Oral exfoliative DNA aneuploidy cytology by image cytometry (DNA-ICM) has been introduced for the early diagnosis of malignant transformation. This study aimed to investigate the optimal cut-off values of DNA content and assess the diagnostic accuracy of artificial intelligence (AI)-assisted DNA-ICM for detecting the risk of malignant transformation in oral potentially malignant disorders (OPMDs). A total of 6,874 consecutive patients with clinical lesions of OPMDs were subjected to AI-assisted DNA-ICM analysis. In this study, both oral exfoliative cytobrushing and surgical biopsy samples were obtained from 415 patients with OPMDs and then allocated by AI-assisted DNA-ICM and histopathological examination, respectively. The diagnostic efficiency and accuracy of AI-assisted DNA-ICM analysis were evaluated in this OPMDs cohort. Fisher's exact test was applied to evaluate differences in qualitative variables. Logistic regression analysis was used to estimate the odds ratio (OR) and the associations among confounding variables. For the optimal cut-off values of at least 1 aneuploid cell with DNA content ≥4.86c, the area under the curve (AUC) was 0.840 (sensitivity, 76.19%; specificity, 74.40%) for the discrimination of dysplasia from OPMDs. For the DNA content ≥6.2c, the AUC was 0.766 (sensitivity, 80.52%; specificity, 64.38%) for detecting carcinoma from OPMDs. Notably, OR analysis by logistic regression suggested that ages ≥60 years and the areca nut chewing habit were significantly associated with dysplasia or worse in OPMDs. In addition, DNA aneuploidy criteria and their diagnostic accuracy vary among different subtypes of OPMDs. This diagnostic study optimised the criteria of DNA aneuploidy cytology for detecting malignant transformation of OPMDs. AI-assisted DNA-ICM may be an efficient noninvasive diagnostic tool for screening different high-risk OPMDs and early-stage oral cancers. This clinicopathological diagnostic study establishes a methodological foundation for future broader screening applications of oral exfoliative AI-assisted DNA-ICM.
Tan et al. (Thu,) studied this question.