Conversational AI Models in Periodontitis Classification

NCT ID: NCT05926999

Last Updated: 2023-08-15

Basic Information

• Recruitment Status: COMPLETED
• Total Enrollment: 62 participants
• Study Classification: OBSERVATIONAL
• Study Start Date: 2023-06-25
• Study Completion Date: 2023-07-30

Brief Summary

The goal of this observational study is to test periodontitis classification ability of ChatCPT. The main questions it aims to answer are:

question 1: Could ChatCPT classify periodontitis? question 2: Is there a better result if ChatCPT is trained for perodontitis classification?

Detailed Description

The descriptive pilot research will be based on an analysis of the baseline digital record and subsequent stage, extent, and grade characterization of 75 untreated patients diagnosed with periodontitis. All cases will be evaluated by 4 examiners and the obtained standardized diagnosis for each case. The information will be used to determine the stage, grade, and extent of 75 periodontitis cases will be copied directly for each case as input to a large AI model called ChatGPT (https://chat.openai.com/), followed by query: "What is the stage, grade, and extent of periodontitis?" The received replies will be compared to standardized diagnosis.

2.2 Ethical Considerations From December 2022 to May 2023, baseline clinical and radiographic documentation of periodontitis patients were collected in the context of regular visits at the Necmettin Erbakan University Periodontology Clinic. In the study, anonymized data will be used. All participants have given their agreement in writing for the data to be used for training and research. This study will adhere to the 2013 revision of the 1975 Declaration of Helsinki and will be authorized by the Necmettin Erbakan University Faculty of Dentistry Ethics Committee for Non-Pharmac eutical and Medical Device Clinical Research.

2.3.Selection and preparation of the documentation of the periodontitis cases Using randomization software, a database of 150 patients receiving periodontal therapy will be used for selecting 85 cases of periodontitis from the archive of patients of the periodontology department at Necmettin Erbakan University. Acute periodontal lesions, gingival diseases, the presence of dental implants, and periodontitis as a manifestation of systemic diseases will be considered as exclusion criteria.

The case description included a comprehensive summary of the patient's medical and dental history, intra-oral photographs, a panoramic radiograph, a complete set of periapical radiographs, and periodontal charting that encompassed various clinical measures related to periodontal health. These measures will be plaque scores (visually assessed after the use of a revealing solution, as present or absent), probing depth, bleeding on probing, clinical attachment loss (CAL), furcation involvement (Hamp, Nyman, & Lindhe, 1975) and tooth mobility (Miller, 1985). The medical history will be also supplied, including details regarding pertinent medical issues including glycemic management and cigarette usage. The clinical, photographic, and radiological records for the 85 patients that will be chosen will be of excellent quality and good diagnostic sensitivity.

The whole documentation of the periodontitis instances will be compiled into four presentation files. They will be presented in different orders in all four presentations for four expert evaluations. The first presentation is provided 2.4 Experts' Evaluation Four experienced periodontists who work as full-time faculty members (Z.T.E, O.B, D.O.S, and F.U.Y.) will evaluate the cases in the first phase utilizing the prepared presentations. The four experts had gone over the consensus reports for the 2018 periodontal classification (Papapanou et al., 2018; Tonetti, Greenwell, & Kornman, 2018) multiple times and had been using it to make clinical diagnoses for at least 4 years. These diagnoses will be considered standardized diagnoses and served as the reference for each respective case. The cases that did not achieve a consistent diagnosis among the experts will be excluded from the study.

2.5 Staging, grading, and determining extent of periodontitis cases using ChatCPT

ChatGPT is an implementation of the Generative Pre-Training Transformer 3 (GPT-3) language model developed by OpenAI, which is publicly accessible and freely available for use. (Brown et al., 2020). GPT-3 is a highly expansive neural network-based natural language processing (NLP) model, currently one of the largest in existence. With training on 175 billion parameters, its primary purpose is to generate text that closely resembles human language. Acting as a versatile chatbot, GPT-3 is capable of performing diverse NLP tasks such as language translation, summarization, and question-answering (Balas & Ing, 2023). Among its many possible uses, we will evaluate the performance of GPT-3 in stage, grade, and extent determination of periodontitis using case descriptions. its ability to stage, grade, and determine extent when given case descriptions of periodontitis. Since ChatCPT is a language model and cannot use images, the radiographs of the cases will be evaluated by the four experts. Bone loss amounts and bone loss rates will be measured and turned into numerical data that ChatCPT could use. Standardized texts containing the information that should be used to determine the stage, grade, and extent of each case will be created. This information is as stated below.

