EMERGING ROLE OF AI IN DIAGNOSIS & TREATMENT PLANNING

EMERGING ROLE OF AI IN DIAGNOSIS & TREATMENT PLANNING OF PERIODONTAL DISEASES

1. EMERGING ROLE OF AI IN DIAGNOSIS & TREATMENT PLANNING Dr R VISWA CHANDRA MDS;DNB;PhD Prof and Head SVS Institute of Dental Sciences Mahabubnagar TS

2. What is AI? Artificial Intelligence (AI) refers to the simulation of human intelligence in machines that are programmed to think, learn, and make decisions. These systems can perform tasks that typically require human cognition including Learning Reasoning Perception and Autonomy

3. Classification of AI By Capability By Learning

4. Classification and Applicability By Capability By Learning

5. Protocols of AI 1. Problem Definition 2. Data Acquisition & Preprocessing 3. Model Selection 4. Training & Optimization 5. Evaluation 6. Deployment Staging & Grading

6. 1. Problem Definition 2. Data Acquisition & Preprocessing 3. Model Selection 4. Training & Optimization 5. Evaluation 6. Deployment Staging & Grading Accurately detect key periodontal landmarks and automate disease staging and grading. YOLOv8 uses an anchor-free approach, simplifying training and improving generalization.

7. DrOROSCOPE® Our Experiences

8. TRAUMATIC ULCER LICHEN PLANUS

9. Previous Model

10. Latest Model IMAGE CAPTURE AI Trained Model DIAGNOSIS PICTURES THROUGH LIVE CAMERA & STORED DATA Image Processing Feature Extraction Training GENERATION OF AI MODEL Data cleansing & Preprocessing Data assessed at three levels By AI model itself By Trained and Calibrated staff Expert through Cloud support Provisional Diagnosis Differential Diagnosis Generated Data Validation

11. Novelty MEASUREMENT OF KERATINISED MUCOSA IS THROUGH ARTIFICIAL INTELLIGENCE photographs taken before and after applying a disclosing agent will be used to identify band of keratinised mucosa around implants Aykol-Sahin G, Yucel O, Eraydin N, Keles GC, Unlu U, Baser U. Efficiency of oral keratinized gingiva detection and measurement based on convolutional neural network. J Periodontol. 2024 Jul 15. doi: 10.1002/JPER.24-0151. Epub ahead of print. PMID: 39007745. Laser Patterned Microcoagulation (LPM) is a laser-based technique that uses small, precise laser burns to create micro-scale coagulation zones in tissue Possible applications of LPM treatment of oral soft tissues include the management of gingival recession, stimulation of the growth and regrowth of interdental and inter-implant papillae.

12. Aykol-Sahin G, Yucel O, Eraydin N, Keles GC, Unlu U, Baser U. Efficiency of oral keratinized gingiva detection and measurement based on convolutional neural network. J Periodontol. 2024 Jul 15. doi: 10.1002/JPER.24-0151. Epub ahead of print. PMID: 39007745. Novelty Red Filter Blue Filter Green Filter

13. Clinical Validation Comprehensive analysis of AI system performance across multiple clinical studies demonstrates consistent superiority over traditional diagnostic methods. Patel, Maitri & Kumar, Santosh & Patel, Bhavin & Patel, Shirishkumar & Girdhar, Gaurav & Patadiya, Hiren & Hirani, Tanvi & Haque, Mainul. (2025). Review Article: Impact of Artificial Intelligence on Periodontology: A Review. Cureus. 17. e80636.. 10.7759/cureus.81162.

14. Clinical Validation Study / Source AI Model / Approach Dataset Clinical Task Validation Metrics Key Findings Do et al., BMC Oral Health (2025) YOLOv8 deep learning 500 panoramic radiographs Staging and grading periodontitis (2017 AAP classification) Accuracy, precision, recall (values not disclosed) High diagnostic accuracy; supports standardized staging and grading Jundaeng et al., Frontiers in Medical Technology (2025) Hybrid CNN + clinical data integration 1,200 patient records + radiographs Periodontal disease classification and risk prediction AUC: 0.91; Sensitivity: 0.88; Specificity: 0.85 Demonstrated strong performance in real-world clinical settings Ferrara et al., J Med Artif Intell (2025) Systematic review of DL models (CNNs, GANs, transformers) 30+ studies (2017–2025) Diagnosis, prognosis, treatment planning Varies; AUCs up to 0.95 in some models Emphasized need for prospective trials and external validation

15. Study / Source AI Model / Approach Dataset Clinical Task Validation Metrics Key Findings MDPI Review (2024) CNNs, GANs, transformer networks 22 studies Radiographic bone loss detection, lesion classification Accuracy: 85–95%; AUC: 0.87–0.96 AI models consistently outperformed manual assessments in speed and reproducibility JADA Systematic Review (2023) CNN-based radiographic analysis 15 studies Bone loss detection and lesion classification Sensitivity, specificity, AUC Reduced inter-observer variability; promising for clinical deployment Clinical Validation

16. THANK YOU