Evaluation of Transfer Learning with CNN to classify the Jaw Tumors
Ahmed Khalid Ismael, AbdulSattar Mohammed Khidhir · IOP Conference Series Materials Science and Engineering · 2020
Abstract Artificial Intelligence” (AI) This term refers to the idea that the machines can perform human tasks. Recently, researchers, professionals and companies around the world introduce deep learning and image processing systems that can analyze hundreds of X-Ray and Computer Tomography (CT) images rapidly to speed up the diagnosis of medical image and help to contain them. Dental diseases analysis is among the most innovative research fields, offering diagnostic and decision-making facilities for a variety of diseases, such as oral and maxillofacial diseases. Inside this paper, we present a comparison of recent architectures of the Deep Convolutional Neural Network (DCNN) for the automatic classification of two diseases depending on transfer learning with fined tuned using a pre-trained network (VGG16, VGG19). The proposed work was tested using a small scale X-Ray panoramic dataset containing 116 images (58 ameloblastoma and 58 Complex Odontoma). As a result, we can assume that the pre-trained network (VGG19) demonstrates highly satisfactory results with a rate of increase in the accuracy of training and validation. Unlike CNN, pre-trained network (VGG16) demonstrates less performance when a small image dataset is available.