Predicting Malignant Transformation in Oral Leukoplakia: A Multilayer Perceptron Approach Incorporating...

Area of Science:

• Oncology
• Biotechnology
• Computational Biology

Background:

• Oral leukoplakia (OL) is a potentially malignant oral mucosa disorder.
• Predicting malignant transformation (MT) in OL is a significant clinical challenge.
• Current methods lack sufficient accuracy in risk stratification.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for predicting MT risk in OL.
• Integrate histopathological, demographic, and DNA content features.
• Improve accuracy in identifying high-risk OL cases.

Main Methods:

• Retrospective cohort study of 97 OL cases (18 MT, 79 controls).
• Utilized clinicopathological features and DNA content (cell cycle phases, 4cER) via flow cytometry.
• Trained a multilayer perceptron (MLP) model with 27 features, employing cross-validation and oversampling.

Main Results:

• Significant MT predictors included 4cER, G2 phase, dysplasia grading, and inflammatory infiltrate.
• The optimized ML model achieved 72% sensitivity, 96% specificity, and 85.4% AUC.
• Survival analysis indicated significantly poorer outcomes for high-risk cases identified by the model.

Conclusions:

• Integrating DNA content analysis with ML offers an objective model for stratifying OL malignant risk.
• This approach complements conventional histopathology.
• Supports personalized patient management for potentially malignant disorders.