Development of a quantitative statistical analysis system for double inversion recovery (DIR) MRI: A preliminary clinical study

Abstract

BACKGROUND:

Gray matter (GM) imaging is important in the investigation of many neurological diseases, including schizophrenia, multiple sclerosis, stroke, Alzheimer’s disease, tuberous sclerosis, and epilepsy, which are all associated with changes in cortical GM.

OBJECTIVE:

The aim of this study was to develop a quantitative statistical analysis system for double inversion recovery (DIR) MRI and to evaluate the new system using preliminary clinical data.

METHODS:

The study population comprised of 10 healthy volunteers and six patients with or without brain degeneration. A quantitative statistical analysis system for DIR images was developed using the following steps: 1) brain spatial normalization, 2) mean and standard deviation (SD) map creation, and 3) Z-score map creation. To evaluate the new voxel-based morphometry system, Z-scores of lesions in patients with brain degeneration were measured and then compared with Z-scores of normal regions.

RESULTS:

All DIR images were adequately spatially normalized to Montreal Neurological Institute MNI coordinate. Lesions in each patient were indicated by high Z-score values on a Z-score map, which were significantly higher than Z-scores of normal regions ( $p<$ 0.05).

CONCLUSIONS:

In this study, we developed a quantitative statistical analysis system for DIR MRI. Using our system, clinicians might accurately diagnose early brain degeneration.

Keywords

Gray matter Alzheimer’s disease spatial analysis neurodegenerative diseases

1. Introduction

The double inversion recovery (DIR) method is an imaging method that adds an additional 180 ${}^{\circ}$ inversion pulse to the conventional inversion recovery (IR) method [1]; two inversion times (TIs) are set: TI-1st and TI-2nd, separated by two 180 ${}^{\circ}$ inversion pulses. By appropriately setting these two inversion times, according to the T1 relaxation time of the target tissue, the signals of these two tissues can simultaneously be made null. In the central nervous region, we can obtain white matter (WM) attenuated inversion recovery (WAIR) images, with suppression of cerebrospinal fluid and white matter, using the DIR method; similarly, we can obtain gray matter (GM) attenuated inversion recovery (GAIR) images, with suppression of cerebrospinal fluid and gray matter [2].

WAIR images, obtained by the DIR method, have been useful in diagnosis of brain degeneration, as in multiple sclerosis (MS) [3, 4, 5]. During detection of MS lesions, the DIR method has been reported as more useful than the fluid attenuated inversion recovery (FLAIR) method [5, 6]. Also, in WAIR images that are obtained by the DIR method, the GM region is selectively emphasized, and in GAIR images, the WM region is selectively emphasized [2]. Therefore, white matter and gray matter can be extracted without post-processing of images [7]. Further, the influence of the partial volume effect is small when extracting WM and GM in voxel-based morphometry (VBM).

VBM is clinically used as a quantitative and valuable imaging method for early diagnosis of Alzheimer’s disease (AD) [8, 9, 10]. In Japan, a local analysis system for Alzheimer’s disease was developed, based on voxels (VSRAD). This software allows quantitative analysis of the extent of hippocampal atrophy in a voxel fashion, using T1-weighted MR images, and has been used for clinical evaluation [11, 12, 13, 14]. However, these VBM systems cannot detect early brain degeneration because they utilize T1-weighted MR images; thus, it is difficult to detect WM lesions, such as those present in MS. DIR images can detect AD, as well as degenerative brain diseases with white matter lesions (e.g., MS), which are visualized as abnormal findings on the image. Therefore, if a VBM system is developed for DIR images, quantitative assessment may be available for more brain diseases, facilitating earlier diagnosis and quantitative judgment of therapeutic effects. The aim of this study was to develop a quantitative statistical analysis system for double inversion recovery (DIR) MRI and to evaluate the new system using preliminary clinical data.

2. Materials and methods

Figure 1 depicts a flowchart of a quantitative statistical analysis system for DIR images, which was developed by Z-score mapping technique. This system indicates an abnormal region as a high Z-score on each voxel. To evaluate the new VBM system, patients with brain degeneration were examined and Z-scores of their brains were compared with Z-scores of normal regions. Image processing was performed by a custom-designed program, using MATLAB 2017b (The MathWorks Inc., Natick, MA, USA) and normalization was performed by Statistical Parametric Mapping 12 (SPM12) (Wellcome Department of Imaging Neuroscience, London, UK).

Figure 1.

A flowchart representing our quantitative statistical analysis system.

