Automated Separation of Caries Requiring Filling and Caries Requiring Root Canal from Full Dose and Low Dose Bitewing Images Using Machine Learning Tam Doz ve Düşük Doz Isirma Görüntülerinden Dolgu ve Kanal Tedavisi İhtiyacinin Makina Öǧrenmesi ile Otomatik Olarak Belirlenmesi
2024 Electrical, Electronics and Biomedical Engineering Conference at 15th National Conference on Electrical and Electronics Engineering, ELECO 2024, Bursa, Türkiye, 28 - 30 Kasım 2024, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/eleco64362.2024.10847112
- Basıldığı Şehir: Bursa
- Basıldığı Ülke: Türkiye
- Anahtar Kelimeler: Dentistry, Machine learning, Recommendation System, Segmentation
- Sağlık Bilimleri Üniversitesi Adresli: Evet
Özet
In this study, an innovative computer-aided system based on deep learning has been developed for caries requiring filling and root canal treatment. With this system, dentists will be able to determine the type of caries in the tooth and decide on the appropriate treatment for the detected caries. Various variations with augmented 1250 bitewing images were used to automatically detect caries types in the first molar teeth. In the test phase, full-dose and low-dose versions of 84 bitewing images were used. The accuracy rates produced by the Detectron- 2 network in classifying caries types in fully dosed bitewing and low-dose bitewing images labeled by experienced dentists were observed as % 95.03 and 87.29 %, respectively. Success rates in distinguishing between canal and filling treatments for full-dose and low-dose imaging were observed as 88.09 % and 61.90 %, respectively. Additionally, the successful results of the network in low-dose images will also enable a reduction in the radiation level applied to patients. As a result, the proposed system has been successful in distinguishing teeth requiring canal and filling treatment.