Technological developments in medical imaging are resulting in a large increase of clinical data available for diagnosis. and prognosis of diseases. Texture is a key component in medical image understanding and analysis. Computer-assisted diagnosis based texture analyses (CAD-TA) have focused on segmentation and classification of visible focal lesions into benign or malignant. However, characterizing abnormalities, predicting disease severity, patient survival and disease prognosis are more complex and demanding problems. In this thesis we propose a CAD-TA approach applicable in these areas of medical imaging. This approach combines two methods: transform based using a Laplacian of Gaussian (LoG) spatial filter and statistical based texture quantification. The transform based selective-scale approach-generates a number of sub-images at different bands of spatial frequencies extracting and enhancing features based on scale and intensity variations (eg. fine-coarse), whilst the relative-scale approach analyses relative contributions made by texture at two different texture scales (e.g. fine/coarse). .Statistical texture parameters quantify brightness, heterogeneity and in~ensity distribution within these derived images. Also, three-dimensional (3-D) CAD-TA of the whole organ provides a novel extension to the two-dimensional (2-D) approach. Relative TA provides imaging biomarkers that reflect hepatic physiology and identifies colorectal cancer patients with reduced· survival from routine computed tomography (CT) images of apparently disease-free areas of the liver. Relative TA I . predicts breast cancer invasion and receptor status from mammographic abnormalities. 3-D selective-scale TA of pulmonaiy CT provides texture correlates for ventilated and vascular lung, thereby demonstrates this approach to be useful in the assessment and distinction of pulmonary disorders. CAD-TA extracts subtle, but crucial information not easily perceptible to the naked eye. Relative TA is demonstrated to be least sensitive to image acquisition parameters and provides an intuitive rationale about the underlying biology that alters image texture. CAD-TA may assist in better patient management and optimal use of surgery/treatment.