The purpose of this research is to provide a method to automatically classify fingerprints into three groups: whorl, loop and arch, which can help medical scientists to study the relationship between fingerprint patterns and medical disorders, such as women breast cancer . In the research, a frequency domain approach was developed, which uses the feature of Fourier spectrum. That is, prominent peaks in the spectrum give the principal direction of fingerprint patterns. Using the above feature, we can obtain the principal direction of every subregion. The pattern of the whole image then can be determined. The frequency domain approach allows one to classify whorl faster and is less sensitive to the quality of fingerprint image, but it does not easily allow for the classification of arch and loop when triradii areas are too small.