Diagnosing Gingiva Disease Using Artificial Intelligence Techniques
Uložené v:
| Názov: | Diagnosing Gingiva Disease Using Artificial Intelligence Techniques |
|---|---|
| Autori: | Rana Khalid Sabri, Lujain Younis Abdulkadir, AbdulSattar Mohammed Khidhir, Hiba Abdulkareem Saleh |
| Zdroj: | Diyala Journal of Engineering Sciences, Vol 18, Iss 2 (2025) |
| Informácie o vydavateľovi: | University of Diyala, College of Science, 2025. |
| Rok vydania: | 2025 |
| Predmety: | InceptionV3, Mechanics of engineering. Applied mechanics, Environmental engineering, MobileNet, TA213-215, TA349-359, TA170-171, Sequential, TK1-9971, Engineering machinery, tools, and implements, Chemical engineering, VGG16, TP155-156, Electrical engineering. Electronics. Nuclear engineering, Periodontal Diseases |
| Popis: | Gingival and periodontal diseases, such as gingivitis and periodontitis, are critical public health concerns that can lead to severe complications if left untreated. Early and precise diagnosis is crucial to mitigate the progression of these conditions and improve oral health outcomes. This study investigates the application of convolutional neural networks (CNNs) in diagnosing gingival diseases using medical images, including X-rays and intraoral photographs. Several CNN architectures, including VGG16, Sequential CNN, MobileNet, InceptionV3, and suggestions for a voting method to enhance the prediction, were evaluated for their performance in classifying gingival conditions. MobileNet emerged as the most effective model, achieving a test accuracy of 92.73%; the suggested method relies mainly on its positive result. When the MobileNet's result is false, the process takes the voting result using the other methods. This boosts the accuracy to 96%. Surpassing other models in precision and recall metrics. Pre-processing techniques such as normalization using the CIELAB color space and data augmentation significantly enhanced model accuracy. The study employed robust evaluation methods, including 10-fold cross-validation and hyperparameter tuning, to ensure model reliability and generalizability. The findings highlight the transformative potential of AI-powered diagnostic tools in dental healthcare. By leveraging lightweight and efficient architectures like MobileNet, these tools can be deployed in resource-limited settings, offering real-time diagnostic support to healthcare professionals. Future work will focus on expanding datasets, exploring ensemble models, and improving interpretability to further enhance diagnostic accuracy and clinical applicability. |
| Druh dokumentu: | Article |
| ISSN: | 2616-6909 1999-8716 |
| DOI: | 10.24237/djes.2024.18211 |
| Prístupová URL adresa: | https://doaj.org/article/2f189a7058dd4b0ca8892b94a08e6a87 |
| Rights: | CC BY |
| Prístupové číslo: | edsair.doi.dedup.....8fe3acf2d5b7fe59e49647e36c84ca5a |
| Databáza: | OpenAIRE |
Nájsť tento článok vo Web of Science | Abstrakt: | Gingival and periodontal diseases, such as gingivitis and periodontitis, are critical public health concerns that can lead to severe complications if left untreated. Early and precise diagnosis is crucial to mitigate the progression of these conditions and improve oral health outcomes. This study investigates the application of convolutional neural networks (CNNs) in diagnosing gingival diseases using medical images, including X-rays and intraoral photographs. Several CNN architectures, including VGG16, Sequential CNN, MobileNet, InceptionV3, and suggestions for a voting method to enhance the prediction, were evaluated for their performance in classifying gingival conditions. MobileNet emerged as the most effective model, achieving a test accuracy of 92.73%; the suggested method relies mainly on its positive result. When the MobileNet's result is false, the process takes the voting result using the other methods. This boosts the accuracy to 96%. Surpassing other models in precision and recall metrics. Pre-processing techniques such as normalization using the CIELAB color space and data augmentation significantly enhanced model accuracy. The study employed robust evaluation methods, including 10-fold cross-validation and hyperparameter tuning, to ensure model reliability and generalizability. The findings highlight the transformative potential of AI-powered diagnostic tools in dental healthcare. By leveraging lightweight and efficient architectures like MobileNet, these tools can be deployed in resource-limited settings, offering real-time diagnostic support to healthcare professionals. Future work will focus on expanding datasets, exploring ensemble models, and improving interpretability to further enhance diagnostic accuracy and clinical applicability. |
|---|---|
| ISSN: | 26166909 19998716 |
| DOI: | 10.24237/djes.2024.18211 |