Analysis of Intraoral Images Using Deep Learning: Classification Performance of CNN Models

Öz

The classification of intraoral images has become a critical aspect of modern dental diagnosis, particularly with the emergence of advanced imaging technologies and artificial intelligence (AI). The integration of transfer learning and ensemble techniques has shown promising results in enhancing the performance of models designed for this purpose. In this study, deep learning models such as ResNet152V2, DenseNet201, InceptionResNetV2, ConvNeXtBase, and Xception were tested individually and as an ensemble. The dataset includes intraoral images from individuals of various age groups. Data preprocessing, normalization, and fine-tuning techniques were applied during model training and evaluation. Performance analysis was conducted using accuracy, precision, recall, F1-score, and ROC-AUC metrics. The results indicate that the ResNet152V2 + DenseNet201 ensemble model achieved the highest accuracy (89.9%) and ROC-AUC score (0.9934). These findings highlight the great potential of deep learning-based approaches in dental applications, such as automatic diagnosis, treatment planning, and remote patient assessment. Furthermore, the development of AI-based systems for analyzing dental morphology and jaw structures can accelerate clinical workflows by providing decision-support mechanisms for dentists. Future studies should focus on expanding dataset diversity to improve model generalization.

Anahtar Kelimeler

• Deep learning
• Intraoral image analysis
• CNN models
• Dental classification

Ağız İçi Görüntülerin Derin Öğrenme İle Analizi: CNN Modellerinin Sınıflandırma Performansı

Öz

Ağız içi görüntülerin sınıflandırılması, özellikle gelişmiş görüntüleme teknolojileri ve yapay zekanın (AI) ortaya çıkmasıyla birlikte, modern diş teşhisinin kritik bir yönüdür. Transfer öğrenme ve topluluk tekniklerinin entegrasyonu, bu amaçla tasarlanan modellerin performansını artırmada ümit verici sonuçlar göstermiştir. Bu çalışmada, ResNet152V2, DenseNet201, InceptionResNetV2, ConvNeXtBase ve Xception gibi derin öğrenme modelleri tek tek ve topluluk modeli olarak test edilmiştir. Kullanılan veri seti, farklı yaş gruplarından bireylerin ağız içi görüntülerini içermektedir. Modellerin eğitimi ve değerlendirilmesi aşamalarında veri ön işleme, normalizasyon ve ince ayar (fine-tuning) gibi teknikler uygulanmıştır. Doğruluk, kesinlik, duyarlılık, F1 skoru ve ROC-AUC gibi metrikler kullanılarak performans analizleri gerçekleştirilmiştir. Sonuçlar, ResNet152V2 + DenseNet201 topluluk modelinin en yüksek doğruluk oranına (%89,9) ve 0,9934 ROC-AUC değerine ulaştığını göstermektedir. Bu bulgular, derin öğrenme tabanlı yaklaşımların diş hekimliği uygulamalarında otomatik teşhis, tedavi planlaması ve uzaktan hasta değerlendirme gibi alanlarda büyük potansiyele sahip olduğunu ortaya koymaktadır. Ayrıca, diş morfolojisi ve çene yapılarının analizi için yapay zeka tabanlı sistemlerin geliştirilmesi, klinik süreçleri hızlandırarak diş hekimlerine karar destek mekanizmaları sağlayabilir. Gelecekteki çalışmalar, daha büyük ve çeşitli veri setleri kullanarak modelin genelleme yeteneğini artırmaya odaklanmalıdır.

Anahtar Kelimeler

• Derin öğrenme
• Ağız içi görüntü analizi
• CNN modelleri
• Diş sınıflandırma

Kaynakça

1. Abunasser BS, Al-Hiealy MRJ. 2022. Prediction of instructor performance using machine and deep learning techniques. Semant Scholar, 21(2): 1-12.
2. Alalharith D, Alharthi H, Alghamdi W, Alsenbel Y, Aslam N, Khan I, Barouch K. 2020. A deep learning-based approach for the detection of early signs of gingivitis in orthodontic patients using faster region-based convolutional neural networks. Int J Environ Res Public Health, 17(22): 8447.
3. Al-Khannaq M, Nahidh M, Al-Dulaimy D. 2022. The importance of the maxillary and mandibular incisors in predicting the canines and premolars crown widths. Int J Dent, 2022: 1-6.
4. Chaudhary S, Shah P, Paygude P, Chiwhane S, Mahajan P, Chavan P, Kasar M. 2024. Varying views of maxillary and mandibular aspects of teeth: A dataset. Data Brief, 56: 110772.
5. Chicco D, Jurman G. 2020. The advantages of the Matthews correlation coefficient (MCC) over F1 score and accuracy in binary classification evaluation. BMC Genomics, 21(1): 1-13.
6. Chollet F. 2017. Xception: Deep learning with depthwise separable convolutions. Proc IEEE Conf Comput Vis Pattern Recognit (CVPR), Honolulu, HI, USA, pp: 1251-1258.
7. Corns S, Liversidge H, Fleming P. 2021. An assessment of root stage of canine and premolar teeth at alveolar eruption. J Orthod, 49(2): 122-128.
8. Elshennawy NM, Ibrahim DM. 2020. Deep-pneumonia framework using deep learning models based on chest X-ray images. Diagnostics (Basel), 10(9): 649.

