed DWI technique allowing to provide quantitative maps of ADC, T2, T1 and DWI images with arbitrary contrast. We clarified that our techniques enabled to provide FLAIR-DWI with higher SNR than actually measured FLAIR-DWI suffering from SNR compared to SE-DWI, and furthermore enhanced by using SE-based ADC map and SE- or IR-based T1 map with optimal TI for brain tissue. Although further optimization or systematic design of software including user interface are required, it is expected to be especially useful for clinical brain diagnosis.

Acknowledgements

No acknowledgement found.

References

[1] Blackledge MD, Leach MO, Collins DJ, et al. Computed Diffusion-weighted MR Imaging May Improve Tumor Detection. Radiology: 261,573-581(2011).

[2] Ueno Y, Takahashi S, Kitajima K, Kimura T et al. Computed diffusion-weighted imaging using 3-T magnetic resonance imaging for prostate cancer diagnosis. Eur Radiol 23,3509-16 (2013).

[3] Kimura T, Machi Y. Computed Diffusion Weighted Imaging Under Ricain Noise Distribution. In: Proceedings of the 20th Annual Meeting of ISMRM, 2012;p3574.

[4] Kimura T, Machi Y et al. A short-TE Computed Diffusion Imaging (cDWI). In: Proceedings of the 23th Annual Meeting of ISMRM, 2015;p2929. [5] Warntjes JBM et al. Rapid Magnetic Resonance Quantification on the Brain. Magn Reson Med 60:320–329 (2008).

Figures

Fig. 1 SE-based and IR-based cDWI calculation flow (4 points method)

Fig. 2 Simulated Noise SD ratio vs. b for 3 kinds of FLAIR-cDWI methods.

Parameters were shown in Table 1. Note that noises in the c)ADCbySE_T1bySE and d)ADCbySE_T1byIR were reduced than in the b)ADCbyFLAIR even at bc=1000. However, the difference between c and d were reduced with increasing bc. Comparing method-c and d, SDs of cDWI signals in method-d was smaller at bc<1000 (b2) than in method-c; however, the differences became smaller at bc>1000


Table 1 Parameters for 4 kinds of measured and computed FLAIR-DWI methods


Table 2 Simulation Results for 4 kinds of FLAIR-cDWI methods at Table 1 condition

Regarding the quantitative tissue parameters, the ADC SD with method-b was smaller than with the method-a and the R1SD with method-d was smaller than with the method-c. Regarding DWI signals, SDs were smaller in the order of method-a, b, c, d reflecting noise propagation effects from the ADC and R1. For FLAIR-mDWI, bias effects on DWI signals due to Rican noise were significant in high b.



Fig. 3 Imaging results for brain FLAIR-cDWI method

cDWI images of PDW-FLAIR and T2W-FLAIR each with 3 methods, measured images, and quantification maps were shown. SNR for R1 map by IR was better than by SE similarly as simulation. The image contrasts among different methods of b=0 were slightly different. Further validation is required.




Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3047