Tehran University Medical Journal

Mendeley  

Zotero  

Saligheh Rad H, Fathi Kazerooni A, Nabil M, Alviri M, Hadavand M, Mohseni M. Differentiation of active tumor from edematous regions of glioblastoma multiform tumor in diffusion MR images using heterogeneity analysis method. Tehran Univ Med J 2018; 76 (3) :178-184
URL: http://tumj.tums.ac.ir/article-1-8831-en.html

Background: Due to intrinsic heterogeneity of cellular distribution and density within diffusion weighted images (DWI) of glioblastoma multiform (GBM) tumors, differentiation of active tumor and peri-tumoral edema regions within these tumors is challenging. The aim of this paper was to take advantage of the differences among heterogeneity of active tumor and edematous regions within the glioblastoma multiform tumors in order to discriminate these regions from each other.
Methods: The dataset of this retrospective study was selected from a database which was collected at the medical imaging center, Imam Khomeini Hospital, Tehran University of Medical Sciences, Iran. The quantification was performed as a part of a research study being supported by the Research Center for Molecular and Cellular Imaging, Tehran University of Medical Sciences, Iran, between May and September 2017. Twenty patients with histopathologically-confirmed GBM tumors who had been imaged on a 3T MRI scanner prior to their surgery, were included. Conventional and diffusion weighted MR images had been carried out on the patients. The regions of interest including the regions of active tumor and edema were identified on MR images by an expert and overlaid on ADC-maps of the same patients. Histogram analysis was performed on each of these regions and 14 characteristic features were calculated and the best feature combination for discrimination of active tumor from edema was obtained.
Results: It was shown that by combining 8 out of 14 histogram features, including median, normalized mean, standard deviation, skewness, energy, 25th, 75th, and 95th percentiles, differentiation with accuracy of 96.4% and diagnostic performance of 100% can be achieved. Furthermore, by combining mean, energy, and 75th percentile features of histograms, the active tumor region can be discriminated from the edematous region by 92.7% of accuracy and 98.9% of diagnostic performance.
Conclusion: The present study confirms that the heterogeneity of cellular distribution can be a predictive biomarker for differentiation of edematous regions from active tumor part of GBM tumors.