9. García-Gil M, Alarcón J, Cacho A, Yáñez-Vico R, Palma-Fernández J, Martín C. 2023. Association between eruption sequence of posterior teeth, dental crowding, arch dimensions, incisor inclination, and skeletal growth pattern. Children (Basel), 10(4): 674.
10. Gupta A, Gupta S, Katarya R. 2021. InstaCovNet-19: A deep learning classification model for the detection of COVID-19 patients using chest X-ray. Appl Soft Comput, 101: 107064.
11. He K, Zhang X, Ren S, Sun J. 2016. Identity mappings in deep residual networks. Proc Eur Conf Comput Vis (ECCV), Amsterdam, Netherlands, pp: 123-135.
12. Huang G, Liu Z, Van Der Maaten L, Weinberger KQ. 2017. Densely connected convolutional networks. Proc IEEE Conf Comput Vis Pattern Recognit (CVPR), Honolulu, HI, USA, pp: 4700-4708.
13. Kanter R, Battistuzzi P, Truin G. 2018. Temporomandibular disorders: "Occlusion" matters! Pain Res Manag, 2018: 1-13.
14. Khalaf K, Brook A, Smith R. 2022. Genetic, epigenetic and environmental factors influence the phenotype of tooth number, size and shape: Anterior maxillary supernumeraries and the morphology of mandibular incisors. Genes (Basel), 13(12): 2232.
15. Khanagar S, Al-Ehaideb A, Maganur P, Vishwanathaiah S, Patil S, Baeshen H, Bhandi S. 2021. Developments, application, and performance of artificial intelligence in dentistry – a systematic review. J Dent Sci, 16(1): 508-522.
16. Kühnisch J, Meyer O, Hesenius M, Hickel R, Gruhn V. 2021. Caries detection on intraoral images using artificial intelligence. J Dent Res, 101(2): 158-165.
17. Liu Z, Lin Y, Cao Y, Hu H, Wei Y, Zhang Z, Lin S, Guo B. 2022. A ConvNet for the 2020s. Proc IEEE/CVF Conf Comput Vis Pattern Recognit (CVPR), New Orleans, LA, USA, pp: 11976-11986.
18. Palkit T, Aggarwal I, Malhotra Y, Uppal M, Goyal M, Singh N. 2020. The relationship between maxillary and mandibular base lengths and dental crowding in patients with true class II malocclusions. Int Healthc Res J, 4(7): OR5-OR9.
19. Rithanya M. 2023. Prevalence of dental caries in permanent second molars in teenagers using ICDAS. J Pharm Res Int, 35(35): 28-35.
20. Sitaula C, Shahi TB. 2022. Monkeypox virus detection using pre-trained deep learning-based approaches. J Med Syst, 46(1): 1-9.
21. Soundarya N, Jain V, Shetty S, Akshatha B. 2021. Sexual dimorphism using permanent maxillary and mandibular incisors, canines, and molars. J Oral Maxillofac Pathol, 25(1): 183-188.
22. Szegedy C, Ioffe S, Vanhoucke V, Alemi A. 2017. Inception-v4, Inception-ResNet and the impact of residual connections on learning. Proc AAAI Conf Artif Intell, San Francisco, USA, pp: 4278-4284.
23. Tan R, Zhu X, Chen S, Zhang J, Liu Z, Li Z, Yang L. 2024. Caries lesions diagnosis with deep convolutional neural network in intraoral qlf images by handheld device. BMC Oral Health, 24(1): 1-5.
24. Ueno K, Kumabe S, Nakatsuka M, Tamura I. 2019. Factors influencing dental arch form. Okajimas Folia Anat Jpn, 96(1): 31-46.
25. Vakili M, Ghamsari M, Rezaei M. 2020. Performance analysis and comparison of machine and deep learning algorithms for IoT data classification. arXiv Preprint, arXiv: 2001.09636.
26. Wang S, Shen Y, Fuh L, Peng S, Tsai M, Huang H, Hsu J. 2020. Relationship between cortical bone thickness and cancellous bone density at dental implant sites in the jawbone. Diagnostics (Basel), 10(9): 710.
27. Wang Z, Hu M, Zhai G. 2018. Application of deep learning architectures for accurate and rapid detection of internal mechanical damage of blueberry using hyperspectral transmittance data. Sensors (Basel), 18(4): 1126.
28. Wimalarathna A, Thilakumara I, Jayasinghe V, Amaratunga P, Jayasinghe R. 2021. Location of vital structures and available bone for the placement of dental implants in clinically edentulous patients: A cohort study. J Dent Implant Res, 40(4): 151-157.
29. Yotsuya M, Nakano M, Umehara K, Nomura T, Sato T. 2020. Implant and prosthetic treatment in esthetic zone with alveolar ridge preservation and autotransplantation: Clinical case report with 16-year follow-up. Bull Tokyo Dent Coll, 61(2): 145-150.