2.1 Population

The study population comprised of 10 healthy volunteers (six men, four women; age range, 21–39 years; average age, 25.2 years) and six patients (three men, three women; age range, 14–90 years; average age, 63.5 years) (one patient with no abnormalities and five patients with brain degeneration). Volunteers and patients both underwent MRI using a 1.5-T or 3.0-T MRI scanner (Ingenia 1.5T or Ingenia 3.0T; Philips Medical Systems, Best, The Netherlands). All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants included in the study. This article does not contain any studies with animals performed by any of the authors.

2.2 MRI scanning

In this study, the DIR imaging method was optimized using only volunteer MR images. WAIR images of patients were obtained using the optimized method. However, it is difficult to optimize typical parameters for individual DIR MR imaging. Therefore, we optimized imaging parameters by applying previously published optimization methods for obtaining WAIR images [2, 15]. Scanning parameters on the 3.0-T scanner were as follows: TR $=$ 5500 ms, TE $=$ 286 ms, TI-1st $=$ 2069 ms, and TI-2nd $=$ 480 ms; parameters on the 1.5-T scanner were as follows: TR $=$ 8000 ms, TE $=$ 73 ms, TI-1st $=$ 2630 ms, TI-2nd $=$ 372 ms. Other scanning parameters were as follows: FOV $=$ 240 $\times$ 240 mm; matrix size $=$ 240 $\times$ 240; slice thickness $=$ 1.2 mm.

2.3 Image analysis for Z-score map

For calculation of the Z-score map, we applied the method used to create Z-Score maps for CT images, as reported by Takahashi et al. [16, 17, 18]. Using WAIR images of volunteers and patients, the new VBM system for WAIR images follows these steps: 1) brain spatial normalization to Montreal Neurological Institute (MNI) coordinates, 2) brain region segmentation, 3) mean and standard deviation (SD) map creation, and 4) Z-score map creation. The new VBM system indicated abnormal regions as high Z-scores on each voxel.

2.4 Normalization to MNI space for WAIR images

Normalization to MNI space was achieved using the normalization function of SPM12 for all volunteers and patients. Example of original and normalized images are shown in Fig. 2a and b.

Figure 2.

Representative image analysis in our quantitative statistical analysis system (a) original image, (b) image normalized to Montreal Neurological Institute (MNI) space, (c) gray matter (GM) segmented image, (d) white matter (WM) segmented image, (e) mean map, (f) standard deviation (SD) map.

2.5 Segmentation of the brain region

For normalized images, brain regions (WM, GM) were segmented using the Segment function of SPM12. Using the segmented WM and GM map as a mask image, only the brain region was segmented from the normalized image. Examples of the extraction of brain regions (WM and GM) are shown in Fig. 2c and d.

2.6 Creation of mean map and standard deviation (SD) map

Mean and SD of image voxel values were calculated from healthy volunteer data sets of WAIR images of segmented brain regions, in units of voxels, to create mean maps and standard deviation maps. Examples of average value and SD maps are shown in Fig. 2e and f.

2.7 Calculation of the Z-score map [16, 17, 18]

Z-score was calculated in units of voxels and was defined as follows:

$\text{Zcore}\ w(x,y,z)=(C_{\text{mean}}(x,y,z)-\text{Input}(x,y,z))/N_{\text{% SD}}(x,y,z)$

where $C_{\text{mean}}$ ( $x, y, z$ ) and $N_{\text{SD}}$ ( $x, y, z$ ) represent creation averages at the coordinates ( $x, y, z$ ) and mean and SD of the SD map, respectively. Input ( $x, y, z$ ) is the value of the spatially normalized dataset in the same coordinate system. Finally, the Z-score map was superimposed on the normalized WAIR data set and color-coded to reflect the range of Z-scores.

2.8 Clinical evaluation

In order to evaluate the developed DIR-VBM system, we evaluated the Z-score map using five clinical images. We examined whether the Z-score of the site where an abnormality was identified in the WAIR image, on diagnostic imaging by the radiologist, was significantly different from the Z-score of the normal region. Statistical analysis was performed using commercial software (Prism 5; GraphPad Software, Inc., San Diego, CA, USA). Z-score values between lesion and normal regions were compared using a Wilcoxon rank-sum test.

Figure 3.

A normalized double inversion recovery (DIR) image and the Z-score map in a patient with no abnormalities.

