Dental image enhancement network may facilitate early diagnosis of oral and dental diseases

Recent research proposed a new dental image enhancement framework for the early detection and diagnosis of oral dental diseases. Researchers proposed a practical and clinically oriented adaptive enhancement strategy to act adaptively in practical scenarios for the early diagnosis of dental diseases embedded in a denticle edification network called the Ded-Net. The proposal was published in the journal Scientific Reports.

Due to the uneven brightness and poor contrast in the recorded pictures, intelligent robotics and expert system applications in dentistry have been suffering from identification and detection issues. Certain drawbacks like sensitive facial parts' exposure to ionizing radiations during the diagnostic process confine the view of vision. Modern digital technologies have difficulty capturing high-quality medical photos, and processing these images degrades contrast and visual clarity. It limits the capabilities of future intelligent and expert systems and discourages the early detection of dental and oral disorders. While network-based approaches rely enormously on large-scale datasets with limited flexibility to varying situations, traditional enhancement methods are designed for particular settings.

Hence researchers proposed a novel and adaptive dental image enhancement strategy based on a small dataset and a paired branch Denticle-Edification network (Ded-Net). The input dental images are reflected and illuminated using a multilayer Denticle network called the De-Net. Any unseen degradation of reflection or illumination is removed by subsequent enhancement operations. An Edification network called the Ed-net is used to maintain adaptive illumination consistency.

The network is regularized after determining that the input data's decomposition is congruent, and it offers the user-specific adaptation flexibility they seek for desired contrast levels. Luminance in the final image is balanced by the practical trade-off. The advantage of this proposed method is that it predicts dynamically in resilient conditions without any requirement for large-scale / paired or carefully selected datasets.

Thus, the experimental results of this proposed study improved visibility and contrast and preserved the edges and boundaries of the low-contrast input images. They also suggested that future dental imaging can be modified by these intelligent and expert system applications.

Further reading: Khan, R., Akbar, S., Khan, A. et al. Dental image enhancement network for early diagnosis of oral dental disease. Sci Rep 13, 5312 (2023).https://doi.org/10.1038/s41598-023-30548-5