Protocol for an interpretable self-attention artificial neural network framework to predict cancer risk in oral potentially malignant disorders

Oral potentially malignant disorders (OPMDs) can directly progress to cancer, necessitating accurate risk prediction to guide clinical intervention. Here, we present a protocol for predicting cancer risk in OPMDs using a customized self-attention artificial neural network (SANN). We describe steps for full preprocessing, defining and training the model, and model validation. We then detail procedures for evaluation and visualization. The model is trained and validated on 1,094 cases and compared with artificial neural networks, random forests, and DeepSurv. For complete details on the use and execution of this protocol, please refer to Li et al. 1 • Instructions for constructing the SANN model and setting hyperparameters • Steps for handling class imbalance using SMOTE and normalizing data • Guidance on model training and 10-fold cross-validation execution • Steps for evaluating performance metrics and saving prediction results Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics. Oral potentially malignant disorders (OPMDs) can directly progress to cancer, necessitating accurate risk prediction to guide clinical intervention. Here, we present a protocol for predicting cancer risk in OPMDs using a customized self-attention artificial neural network (SANN). We describe steps for full preprocessing, defining and training the model, and model validation. We then detail procedures for evaluation and visualization. The model is trained and validated on 1,094 cases and compared with artificial neural networks, random forests, and DeepSurv.