3. Results

Figure 3 shows a normalized DIR image and the Z-score map in a patient with no abnormalities. There was no region of high Z-score on the map. Figure 4 shows a normalized DIR image and the Z-score map in a patient with cerebral infarction. On the fusion image, red regions indicate high Z-score regions (Z-score $>$ 10), and the blue regions indicate low Z-score regions (Z-score $<$ $-$ 5). Figure 5 shows a normalized DIR image and the Z-score map in a patient with spinocerebellar degeneration. On the fusion image, red regions indicate high Z-score regions (Z-score $>$ 10), and the blue regions indicate low Z-score regions (Z-score $<$ $-$ 5). Z-scores of lesions of patients with brain degeneration were significantly higher than Z-scores of normal regions of all volunteers and patients ( $p<$ 0.05).

Figure 4.

A normalized double inversion recovery (DIR) image and the Z-score map in a patient with cerebral infarction.

Figure 5.

A normalized double inversion recovery (DIR) image and the Z-score map in a patient with spinocerebellar degeneration.

4. Discussion

Using WAIR images obtained using the optimized imaging method, we developed a new VBM system for WAIR images, which can spatially normalize WAIR images to the MNI coordinate and can calculate the Z-score map for each patient. Moreover, Z-score calculated by the VBM system can emphasize lesions, including brain degeneration, on WAIR images. Evaluation of degenerated brain regions can be performed more accurately with WAIR images than FLAIR images [5, 6]. However, even for WAIR images, the evaluation is subjective (performed by the clinician); thus, if evaluation can be performed with a quantitative value, such as a Z-score, clinicians may quantitatively judge therapeutic effects. Further, by using quantitative images, clinicians might accurately diagnose brain degeneration at the stage of small lesions, and a computer-aided diagnostic (CAD) system might be developed for patients with brain degeneration.

Numerous studies have been conducted on the ability of VBM analysis of T1-weighted MRI to diagnose early AD [8, 9, 10]. However, since T1-weighted MR images largely rely on evaluation of brain atrophy, they do not greatly facilitate earlier AD detection. Therefore, to diagnose AD at a very early stage, attempts have been made to detect microinfarction with WAIR images [19, 20]. WAIR is also useful for detection of WM lesions, and provides significant clinical utility [4, 5, 21]. However, since WAIR and T1-weighted images utilize different contrasts, there have not yet been reports on VBM system development and clinical evaluation for WAIR images. Therefore, in this research, we developed a VBM system to use in analyzing WAIR images, using Takahashi’s Z-score mapping method. Using the DIR-VBM system developed in this study, we were able to quantitatively detect lesions during early clinical evaluations. Thus, we believe that the system developed in this study will enable diagnosis of very early AD and quantitative assessment of degenerative brain diseases.

There are several limitations in this research. Firstly, the quantity of WAIR data is small. Since the purpose of this research was systems development, the quantity of data in the Mean and SD maps was small and the age range of the patients was narrow. Therefore, in the future it will be necessary to accumulate clinical data and create Mean and SD maps from patients across a wider range of ages. Moreover, in current clinical evaluation, this system cannot classify lesions according to disease; hence, we plan to examine the clinical usefulness of this system using additional clinical data.

5. Conclusion

In this study, we developed a quantitative statistical analysis system for DIR MRI. This system can evaluate any lesions with abnormally high intensities on WAIR images, using Z-score values to quantify each voxel of an entire brain region. Using our system, clinicians might accurately diagnose early brain degeneration; additionally, our system can be used to develop a computer-aided diagnostic (CAD) system for patients with brain degeneration.

Footnotes

Acknowledgments

This work was supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI Grant (Number JP15K19206).

Conflict of interest

None to report.

References

Redpath

Smith

. Technical note: use of a double inversion recovery pulse sequence to image selectively grey or white brain matter. Br J Radiol. 1994 Dec; 67(804): 1258-63.

Motegi

Shimada

Hayashi

Nagase

Taketomi-Takahashi

Tsushima

. Double inversion recovery imaging of the brain: deriving the most relevant sequence through real images. Radiol Phys Technol. 2017 Apr 9; 10(3): 364-75.

de Graaf

Zwanenburg

Visser

Wattjes

Pouwels

Geurts

, et al. Lesion detection at seven Tesla in multiple sclerosis using magnetisation prepared 3D-FLAIR and 3D-DIR. Eur Radiol. 2012 Jan; 22(1): 221-31.

Simon

Schmidt

Lukas

Gieseke

Traber

Knol

, et al. Improved in vivo detection of cortical lesions in multiple sclerosis using double inversion recovery MR imaging at 3 Tesla. Eur Radiol. 2010 Jul; 20(7): 1675-83.

