Human fingerprints comprise a series of whorls or ridges. In some special cases, these whorls are broken by so-called secondary creases-colinear breaks across a sequence of adjacent ridges. It is a working hypothesis that the presence of these secondary creases form a physical marker for certain human disorders. A technique to automatically detect such creases in fingerprints is described. This technique utilizes a combination of spatial filtering and region growing to identify the morphology of the locally fragmented fingerprint image. Regions are then thinned to form a skeletal model of the ridge structure. Creases are characterized by colinear terminations on ridges and are isolated by analyzing the Hough transform space derived from the ridge end points. Empirical results using both synthetic and real data are presented and discussed.