[en] In this thesis a computer-aided sketch (CAS) system and a similarity-based retrieval method in medical images are proposed. A CAS system was developed for assisting physicians to draw a sketch when they were reading a mammogram. The sketch is automatically drawn using elements in the mammogram that were extracted by automated detection techniques developed in our computer-aided diagnosis (CAD) research. In the thesis a technique is described for the segmentation of pectoral-muscle regions for the CAD and CAS systems. The pectoral-muscle region was determined from an edge-enhanced image obtained with a Prewitt filter by searching for a boundary between the pectoral-muscle region and mammary region. The boundaries were extracted correctly in 502 out of 513 images. In the thesis the CAS system is described. Its performance was evaluated by comparing its sketches of mammograms to those made by a physician. 430 mammograms were used in this study. Some differences in the sketches made by the physician and CAS system from the same mammogram are illustrated. If the contours of masses were clear, the sketches tended to match well to the physician's. A method of similarity-based image retrieval and demonstrations of its potential usefulness for medical images is described. The purpose was to develop an image retrieval system that searches the image database for an image resembling the one entered without carrying out image segmentation. The approach presented includes a local-pattern matching method and principal component analysis on monochromatic images. The image retrieval method was able to find images with similar characteristics from 30 monochromatic images. Initial results on mammographic masses indicate that the method might provide a new CAD system for mammograms