Predicting Progression of Oral Lesions to Malignancy Using Machine Learning

The study aimed to develop machine learning models to predict malignant transformation of oral lesions using electronic health record (EHR) data. The retrospective cohort consisted of 2192 patients with a biopsied oral lesion, and machine learning models were trained on the data in two experiments. The best model predicted malignant transformation among biopsied oral lesions with an area under the curve (AUC) of 86%, while the random forest model predicted malignant transformation among lesions with dysplasia with an AUC of 0.75. Dysplasia grade and the presence of multiple lesions were the most influential features in predicting malignant transformation. The study concluded that machine learning approaches are feasible and effective for generating models that predict which oral lesions are likely to progress to malignancy.