Feasibility of computer evaluation of the histopathologic findings in human skeletal muscle disease.

Optical Fourier methods of image processing coupled with mutual information analysis were applied to the separation of human muscle specimens into four diagnostic classes. This process allowed rapid, relatively inexpensive data acquisition, reduction, and classification. The accuracy of the method when compared to that of experienced surgical pathologists is acceptable. Difficulties arise when material outside of the previous experience of the computer is presented.