CNN in periodontology: A review

Periodontal Disease (PD) may be characterized by gingival inflammation, loss of connective tissue attachment, or the destruction of alveolar bone. It accounts for the sixth most common inflammatory disease known to mankind¹. The conventional methods in use, today, to diagnose PD include clinical assessment and the analysis of radiographs² by subject experts. Increasingly, there is a need for a more standardized method to diagnose PD, owing to various limitations of the conventional diagnosis paradigm – inaccuracy from human error, contradictory assessments by individual examiners, and the dearth of time to parse through large amounts of data to arrive at precise diagnoses. The advent of Artificial Intelligence, with its nascent diagnostic tools, seeks to make crucial inroads into this gap in various fields of research. Breakthroughs in Machine Learning (ML) and its more robust subset, Convolutional Neural Networks (CNN), over the last decade3 have opened doors to wide-ranging applications – facial and speech recognition, climate mapping, colorization of black and white images, document analysis and preservation, etc. The use of CNN in healthcare has successfully aided medical imaging analysis, cancer screenings, and prediction of neurological conditions. ML may also soon be harnessed to advance dentistry as a discipline. The use of CNN as an unsupervised diagnostic tool in Periodontology, however, has been extremely limited, primarily due to hardware constraints, but also because of the lack of efficient interaction between the exclusive disciplines of AI and dentistry. With the right algorithms and the expertise of examiners trained both in dentistry and neural networks, it is safe to posit that revolutionary leaps can be made in dental diagnostics and prognostics.