MR-imaging: a new approach for glioma characterization

Gliomas are the most common primary brain tumors. The diffuse infiltration of white matter tracts by cerebral gliomas is a major cause of their appalling prognosis: tumor cells invade, displace, and possibly destroy WM. An early diagnosis and a comprehensive evaluation of tumor extent and relationships with surrounding anatomical structures are crucial in determining prognosis and treatment planning. Conventional Magnetic Resonance (MR) sequences (e.g. T1- or T2—weighted images) have limited sensitivity and specificity in diagnosing brain tumors,[1] because they do not always allow precise delineation of tumor margins, or tumor differentiation from edema and /or treatment effects. In particular, contrast-enhanced MR images may underestimate lesion margins, which is critical for image-guided tumor resection, radiotherapy planning, and for assessing the response to chemotherapy. On the contrary, Diffusion Tensor Imaging (DTI) can identify peritumoral white—matter abnormalities, by detecting the presence of small areas with tumor—cell infiltration in WM around the edge of the gross tumor, as confirmed by image guided biopsies. In particular the tumor core is characterized by reduced anisotropy and increased isotropy, while, around this area, tumor infiltration shows increased isotropy, but normal anisotropy. The aim of this study was to characterize pathological and healthy tissue in DTI datasets by 3D statistical analysis. In order to investigate the pathological tissues, greyscale digital FLAIR images have been processed. Hence, several well—known statistical quantities have been used to gather meaningful information from the available dataset. The most commonly used indexes of location are mean, mode, median and quartiles. The dispersion (or variability) is given by the variance s2, which is related with its second order moment of the distribution, and its square root, the standard deviation s; dividing the latter by the absolute value of the mean one obtains the coefficient of variation CV, i.e. a non-dimensional measure of spread. Another feature of interest is the heterogeneity, usually characterized by the Gini concentration index and entropy, scaling range from 0 (minimum concentration) up to 1 (maximum concentration). Skewness and kurtosis represent the 3rd and 4th order moments of the distribution, and locate the asymmetry and the “distance” from a perfectly normally distributed variable. Finally, an estimation of the fractal dimension is performed using by box counting. Box counting is a method of gathering data for analyzing complex patterns by breaking a dataset, object, image, etc. into smaller and smaller pieces, typically "box"—shaped, and analyzing the pieces at each smaller scale[2]. This arsenal of instruments allowed us to determine the statistical differences among different gliomas.