Computerized image recognition for morphometry of nerve attribute of shape of sampled transverse sections of myelinated fibers which best estimates their average diameter.

A computerized image recognition method was used to measure various attributes of shape of cross-sections of myelinated nerve fibers. Measurements were made at intervals over 1/2 internode of each fiber on 20 fibers from each of 4 sural nerves from rats. Diameters were computed in 6 different ways from the computer measurements and compared for bias, precision, and accuracy between sections and to the diameter of an idealized cylinder reconstructed for each fiber from multiple actual cross sections. The diameter computed from cross-sectional areas of transversely sectioned myelinated fibers, converted into a circle, showed the highest precision, greatest accuracy and least bias. Fibers were classified by shape and the frequency was determined in defined regions (I1 = paranodal, I3= nuclear and I2=region between I1 and I3) of the 1/2 internode. A crenated shape is highly characteristic of the I1 region. The boomerang shape was found most frequently in I3 whereas the circular shape was found most frequently in I2. Epileptical and boomerang shapes of myelinated fibers within fascicles which have been orientated carefully to obtain transverse sections, are not due to obliquity of section. Therefore, using the minor axis to determine the diameter of such profiles, as we had done previously in our laboratory, is in error. We conclude from these studies, that in carefully orientated transverse sections of nerve trucks, the diameter calculated from measurement of area converted to a circular shape is the best among the various estimates of myelinated fiber diameter and is the most suitable one for use in computerized image recognition systems for nerve morphometry. It seems reasonable to extrapolate this general conclusion to myelinated fibers of man.