Publication: Evaluation of Convolutional Neural Network based on Dental Images for Age Estimation
Date
2019
Authors
Alkaabi S.
Yussof S.
Al-Mulla S.
Journal Title
Journal ISSN
Volume Title
Publisher
Institute of Electrical and Electronics Engineers Inc.
Abstract
Age estimation is a crucial component of medical forensic. Age can be estimated using various biological features such as face, bones, skeletal and dental structures. This paper attempts to evaluate the use of dental images for age estimation. Age estimation using image processing algorithms have seen a great deal of transformation after the invention of machine learning and deep learning algorithms. This transformation is supported by large scale availability of labelled image data and complex computational units which can process these data. This paper attempts to evaluate various Convolutional Neural Network (CNN) architectures for age estimation using dental panoramic X-ray Images. The evaluations use CNN for end to end to address drawback of automated age estimation in forensic dentistry without any transformations. The custom dataset of more than 2000 X-ray images divided to 7 different classes is used for training the CNN architectures. The concept of transfer learning is also used for training the popular CNN architectures like AlexNet, VGGNet and ResNet for age estimation. The performance of age estimation is evaluated by analysing its recall, precision, F1-scor, accuracies and average accuracies for all the architectures have performed. Due to rotation and tilt orientation, overlap teeth, missing teeth, our investigation yielded low accuracy with less than 40% in using dental images for age estimation using CNN architectures. To the best of our knowledge, this is the first paper that attempts to predict age estimation from dental images using Capsule-Net. However, the proposed architecture shows that Capsule Network has improved 36% than CNNs and transfer learning to achieve totally 76%. � 2019 IEEE.
Description
Convolution; Convolutional neural networks; Deep learning; Deep neural networks; Forensic science; Learning algorithms; Magnetic resonance imaging; Metadata; Transfer learning; Age estimation; Biological features; Computational units; Dental Panoramic X-ray images; Image processing algorithm; Magnetic Resonant Imaging (MRI); Proposed architectures; X-ray image; Network architecture