Artificial intelligence for detection of periapical lesions on intraoral radiographs: Comparison between convolutional neural networks and human observers

Abstract

Objective. The aim of this study was to compare the diagnostic performance of convolutional neural networks (CNNs) with the performance of human observers for the detection of simulated periapical lesions on periapical radiographs. Study Design. Ten sockets were prepared in bovine ribs. Periapical defects of 3 sizes were sequentially created. Periapical radiographs were acquired of each socket with no lesion and with each lesion size with a photostimulable storage phosphor system. Radiographs were evaluated with no filter and with 6 image filter settings. A CNN architecture was set up using Keras-TensorFlow. Separate CNNs were evaluated for randomly sampled training/validation data and for data split up by socket (5-fold cross-validation) and filter (7-fold cross-validation). CNN performance on validation data was compared with that of 3 oral radiologists for sensitivity, specificity, and area under the receiver operating characteristic curve (ROC-AUC). Results. Using random sampling, the CNN showed perfect accuracy for the validation data. When data were split up by socket, the mean sensitivity, specificity, and ROC-AUC values were 0.79, 0.88, and 0.86, respectively; when split up by filter, they were 0.87, 0.98, and 0.93, respectively. For radiologists, the values were 0.58, 0.83, and 0.75, respectively. Conclusions. CNNs show promise in periapical lesion detection. The pretrained CNN model yielded in this study can be used for further training on larger samples and/or clinical radiographs. (Oral Surg Oral Med Oral Pathol Oral Radiol 2021;000:1À7) Statement of Clinical RelevanceConvolutional neural networks are used to develop artificial intelligence systems for diagnostic tasks in oral and maxillofacial radiology. This study indicates that convolutional neural networks can yield diagnostic performance comparable to or better than that of human observers for detection of periapical lesions.

Analysis of Advances in Research Trends in Robotic and Digital Dentistry

“…For treatment and rehabilitation, robotics is used and for the prognosis, imaging and the diagnosis, virtual AI is used. [ 8 9 10 ] Our findings suggest a lower awareness and application of these technologies among clinicians. We had difficulty comparing our results as this is one of the first studies evaluating the awareness of advanced technologies in dentistry.…”

Where Is the Artificial Intelligence Applied in Dentistry? Systematic Review and Literature Analysis

“…These models can be of great assistance for less experienced dentists and non-specialists, as they can be used in clinical applications [ 118 , 125 ]. Artificial intelligence technology is widely used in the diagnosis (radiographic detection, CBCT images) of periapical pathologies, demonstrating satisfactory results with high sensitivity and moderate specificity [ 127 , 128 , 129 ]. It is especially important to detect vertical root fractures (VRFs) at an early stage to prevent damage to supporting structures.…”

Artificial Intelligence in Dentistry—Narrative Review

“…In restorative dentistry, neural networks can detect tooth decay or restorations, moreover, they can facilitate the choice of caries excavation method [ 3 , 22 , 23 ]. In endodontics, neural networks can be useful in detecting periapical lesions and root fractures, root canal system anatomy evaluation, predicting the viability of dental pulp stem cells, determining working length measurements, and predicting the success of retreatment procedures [ 25 , 26 , 27 , 28 , 29 ]. In orthodontics, they can facilitate the diagnosis and treatment planning, cephalometric points marking, anatomic analyses, assessment of growth and development, and the evaluation of treatment outcomes [ 33 , 34 , 36 , 39 , 40 ].…”

“…The periapical radiographs were evaluated to find periapical lesions made in bovine ribs. The results were compared with three oral radiologists and the CNN showed a perfect accuracy of 87% [ 26 , 29 ]. Ekert et al assessed panoramic images for the presence of periapical lesions with the help of CNN.…”