Oral mucosal disease recognition based on dynamic self‐attention and feature discriminant loss

In a breakthrough, researchers have unveiled a dynamic self-attention and feature discrimination loss function (DSDF) model to enhance the identification of oral mucosal diseases. The DSDF model addresses data imbalances, complex image backgrounds, and variations in visual characteristics among lesion types. Leveraging dynamic self-attention networks improves context comprehension between image regions, while the feature discrimination loss function enhances feature distinctiveness in similar areas. Remarkably, the DSDF model achieves an impressive 91.16% recognition accuracy, outperforming other advanced methods by 6%. With a recall of 90.87% and an F1 score of 90.60%, DSDF shows promise for aiding in oral mucosal disease diagnosis.