Fractal Dimension Analysis to Facilitate Histopathology Diagnosis in Oral Carcinogenesis

Background: Oropharyngeal and Oral squamous cell carcinoma (OSCC) are diagnosed in approximately 270,000 patients worldwide. Diagnosis usually depends on assessment of tumor invasion patterns thus leading to differences in opinion between pathologists and reducing treatment effectiveness and ultimately survival. Objective: We test the hypothesis that mathematical pattern recognition characterized by fractal dimension (FD) pattern analysis can improve histopathology diagnosis accuracy for OSCC. To accomplish this goal we combined digitalized, public domain software with recognized pathology criteria. We obtained an analysis of patterns of invasion (POI), inflammatory infiltration (II) and a selection of specific cells forming a FD pattern using immunohistochemical (IHC) noted protein expressions. Experimental method: To accomplish FD analysis we used an experimental oral carcinogenesis Golden Syrian hamster (25 animals per groups: 4-5 wks of age) tumor model was induced with dibenz [a, l] pyrene (DBP) (0.0025 mM dissolved in acetone (1.0%, v/v); 2.5 nM/application to the floor of the mouth and lateral border of the tongue) and an application control (1.0% acetone). Accuracy was assessed by comparison of FD+/-POI;+/-II;+/-Lacunarity;+/-IHC: Test set 1 and Test set 2 which were obtained from 125 photographic images (20X magnification) a Test set 3 for IHC analysis used 70 images. Test set 3 tested enhancement of FD after tagged basal/supra basal keratinocyte populations using markers (proliferating cell nuclear antigen (PCNA); CD44; p38 mitogen activated phosphokinase (MAPK), and 8-Hydroxy-2′-deoxyguanosine) in contiguous and non-contiguous tissue sections were identified. Analysis was performed using Java-based image processing tool, ImageJ, which provided the Fractal dimensions (FDs) for each studied element. Statistical analysis such as spearmen correlation and Kappa value analysis were also performed. Results: FD analysis enhanced the accuracy with insignificant variability of diagnosis between normal, premalignant and malignant tissues using two independent examinations or repeated examination by an identical investigator. Furthermore inclusion of IHC enhanced FD analysis when markers identified cells with a significant contribution to the FD pattern using contiguous tissue sections. Conclusion: FD analysis should be considered for improved diagnosis and understanding of experimental oral carcinogenesis tissues and this study provides a template to study human tissues to determine improved accuracy.