Vural

Keklikoglu

Temel

Deniz

Ercan

. Comparison of double inversion recovery and conventional magnetic resonance brain imaging in patients with multiple sclerosis and relations with disease disability. Neuroradiol J. 2013 Apr; 26(2): 133-42.

Seewann

Kooi

Roosendaal

Pouwels

Wattjes

van der Valk

, et al. Postmortem verification of MS cortical lesion detection with 3D DIR. Neurology. 2012 Jan 31; 78(5): 302-8.

Bedell

Narayana

. Volumetric analysis of white matter, gray matter, and CSF using fractional volume analysis. Magn Reson Med. 1998 Jun; 39(6): 961-9.

Busatto

Diniz

Zanetti

. Voxel-based morphometry in Alzheimer’s disease. Expert Rev Neurother. 2008 Nov; 8(11): 1691-702.

Hirata

Matsuda

Nemoto

Ohnishi

Hirao

Yamashita

, et al. Voxel-based morphometry to discriminate early Alzheimer’s disease from controls. Neurosci Lett. 2005 Jul 15; 382(3): 269-74.

10.

Baron

Chetelat

Desgranges

Perchey

Landeau

de la Sayette

, et al. In vivo mapping of gray matter loss with voxel-based morphometry in mild Alzheimer’s disease. Neuroimage. 2001 Aug; 14(2): 298-309.

11.

Sugiyama

Sato

Kimura

Maekawa

Wakasugi

Sone

, et al. Thalamic involvement determined using VSRAD advance on MRI and easy Z-score analysis of (99m)Tc-ECD-SPECT in Gerstmann-Straussler-Scheinker syndrome with P102L mutation. J Neurol Sci. 2017 Feb 15; 373: 27-30.

12.

Matsuda

. MRI morphometry in Alzheimer’s disease. Ageing Res Rev. 2016 Sep; 30: 17-24.

13.

Nakatsuka

Imabayashi

Matsuda

Sakakibara

Inaoka

Terada

. Discrimination of dementia with Lewy bodies from Alzheimer’s disease using voxel-based morphometry of white matter by statistical parametric mapping 8 plus diffeomorphic anatomic registration through exponentiated Lie algebra. Neuroradiology. 2013 May; 55(5): 559-66.

14.

Goto

Suzuki

Abe

Hayashi

Aoki

Mori

, et al. Customization of normal data base specific for 3-tesla MRI is mandatory in VSRAD analysis. Radiol Phys Technol. 2008 Jul; 1(2): 196-200.

15.

Hayashi

Yarita

Sakata

Motegi

Nagase

Ujita

, et al. Optimization method of MRI scan parameters of a double inversion recovery sequence using a T1 map and a developed analysis algorithm. J Xray Sci Technol. 2017 May 6; 25(5): 803-12.

16.

Takahashi

Tsai

Lee

Kinoshita

Ishii

. Z-score mapping method for extracting hypoattenuation areas of hyperacute stroke in unenhanced CT. Acad Radiol. 2010 Jan; 17(1): 84-92.

17.

Takahashi

Tsai

Lee

Kinoshita

Ishii

Tamura

, et al. Usefulness of z-score mapping for quantification of extent of hypoattenuation regions of hyperacute stroke in unenhanced computed tomography: analysis of radiologists’ performance. J Comput Assist Tomogr. 2010 Sep-Oct; 34(5): 751-6.

18.

Takahashi

Kinoshita

Ohmura

Lee

Matsuyama

Toyoshima

, et al. Z-score-based semi-quantitative analysis of the volume of the temporal horn of the lateral ventricle on brain CT images. Radiol Phys Technol. 2016 Jan; 9(1): 69-76.

19.

Maeda

Kida

Matsuo

Shindo

Taniguchi

, et al. In vivo detection of cortical microinfarcts on ultrahigh-field MRI. J Neuroimaging. 2013 Jan; 23(1): 28-32.

20.

Kida

Satoh

Fukuyama

Maeda

Tomimoto

. Detection of cerebral amyloid angiopathy by 3-T magnetic resonance imaging and amyloid positron emission tomography in a patient with subcortical ischaemic vascular dementia. Psychogeriatrics. 2016 Jan 18.

21.

Geurts

Pouwels

Uitdehaag

Polman

Barkhof

Castelijns

. Intracortical lesions in multiple sclerosis: improved detection with 3D double inversion-recovery MR imaging. Radiology. 2005 Jul; 236(1): 254-60.