Detection and diagnosis of teeth periapical lesions using machine learning techniques /
اكتشاف و تشخيص آفات الأنسجه تحت الجذرية للأسنان بإستخدام تقنيات التعليم الآلى
Yasmine Eid Mahmoud Yousef ; Supervised Hoda M. O. Mokhtar , Soha Safwat Labib
- Cairo : Yasmine Eid Mahmoud Yousef , 2016
- 87 Leaves : charts , facsimiles ; 30cm
Thesis (M.Sc.) - Cairo University - Faculty of Computers and Information - Department of Information Systems
Dentists diagnose teeth periapical lesion according to patients dental x-ray, but in most time there is a problematic issue to reach a definitive diagnosis. They exert a lot of effort to reach the ideal clinical diagnosis. Even with these difficulties, sometimes reaching definitive diagnosis before starting the treatment is still difficult to achieve. Therefore, the objective of this research is to propose a new machine learning based methodology that can help dentists achieve their goals in easier and more accurate way. The proposed methods predict whether the patient has teeth periapical lesion or not and its type an eventually the dentist can select the best treatment. The proposed system consists of four main steps: 1) Data collection, 2) data preprocessing, 3) feature extraction, and finally 4) classification. Median and average filters and Histogram equalization were used in data preprocessing for removing noise and image enhancement. Image segmentation with expectation maximization algorithm or discrete wavelet transform was used for feature extraction. Feed forward neural networks and K-Nearest neighbor classifier were used for classification. Experiments were conducted using a real dataset. These experiments show that for detecting whether there was an infection or notimage segmentation with expectation maximization or 2D wavelet transform gave us the same results, and K-NN classifier performs better than feed forward neural network. However, for detecting the teeth periapical lesion type (AAP, APA, CPA or Normal) 2D wavelet transform performs better than image segmentation with expectation maximization, and K-NN classifier performs better than Feed Forward Neural Network