Automated analysis of CT images using fractal theory

Rajnish K. Jauhari¹, P. Munshi², and K.K. Pandey³

rajnish.jauhari@gmail.com, pmunshi@iitk.ac.in

¹J. K. Cancer Institute, G.S.V.M. Medical College, Kanpur, India

²Indian Institute of Technology, Kanpur, India

³Kanpur Scans Pvt. Ltd., Kanpur, India

The diagnostic interpretation of medical images is a multifaceted task. Its objective is the accurate detection and precise characterization of potential abnormalities-a crucial step toward the institution of effective treatment. Achieving this goal relies on the radiologists’ successful interpretation of two distinct processes: (a) the process of image perception to recognize unique image patterns and (b) the process of reasoning to identify relationship between perceived patterns and possible diagnosis. Both the processes depend heavily on the radiologists’ empirical knowledge, memory, institution and diligence. There is evidence that the human visual system has difficulty in the discrimination of textural information that is related to higher-order statistics or spectral properties on an image.

Automated analysis of medical images is the need of hour for accurate and error free diagnosis and treatment. The present study is an attempt to use the fractal theory, developed by Mandelbrot, for automated analysis of CT images of brain. Two sets of CT images, normal and diseased, are included in this study. Computation of fractal dimensions of each image is done by computer code developed in C using V20Z SUN machines. The results show that the difference in the fractal dimensions, with proper choice of scales, can be used as a parameter for differentiation between normal and diseased brain CT images. The fractal dimensions of normal images (smooth) found to be higher than the diseased images (rough). Higher fractal dimension of normal images also indicates that these images are relatively ‘smooth’ than the diseased brain images, which is also intuitively correct in the clinical situation. The approach discussed here will be a supplementary tool in the hands of radiologists for better diagnosis even in the difficult situations where abnormality may not be visualized by the human eye.