Detection of dental restorations on panoramic radiographs using a no-code computer vision platform

Objective

To use a no-code computer vision platform (LandingLens) to develop, train, and evaluate an artificial intelligence model specifically designed for the detection of dental restorations on panoramic radiographs. No-code computer vision platforms, driven by deep learning neural networks, offer a versatile solution that effectively addresses challenges associated with the need for extensive machine learning expertise, expensive training costs, and operational proficiency.

Study Design

Institutional review board approval was obtained for this study. A convenient sampling method was employed to select one hundred panoramic radiographs from the AxiUm records of the dental school. Exclusion criteria were applied to ensure the selection of diagnostic radiographs. Accurate labeling of dental restorations was performed by calibrated dental faculty and students, with subsequent final review by a radiologist.

The radiographs were randomly split into training (70%), development (20%), and testing (10%) subgroups. The model was trained for 40 epochs using a medium model size. Data augmentation techniques such as horizontal flip and vertical flip were employed to enhance the training process.

Results

At a confidence threshold of 0.95, the model achieved a sensitivity of 86.64%, specificity of 99.78%, accuracy of 99.63%, and precision of 82.4%. These metrics indicate the model's ability to accurately detect dental restorations on a limited set of panoramic radiographs.

Conclusion

This study highlights the potential of no-code computer vision platforms in radiology. However, further research and validation are required to evaluate performance on larger and more diverse datasets, as well as for other detection tasks. Continued exploration of these platforms can contribute to advancements in dental imaging by democratizing computer vision development.