Quantitative Structural Analysis of Hyperchromatic Crowded Cell Groups in Cervical Cytology: Overcoming Diagnostic Pitfalls

The diagnostic challenges presented by hyperchromatic crowded cell groups (HCGs) in cervical cytology often result in either overdiagnosis or underdiagnosis due to their densely packed, three-dimensional structures. The objective of this study is to characterize the structural differences among HSIL-HCGs, AGC-HCGs, and NILM-HCGs using quantitative texture analysis metrics, with the aim of facilitating the differentiation of benign from malignant cases. A total of 585 HCGs images were analyzed, with assessments conducted on 8-bit gray-scale value, thickness, skewness, and kurtosis across various groups. HSIL-HCGs are distinctly classified based on 8-bit gray-scale value. Significant statistical differences were observed in all groups, with HSIL-HCGs exhibiting higher cellular density and cluster thickness compared to NILM and AGC groups. In the AGC group, HCGs shows statistically significant differences in 8-bit gray-scale value compared to NILM-HCGs, but the classification performance by 8-bit gray-scale value is not high because the cell density and thickness are almost similar. These variations reflect the characteristic cellular structures unique to each group and substantiate the potential of 8-bit gray-scale value as an objective diagnostic indicator, especially for HSIL-HCGs. Our findings indicate that the integration of gray-scale-based texture analysis has the potential to improve diagnostic accuracy in cervical cytology and break through current diagnostic limitations in the identification of high-risk lesions.