Objective: Nuclear pleomorphism (nuclear membrane irregularity) was investigated using transmission electron microscope (TEM) micrographs of 1,419 nuclei (32 oral carcinomas and normal cells from surgical margins).
Study design: Nuclear profiles (1,400 x) were digitized (1 pixel = 35 nm) and fractal dimension estimated using the "yardstick" method.
Results: Log-log plots of yardstick length vs. perimeter showed a significant effect on length measurement typical of fractals at low resolutions (large yardsticks), but this effect disappeared at higher resolution (small yardsticks); that is compatible with Rigaut's asymptotic fractal model. Analysis of the asymptotic fractal parameters c, L and Bm showed that c was higher in normal nuclei, but log(L) and Bm were higher in malignant nuclei. A linear discriminant analysis using c, log(L) and Bm reclassified correctly 78.8% of the nuclei (normal 88.0%, tumor 70.2%).
Conclusion: Asymptotic fractal analysis of nuclear profiles appears to show great potential for quantitative discrimination of oral cancer cell features.