Visual Diagnostics of Dental Caries through Deep Learning of Non-Standardised Photographs Using a Hybrid YOLO Ensemble and Transfer Learning Model

Abstract

Background: Access to oral healthcare is not uniform globally, particularly in rural areas with limited resources, which limits the potential of automated diagnostics and advanced tele-dentistry applications. The use of digital caries detection and progression monitoring through photographic communication, is influenced by multiple variables that are difficult to standardize in such settings. The objective of this study was to develop a novel and cost-effective virtual computer vision AI system to predict dental cavitations from non-standardised photographs with reasonable clinical accuracy. Methods: A set of 1703 augmented images was obtained from 233 de-identified teeth specimens. Images were acquired using a consumer smartphone, without any standardised apparatus applied. The study utilised state-of-the-art ensemble modeling, test-time augmentation, and transfer learning processes. The “you only look once” algorithm (YOLO) derivatives, v5s, v5m, v5l, and v5x, were independently evaluated, and an ensemble of the best results was augmented, and transfer learned with ResNet50, ResNet101, VGG16, AlexNet, and DenseNet. The outcomes were evaluated using precision, recall, and mean average precision (mAP). Results: The YOLO model ensemble achieved a mean average precision (mAP) of 0.732, an accuracy of 0.789, and a recall of 0.701. When transferred to VGG16, the final model demonstrated a diagnostic accuracy of 86.96%, precision of 0.89, and recall of 0.88. This surpassed all other base methods of object detection from free-hand non-standardised smartphone photographs. Conclusion: A virtual computer vision AI system, blending a model ensemble, test-time augmentation, and transferred deep learning processes, was developed to predict dental cavitations from non-standardised photographs with reasonable clinical accuracy. This model can improve access to oral healthcare in rural areas with limited resources, and has the potential to aid in automated diagnostics and advanced tele-dentistry applications.

Informatic tools for diagnosis in dentistry. A compilation review

“…Computer tools for diagnosing tooth decay from images offer objective verification, aid in doctor-patient communication, teledentistry, and potentially improve diagnostic accuracy and efficiency in the detection of oral diseases. (61) Studies such as Tareq et al (62) argue that these applications make it possible to predict dental cavitations from non-standardized photographs with reasonable clinical accuracy, improving access to oralhealthcaree in resource-limited areas. In addition, the use of deep learning in panoramic images makes it possible to accurately detect various tooth-related diseases in real-time, helping to plan treatment in time and reducing the risk of misdiagnosis.…”

A 3D printed electronic wearable device to generate vertical, horizontal and phono-articulatory jaw movement parameters: A concept implementation

“…To enhance the current design of the wearable device, several modifications can be made. These include creating a thinner chassis, increasing modularity to accommodate varying head sizes, and installing fiducial markers to enable easy tracking through various object detection algorithms for real-time image tracing [ 5 , 29 ]. The use of image tracing has enabled accurate dental diagnoses even with low-resolution, noise-prone imaging [ 30 ], which suggests that low-cost cameras may be a viable option for this purpose.…”

“…The methods of diagnostic analysis vary in their approaches and technologies, and each possesses unique advantages and limitations depending on the patient’s requirements and the research or clinical application. However, the high cost of specialized equipment and maintenance can make it difficult for remote rural practices and developing economies to access them [ 5 ]. Additionally, most primary care practices may not require such complex and expensive equipment and would only need basic evaluation tools to aid in estimating whether any side of the jaws are affected, scheduling appropriate referrals, and reducing waiting periods.…”

Clinical Annotation and Segmentation Tool (CAST) Implementation for Dental Diagnostics

“…However, this strategy necessitates the use of robust object detection methods, a form of deep learning capable of learning from images, to initially train CAST. You Only Look Once (YOLO) has been a robust object detection method extensively used in previous research for the same task [ 7 , 8 ]. YOLOv8, the most recent iteration of YOLO introduced in 2023, surpassed its predecessors, including YOLOv3, scaled YOLOv5, YOLOv6, and YOLOv7, in terms of precision and recall [ 9 ].…”