An Approach to Imbalanced Cavity Detection With Adaptive Sampling and Ensemble Learning: A Hybrid of Deep Learning and Machine Learning

Detection of dental cavities is one of the most important ways of early diagnosis and prevention of caries. The huge disparity between the cavity and non-cavity samples in dental datasets inflicts more problems in achieving an accurate classification. In this work, we propose a new strategy that merges adaptive sampling with K-Means clustering, the generation of synthetic data through SMOTE and ADASYN, and ensemble learning with deep and traditional machine learning classifiers to enhance classification performance. First, K-Means clustering is applied to the dataset to perform adaptive sampling, ensuring improved misrepresentation of minority classes. Then, with the help of SMOTE and ADASYN, synthetic samples are generated to further balance the data. Finally, two pre-trained models such as InceptionNetV3, ResNet50, ResNet18, DenseNet121, Efficient-NetB0, DenseNet169, and VIT B16 etc. are parallelly used for feature extraction, and traditional machine learning classifiers such as Random Forest, Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost) used for classification. Advanced Adaptive Sampling via Clustering with SMOTE (AASC with SMOTE) performed best compared to other sampling techniques with an accuracy of 87%. The InceptionNetV3 and ResNet18 models, combined with Random Forest, achieved the highest performance among all combinations.