Framework for Adaptive Intelligent Multi-classification and Segmentation of Dental Caries Classes and Oral Diseases

Oral disease prediction models based on artificial intelligence have been highly popularized recently within healthcare-related applications. Nevertheless, most available models still suffer from inadequate feature selection, overfitting, high computational complexity, and reduced predictive accuracy in high-dimensional health datasets. To address the aforementioned issues, an innovative Orangutan eXtreme Gradient Prediction (OXGP) scheme is introduced for successful prediction and classification in the field of oral diseases. At the first stage, preprocessing is carried out with the help of Gaussian filter and min-max normalization in order to minimize noise from images. Secondly, exploration and exploitation behaviors, inspired by the orangutans' intelligence while foraging, are used by the OOA algorithm for choosing the most useful feature. Moreover, L2 regularization is employed to avoid overfitting and increase generalizability. The OXGP method was tested on an oral disease dataset comprising several disease types using the same experimental setup. According to experimental results, the OXGP approach achieved high accuracy, precision, recall, F1-score, and AUC, with values of 97.36%, 96.92%, 96.41%, 96.66%, and 98.11%, respectively. Moreover, statistical testing through confidence interval analysis and p-value revealed the significance of the suggested technique. The results clearly show that the OXGP framework significantly enhances feature selection and classification performance, thereby providing a promising technique for intelligent diagnosis of oral diseases. • A new hybrid approach Orangutan Extreme Gradient Prediction (OXGP) has been developed. • Regularization is performed during the hidden phase to avoid overfitting. • Orangutan optimization mechanism facilitates intelligent feature tuning by identifying the most relevant features.