Can ChatGPT be used in periodontal diagnostics?

30.09.2024

PD Dr Kristina Bertl, PhD MSc MBA

Artificial intelligence has found its way into many areas of our lives. One well-known application of artificial intelligence is ChatGPT, which was unveiled in November 2022. ChatGPT can communicate with users via text messages and/or images, and estimates suggest that it has around 200 million active users per month worldwide.

Artificial intelligence is already being used in dentistry and recently an interesting study in this area, involving ChatGPT, was conducted (Eroglu et al. 2024). The group from Türkiye tested ChatGPT in periodontal diagnostics. There has been a new classification for periodontal and peri-implant diseases and conditions since 2018. As part of this classification, a system involving stages (I to IV) and grades (A to C) was introduced for periodontitis. This required some readjustment for clinicians and help with reaching the correct diagnosis is obviously always welcome!

The study by Eroglu et al. tested ChatGPT’s ability to make the correct periodontal diagnosis. For this purpose, 200 patient cases with untreated periodontitis were comprehensively summarized with all necessary information, and assessed and classified by four experts. These cases were then given to ChatGPT for assessment by asking the question, “What is the stage, grade and extent of periodontitis?”

Bearing in mind that ChatGPT carried out this diagnosis without prior training, the following results were achieved:

The stage, grade and extent of periodontitis were correct in 60, 51 and 84% of the cases. This means that only the classification of the extent of periodontal disease was correct for a high percentage of cases.
Stage I was correctly diagnosed in 91% of cases, while stage IV was only correctly diagnosed in 12% of cases. In general, the discrimination between stage III and IV in particular was often incorrect.
Similar results were achieved for the classification of grade. Grade A was correctly diagnosed in 94% of cases, while grade C was only correctly diagnosed in 29% of cases.

Overall, it can therefore be seen that ChatGPT definitely has potential, but for more complex cases in particular it often gives the wrong diagnosis and therefore cannot yet be used reliably in periodontal diagnostics.

Reference

Zeynep Tastan Eroglu, Osman Babayigit, Dilek Ozkan Sen, Fatma Ucan Yarkac (2024) Performance of ChatGPT in classifying periodontitis according to the 2018 classification of periodontal diseases. Clin Oral Investig, 28(7):407. doi: 10.1007/s00784-024-05799-9.