Evaluation of malignancy in tumors of the central nervous system using fractal dimension

We propose the use of the concepts of fractal dimension and digital image processing, as a possible methodology to characterize the degree of malignancy of neoplastic structures located in the central nervous system. The images were detected by MRI techniques including proton density, T₁, and T₂ images. Malignant lesions (Gliomas) were compared with benign ones (cysts). The correlation dimension, Lyapunov exponents and information dimension were used as the relevant geometrical properties to characterize the irregular edge present in a particular structure. The edge was obtained by means of an edge detector operator and afterwards a codification procedure based on Fourier descriptors was used to generate a numerical array or time series. The analysis of the processed images revealed that the relevant geometrical properties exhibit a different behavior in the case of gliomas compared to cystic lesions, a fact that can be used by the physician as an auxiliary tool to evaluate the malignancy of neoplastic structures in the brain