Screening of Oral Squamous Cell Carcinoma Through Color Intensity-Based Textural Features

Abstract

Background Early screening and diagnosis of oral squamous cell carcinoma (OSCC) has always been a major challenge for pathologists. Artificial intelligence (AI)-assisted screening tools can serve as an adjunct for the objective interpretation of Papanicolaou (PAP)-stained oral smears. Aim This study aimed to develop a handy and sensitive computer-assisted AI tool based on color-intensity textural features to be applied to cytologic images for screening and diagnosis of OSCC. Methodology The study included two groups consisting of 80 OSCC subjects and 80 control groups. PAP-stained smears were collected from both groups. The smears were analyzed in Matlab software computed data and color intensity-based textural features such as entropy, contrast, energy, homogeneity, and correlation, were quantitatively extracted. Results In this study, a statistically significant difference was noted for entropy, energy, correlation, contrast, and homogeneity. It was found that entropy and contrast were found to be higher with a decrease in homogeneity, correlation, and energy in OSCC when compared to the control group. Receiver operating characteristic curve analysis was done and accuracy, sensitivity, and specificity were found to be 88%, 91%, and 81%, respectively. Conclusion The gray-level co-occurrence matrix (GLCM) color intensity-based textural features play a significant role in differentiating dysplastic and normal cells in the diagnosis of OSCC. Computer-aided textural analysis has the potential to aid in the early detection of oral cancer, which can lead to improved clinical outcomes.