The revolution in health data began around 2009 with the advent of digital health. This shift to digital platforms enabled significant advancements in data mining, which proved immensely valuable for clinicians and managers, especially during the COVID-19 pandemic. In 2021, we focused on the application of Artificial Intelligence (AI) in our field, shared Dr Mohammed Aljohani, Director, Ministry of Health, Saudi Arabia at 2nd Elets Global Healthcare Summit & Awards, Dubai.

He stated, “One of the common diseases we encounter in dental care is periodontitis, an inflammation leading to bone loss. We aimed to develop a code for a detection model that could distinguish between normal and abnormal states. We used images to train the machine, explaining what constituted normal characteristics like the pointed crystalline alveolar bone.”

In 1999, a classification for periodontitis was established with categories like mild, moderate, and severe. However, there was limited research on predicting treatment outcomes. Our goal was to develop a code that could not only detect periodontitis but also classify its severity. Currently, our model has reached a 95 per cent accuracy rate, but we aim for 99 per cent.

For this project, we obtained IRB approval and collected 1,700 images, which we used to train our AI model. One challenge we faced was the imbalance in data, which could lead to overfitting problems in the model.

He further added, “Our AI model utilises the Visual Geometry Group’s 16-layer framework, which we adapted for our purposes. We particularly focused on preventing overfitting by dropping 50 per cent of neurons during training. The optimizers we chose, RMSprop and Adam, were selected for their efficiency in handling large image datasets.”

Our preliminary results showed a significant distinction between normal and diseased states. However, there’s still a gap in the classification accuracy of different disease stages, which we aim to address. Interestingly, our studies indicate that the system’s accuracy varies between anterior and posterior teeth, a hypothesis that requires further validation.

Concluding the session, he stated, “While we’ve made significant strides in applying AI to dental radiology, ongoing research and development are needed to refine our models, especially in terms of data balancing and expanding to 3D imaging technologies.”

Like

Dislike

Be a part of Elets Collaborative Initiatives. Join Us for Upcoming Events and explore business opportunities. Like us on Facebook , connect with us on LinkedIn and follow us on Twitter , Instagram.