Diffusion kurtosis imaging and diffusion-weighted imaging-based classication of acute stroke lesions in the brain: A pilot study

Background: We postulated that diffusion kurtosis imaging (DKI) could classify heterogeneous stroke lesions on diffusion-weighted imaging (DWI) and improve our understanding of the characteristics of tissue injury. We aimed to retrospectively study different DKI parameters in patients with acute stroke reported in the literature. Methods: We collected the DWI and DKI data of 41 patients (26 men, 15 women), including 86 cases of acute cerebral infarction in different brain regions. Of them, 20 patients had single infarction, whereas others had multiple infarctions. Acute cerebral infarction lesions were classied into two categories based on DKI and DWI parameters: type I, matched DKI and DWI parameters and type II, mismatched DKI and DWI parameters. Regions of interest (ROIs) were outlined within the most severely infarcted areas of each lesion according to each independent parametric map. In the control groups, same-sized ROIs were located in the corresponding region of the normal contralateral hemisphere. In both categories, DKI and DWI parameters followed a normal Gaussian distribution. We used the independent sample t-test to compare the differences in each group. Results: In type I cases, fractional anisotropy, mean diffusivity, axial diffusivity, radial diffusivity, mean kurtosis (MK), and axial kurtosis (Ka) values were signicantly different (P<0.05). In type II cases, only MK and Ka values were signicantly different (P<0.05). Conclusions: DKI can provide more information on acute ischemic brain infarction and enrich our understanding of ischemic tissue injury. This DKI and DWI parameters-based classication of acute stroke lesions may confer a renewed understanding of infarction cores.


Background
Diffusion-weighted magnetic resonance imaging (DWI) has become an integral part of the clinical evaluation of acute stroke because of its high detection sensitivity for cerebral infarctions 1 . Although medical imaging and treatment have advanced, current research is not limited to the detection of ischemic lesions and aimed at distinguishing reversible and irreversible brain damage.
Previous studies suggest that the mismatch between DWI and perfusion-weighted imaging showed marked sensitivity to ischemic brain injury ( Fig. 1) 1,2 . Nevertheless, the signi cance of tissue mismatch, including those of oligemic, salvageable, and irretrievable tissue, is unclear. These tissue characteristics may depend on various intricate factors, such as the duration and nature of the ischemic injury, proximity of the patent vessels, and so on. In conclusion, perfusion-diffusion mismatch can only approximate the features of real ischemic penumbra. Despite the limitations, this mismatch often guides clinical decisions in the management of acute stroke 3 . Therefore, presently, additional methods are being explored to provide a profound knowledge of ischemic brain injury.
Diffusion kurtosis imaging (DKI) is a promising new DWI sequence that measures the non-Gaussianity of water diffusion and can be used to characterize the complexity or heterogeneity of tissue microenvironments with reduced imaging time, hardware requirements, and post-processing effort compared with 3-dimensional q-space imaging 4 . Indeed, DKI has been reported to detect microstructural cerebral changes in aging brains, acute stroke, and tumors.
We postulated that DKI could stratify heterogeneous DWI lesions and improve the characterization of tissue injury. The spatiotemporal dynamics of DKI in patients with acute stroke have been studied systematically in the literature.

Ethics Statement
The study was approved by the institutional review board of the Second A liated Hospital of Xiamen Medical University. The patients (conscious patients) or their lineal relatives (for the unconscious patients) provided written informed consent. The study adhered to the tenets of the Declaration of Helsinki.

Subjects
Overall, 41 patients (26 men, 15 women) with 86 different cases of acute cerebral infarction in different areas of the brain were recruited for this study. Twenty patients had single infarction, whereas the others had multiple infarctions. All patients underwent MRI scans within 72 h of stroke onset. The inclusion criterion was presence of acute cerebral infarction. The exclusion criteria were presence of cerebral hemorrhage (original or secondary), cerebral tumors, degenerative brain disease, craniocerebral trauma, post-craniocerebral operation, dyspnea, and coma.

Classi cation
Herein, we classi ed acute cerebral infarction lesions into two categories according to the manifestations of DKI and DWI parameters. Type I lesions had matched DKI and DWI parameters, and type II lesions had mismatched DKI and DWI parameters. This study included 45 and 41 cases of type I and type II lesions, respectively ( Fig. 2-3).

Imaging acquisition
All images were acquired using a GE 1.5-Tesla HDx Echo Speed Plus MRI scanner with an 8-channel head coil (GE Healthcare Life Sciences, Chalfont, UK). The scanning sequences included T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), T2-weighted uid-attenuated inversion recovery (T2-FLAIR) imaging, DWI, and DKI. The b-value of the DWI sequence was 1000s/mm 2 . The DKI sequence parameters were as follows: repetition time (TR), 6000 ms; echo time (TE), minimum; layer thickness, 5 mm; layer spacing, 1.5 mm; eld of view (FOV), 240 × 240 mm 2 ; matrix, 96 × 130; layer number, 19; diffusion direction, 15; b = 0, 1000, and 2000s/mm 2 ; and scanning time, approximately 6 m and 18 s. Imaging analysis DKI post-processing software provided by GE was used for the post-processing of DKI images. Regions of interest (ROIs; maximal area in the center of the lesion) were outlined within the most severely infarcted areas of each lesion according to each independent parametric map. In the control groups, same-sized ROIs were located in the corresponding normal regions of the contralateral hemisphere (Fig. 4). Multiple DKI parameters [mean kurtosis (MK), radial kurtosis (Kr), and axial kurtosis (Ka)] were measured for all ROIs. Two radiologists independently diagnosed and analyzed the ischemic lesions. In case of disagreements, a third radiologist was consulted.

Statistical analysis
The relative diagnostic value of DKI parameters was analyzed with the independent sample t-test.
Statistical analyses were performed using SPSS 13.0 software (SPSS, Inc., Chicago, IL, USA). P-values < 0.05 were considered statistically signi cant.  Table 1).  A previous study showed that in an experimental model of traumatic brain injury, the MK elevates in neural tissues subjected to subacute stage infarction despite pseudonormalization of MD and FA values 9 . An increase in the MK was associated with increased reactive astrogliosis, which con rms that K is a biomarker of tissue heterogeneity. Similar results were also observed in previous preliminary studies of animals and humans with stroke using DKI.

Patient characteristics
In other words, K is an in vivo measurement of the complications or/and inhomogeneity of the microenvironment of stroke lesions. It provides complementary information to conventional diffusion indices and maybe a more sensitive biomarker to assess the pathophysiological changes associated with stroke 10 .
The results of our study imply that DWI/DKI mismatch may be indicative of a mildly damaged and potentially recoverable ischemic lesion, whereas the matched areas may be indicative of irreparable cellular damage. Here, the type II stroke lesions showed signi cantly different MK and Ka values, which may be an indicative mismatch of the real infarction core in patients with acute stroke.
Conclusions DKI can provide more information than DWI about acute ischemic infarction lesions of the brain and facilitate improved understanding of ischemic tissue injuries. Our method of classi cation of acute stroke lesions based on their DKI and DWI parameters may enable further understanding of the real infarction core.  The red circle (A) indicates the store lesion seen on the DWI scan, while the blue circle (B) indicates the store lesion seen on the PWI scan; C indicates the mis-match of the DWI and PWI. ROIs (maximal areas in the center of the lesions) were outlined within the most severe areas of each lesion according to each independent parametric map. Same-sized ROIs were located in the corresponding normal brain regions of the contralateral hemisphere. ROI, region of interest