3D Texture Analyses of Quantitative Susceptibility Maps to Differentiate Alzheimer’s Disease from Cognitive Normal and Mild Cognitive Impairment
Eo-Jin Hwang1, Hyug-Gi Kim1, Chanhee Lee1, Hak Young Rhee2, Chang-Woo Ryu1, Dal-Mo Yang1, Tian Liu3, Yi Wang3, and Geon-Ho Jahng1

1Department of Radiology, Kyung Hee University Hospital at Gangdong, Kyung Hee University, Seoul, Korea, Republic of, 2Department of Neurology, Kyung Hee University Hospital at Gangdong, Kyung Hee University, Seoul, Korea, Republic of, 3Biomedical Engineering and Radiology, Cornell University, New York, NY, United States

Synopsis

To investigate QSM textures in the subjects with cognitively normal (CN), mild cognitive impairment (MCI) and Alzheimer’s disease (AD) and to compare the QSM texture results with those of 3D T1W images, 18 elderly CN, 18 MCI, and 18 AD subjects were scanned both 3D multi-echo gradient-echo and 3D T1-weighted sequences. The 1st and 2nd ordered texture parameters of the QSMs and 3DT1W images were calculated and compared among the three subject groups to differentiate the subject groups. Our results suggest that the demyelination effect could be more dominant than the metal accumulations in AD progression.

Background and Purpose

Target Audience: Clinicians and Physicists who work for neurodegenerative diseases

Background: Quantitative susceptibility map (QSM) enables quantifying susceptibility-changing materials within a magnetic field (1). Texture analysis of MR images provides quantitative means for describing tissue properties and physiological and pathological stages in order to reveal overall information about the images that is often invisible to the naked eye (2). Texture analyses of Alzheimer’s disease (AD) have been previously run on 3D T1-weighted (T1W) images (2,3), but not on QSM data.

Purpose: To investigate QSM textures in subjects with cognitive normal (CN), mild cognitive impairment (MCI), and AD and to compare the results with those of 3D T1W images.

Materials and Methods

The study was approved by the local institutional review board, and informed consent was obtained from all subjects. The participants were 18 elderly CN, 18 MCI and 18 AD subjects. A fully first-order flow-compensated 3D gradient-echo (GE) sequence was run to obtain axial magnitudes and phase images and to produce QSM data. Sagittal structural 3D T1W (3DT1W) images were also obtained with the magnetization-prepared rapid acquisition of GE sequence to obtain brain tissue images. To generate the QSMs, the magnitude and phase images acquired from the 3D GE sequences were further processed using morphology enabled dipole inversion (MEDI) (4). The 1st and 2nd ordered texture parameters of the QSMs and 3DT1W images were obtained using MaZda software (http://www.eletel.p.lodz.pl/programy/mazda/, Lodz, Poland) to evaluate group differences using a one-way analysis of covariance.

Results

Figure 1 and 2 show results of the 1st order texture analysis of QSM (Fig.1) and the 3 dimensional T1 weighed (3DT1W) (Fig.2). For the first-order QSM analysis, mean, standard deviation (SD) and covariance of signal intensity (COVSI) separated the three subject groups (F = 5.191, p = 0.009). For the 3DT1W images, the means showed no significant differences among the three subject groups (p > 0.07). However, the SD and COVSI showed a significant difference between the subject groups. For the 2nd order QSM textures, AngScMom, contrast, correlation and DifVarnc showed significant differences among the groups. In contrast, for the 2nd order 3DT1W image texture, AngScMom, entropy, InvDfMom and SumEntrp showed significant differences.

Discussion

Because AD is associated with biological substances that are susceptible to magnetic fields, QSMs can provide us with novel knowledge of how the distribution of susceptibility-inducing materials would conform to the pathology. It should be noted that the QSMs and 3DT1W images responded independent of one of each other. QSM mean increased linearly from CN to AD, whereas no linear transitions were observed in the 3DT1W images except for HC, in which the value progressively declined from CN to AD. Our results suggest that the demyelination effect could be more dominant than the metal accumulations in AD progression.

Conclusion

This was the first and the only study to evaluate the textures of QSMs in AD. MCI was better characterized by the QSM textures, which displayed more consistent transitions from CN and AD than did the 3DT1W images. WM QSM means and COVSIs successfully differentiated MCI from CN and AD from CN.

Acknowledgements

This research was supported by a grant of the Korean Health Technology R&D Project, Ministry for Health, Welfare & Family Affairs, Republic of Korea (A092125) and the Basic Science Research Program through the National Research Foundation of Korea (2014R1A2A2A01002728).

References

1. F. Schweser, et al., " Neuroimage, vol. 62, 2083-2100; 2. J. Zhang, et al., Brain Imaging Behav, vol. 6, 61-69; 3. P. A. Freeborough and N. C. Fox, IEEE Trans Med Imaging, vol. 17, 475-479; 4. T. Liu, et al., Magn Reson Med, vol. 66, 777-783.

Figures

The 1st order texture for QSM data

The 1st order texture for 3DT1WI data



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1255