Diffusion kurtosis imaging provides independent and complementary information to that acquired with traditional diffusion techniques. The additional information is thought to indicate the complexity of the microstructural environment of the imaged tissue and may lead to broad-reaching applications in all aspects of neuroradiology 1).
Since tissue structure is responsible for the deviation of water diffusion from the Gaussian behavior typically observed in homogeneous solutions, this method provides a specific measure of tissue structure, such as cellular compartments and membranes. The method is an extension of conventional diffusion weighted imaging that requires the use of somewhat higher b values and a modified image postprocessing procedure. In addition to the diffusion coefficient, the method provides an estimate for the excess kurtosis of the diffusion displacement probability distribution, which is a dimensionless metric of the departure from a Gaussian form. From the study of six healthy adult subjects, the excess diffusional kurtosis is found to be significantly higher in white matter than in gray matter, reflecting the structural differences between these two types of cerebral tissues. Diffusional kurtosis imaging is related to q-space imaging methods, but is less demanding in terms of imaging time, hardware requirements, and postprocessing effort. It may be useful for assessing tissue structure abnormalities associated with a variety of neuropathology's 2).
A noninvasive diagnostic method to predict the degree of malignancy accurately would be of great help in glioma management. This study aimed to create a highly accurate machine learning model to perform glioma grading.
Preoperative MRI acquired for cases of glioma operated from October 2014 through January 2018 were obtained retrospectively. Six types of MRI sequences (T2-weighted image, diffusion-weighted image, apparent diffusion coefficient [ADC], fractional anisotropy, and mean kurtosis [MK]) were chosen for analysis; 476 features were extracted semi-automatically for each sequence (2,856 features in total). Recursive feature elimination was used to select significant features for a machine learning model that distinguishes glioblastoma from lower-grade glioma (grades II/III).
Fifty-five datasets from 54 cases were obtained (14 grade II gliomas, 12 grade III gliomas, and 29 glioblastomas), of which 44 and 11 datasets were used for machine learning and independent testing, respectively. We detected 504 features with significant differences (false discovery rate < 0.05) between glioblastoma and lower-grade glioma. The most accurate machine learning model was created using 6 features extracted from the ADC and MK images. In the logistic regression, the area under the curve was 0.90 ± 0.05 and the accuracy of the test dataset was 0.91 (10/11); using a support vector machine, they were 0.93 ± 0.03 and 0.91 (10/11), respectively (kernel, radial basis function; c, 1.0).
DK imaging is able to depict microstructural changes within glioma tissue and is able to help differentiate among glioma grades
DKI parameters can effectively distinguish between low- and high-grade gliomas 4).
Mean MK is the best independent predictor of differentiating glioma grades 5).
Diffusional kurtosis is a sensitive and complementary measure of microstructural compromise in patients with temporal lobe epilepsy. It provides additional information regarding the anatomic distribution and degree of damage in this condition. Diffusional kurtosis imaging may be used as a biomarker for disease severity, clinical phenotypes, and treatment monitoring in epilepsy 6).
There were significant differences in kurtosis parameters between high grade glioma and primary central nervous system lymphoma, while differences in diffusion parameters between them did not reach significance; hence, better separation was achieved with these parameters than with conventional diffusion imaging parameters 7).