Objectives: This study compared quantitatively the histological architecture of SCC and normal oral mucosa to establish objective morphological criteria to characterise the two entities, in an attempt to automate histological diagnosis.
Methods: The complexity of the epithelial connective tissue interface (ECTI) in 56 profiles from SCC and normal mucosa was analysed at x4 magnification using fractal geometry. The box counting method and the mass radius relation were used to estimate global and local fractal dimensions respectively. High resolution images (x20) of the locally most complex locations in each profile (upper decile) were obtained and both the epithelial and non-epithelial compartments were partitioned into "cells" defined as exclusive areas of the epithelial compartment related to each epithelial nucleus. The morphological and densitometric properties of these cells where then analysed.
Results: The mean box fractal dimension was 1.06 for normal and 1.22 for SCC ECTI profiles. The complexity of ECTI represented by the fractal dimensions alone (global and local at a local window of 47µm and 190µm) allowed 87% correct discrimination between normal and SCC profiles. The discrimination between these two entities using epithelial cellular morphology data alone was 79% but when combined with the fractal analysis, the case-wise correct discrimination rate was 100%.
Conclusion: The architectural and cellular features of normal mucosa and SCC show differences that allow to statistically differentiate the two diagnostic classes. This opens a new possibility to characterise cancerous growth in quantitative terms to be used in an automated or semi-automated diagnosis systems.