For staging;

1. Age and gender;

2. Maximum clinical attachment loss in the interproximal area

3. The percentage of bone loss;

4. Number of tooth loss due to periodontal reasons;

5. Maximum probing depth;

6. The bone loss type;

7. Furcation involvement (FI) according to the Hamp classification (Hamp et al., 1975)

8. the presence of chewing dysfunction;

9. 2nd-degree and above-tooth mobility;

10. The presence of ridge defect;

11. Number of teeth in occlusion; For determining the extent; Periodontitis coverage; For Grading;

a. The amount of bone loss in the last 5 years; b. The age ratio of the percentage of bone loss in the worst area; c. Phenotype of destruction: d. Smokin status and number of cigarettes smoked per day; e. Diabetic status and HbA1c level below or above. For this study, a new account will be created, granting access to ChatGPT through the link provided (https://chat.openai.com/chat). The standardized texts for each case will be written in English, and then the question "What stage, grade, and extent is the periodontitis?" will be asked to ChatGPT. The same current version of the ChatGPT program will be used in the query, and the query process will be done in two ways.

1. To minimize the impact of prior responses, a new chat window will be opened for each question asked, and the responses will be recorded for later analysis.

2. A new chat window will be opened and the basic information needed to determine the stage, grade, and extent of periodontitis according to the 2018 classification will be transmitted to ChatCPT. ChatCPT will be then asked to classify the cases to be forwarded later according to this information, and this request will be approved by ChatCPT. The same standardized texts for each case used in the first query will be transmitted to the ChatCPT, this time using the same chat window, and again "What stage, grade, and extent is the periodontitis?" question will be asked. The responses will be recorded for later analysis.

In 2 different inquiries from chatCPT, an answer will be obtained for the stage, grade and extent of periodontitis for each case. The stage, grade and extent responses obtained for each case will be compared with the standardized diagnosis created by the experts.

Conditions

Periodontitis

Study Design

• Observational Model Type: CASE_ONLY
• Study Time Perspective: CROSS_SECTIONAL

Eligibility Criteria

Inclusion Criteria

• The cases whose clinical, photographic, and radiological records are excellent quality and have good diagnostic sensitivity.

Exclusion Criteria

• Acute periodontal lesions
• gingival diseases
• the presence of dental implants

• Minimum Eligible Age: 18 Years
• Maximum Eligible Age: 65 Years
• Eligible Sex: ALL
• Accepts Healthy Volunteers: No

Sponsors

Necmettin Erbakan University

OTHER

Sponsor Role: lead

Responsible Party

Zeynep Tastan Eroglu

Assistant professor

Responsibility Role: PRINCIPAL_INVESTIGATOR

Locations

Zeynep Taştan Eroğlu

Konya, , Turkey (Türkiye)

Countries

Turkey (Türkiye)

References

Hamp SE, Nyman S, Lindhe J. Periodontal treatment of multirooted teeth. Results after 5 years. J Clin Periodontol. 1975 Aug;2(3):126-35. doi: 10.1111/j.1600-051x.1975.tb01734.x.

Reference Type: RESULT

PMID: 1058213 (View on PubMed)

Miller PD Jr. A classification of marginal tissue recession. Int J Periodontics Restorative Dent. 1985;5(2):8-13. No abstract available.

Reference Type: RESULT

PMID: 3858267 (View on PubMed)

Papapanou PN, Sanz M, Buduneli N, Dietrich T, Feres M, Fine DH, Flemmig TF, Garcia R, Giannobile WV, Graziani F, Greenwell H, Herrera D, Kao RT, Kebschull M, Kinane DF, Kirkwood KL, Kocher T, Kornman KS, Kumar PS, Loos BG, Machtei E, Meng H, Mombelli A, Needleman I, Offenbacher S, Seymour GJ, Teles R, Tonetti MS. Periodontitis: Consensus report of workgroup 2 of the 2017 World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions. J Periodontol. 2018 Jun;89 Suppl 1:S173-S182. doi: 10.1002/JPER.17-0721.

Reference Type: RESULT

PMID: 29926951 (View on PubMed)

Tonetti MS, Greenwell H, Kornman KS. Staging and grading of periodontitis: Framework and proposal of a new classification and case definition. J Periodontol. 2018 Jun;89 Suppl 1:S159-S172. doi: 10.1002/JPER.18-0006.

Reference Type: RESULT

PMID: 29926952 (View on PubMed)

Tastan Eroglu Z, Babayigit O, Ozkan Sen D, Ucan Yarkac F. Performance of ChatGPT in classifying periodontitis according to the 2018 classification of periodontal diseases. Clin Oral Investig. 2024 Jun 29;28(7):407. doi: 10.1007/s00784-024-05799-9.

Reference Type: DERIVED

PMID: 38951256 (View on PubMed)

Related Links

http://arxiv.org/pdf/2005.14165.pdf

Brown T, Mann B, Ryder N, et al. Language models are few-shot learners. Advances in neural information processing systems 2020;33:1877-1901.

Other Identifiers

NEUandChatCPT

Identifier Type: -

Identifier Source: org_study_id