Table 1.

Advanced AI applications in endodontics.

Authors	Summarized Abstract	Methods Used	Results	Conclusions
Hu Z. [7]	-Deep learning models diagnose vertical root fractures (VRF) in vivo on CBCT images.	-Deep learning networks used: ResNet50, VGG19, and DenseNet169, with a training to testing ratio of 3:1.	-ResNet50 showed higher diagnostic efficiency than VGG19, DenseNet169, and radiologist.	-Deep learning models are promising for screening VRF teeth.
Fukuda M. [9]	-CNN system detects vertical root fractures on panoramic radiography images.	-A CNN-based deep learning model developed using DetectNet within DIGITS version 5.0 software.	-In total, 267 out of 330 VRFs were detected in the study. -Recall was 0.75, precision 0.93, and F measure 0.83.	-CNN model shows promise in detecting vertical root fractures on radiography.
Johari M. [6]	-CNN model detects VRF in teeth using radiographs.	-Probabilistic neural network (PNN) trained to diagnose and classify teeth with and without VRFs.	-CBCT images showed higher accuracy, sensitivity, and specificity than periapical radiographs.	-Neural network effective in diagnosing VRFs on CBCT images. -CBCT images more effective than periapical radiographs for VRF diagnosis.
Hiraiwa T. [8]	-Deep learning system accurately classifies root morphology on panoramic radiographs.	-Data augmentation process enhanced training image patches for deep learning. AlexNet and GoogleNet were utilized.	-In total, 21.4% of distal roots had extra roots on CBCT images. -Deep learning system had 86.9% accuracy in root morphology classification.	-Deep learning system accurately diagnose single or extra roots. -High accuracy in differential diagnosis of distal roots in molars.
Altındag A. [5]	-Study on deep learning model for pulp stone detection.	-Pulp stones were marked using the CranioCatch (CranioCatch, Eskişehir, Turkey) labeling program. -Mask R-CNN architecture was utilized for the deep learning model.	-Deep learning model achieved 90% sensitivity in detecting pulp stones.	-Deep learning detects pulp stones, aiding clinicians in diagnosis. -Larger datasets enhance accuracy of deep learning systems.
Pauwels R. [3]	-A comparison study between convolutional neural networks and human observers for detection of periapical lesions on intraoral radiographs.	-CNN performance on validation data compared with three oral radiologists in terms of sensitivity, specificity, and ROC-AUC.	-Mean sensitivity, specificity, and ROC-AUC values for CNN were 0.79, 0.88, and 0.86. -Radiologists had values of 0.58, 0.83, and 0.75, respectively.	-CNNs show promise in periapical lesion detection with high accuracy.
Kirnbauer B. [4]	-Deep CNN for automated detection of osteolytic periapical lesions in CBCT data.	-Two-step approach for automatic detection of periapical lesions: Spatial Configuration-Net for tooth localization, modified U-Net architecture for segmentation of lesions.	-Tooth localization network success rate: 72.6% to 97.3%. -Lesion detection sensitivity: 97.1%, specificity: 88.0%.	-Automated method shows excellent results in detecting osteolytic periapical lesions.
Gao X. [10]	-Study evaluates neural network for predicting postoperative pain after root canal.	-Back propagation (BP) neural network model developed using MATLAB 7.0.	-BP neural network model accuracy was 95.60% for pain prediction. -ANN used to predict postoperative pain with high accuracy.	-ANN model can be used to predict postoperative pain effectively.