G.V Black dental caries classification and preparation technique using optimal CNN-LSTM classifier

Abstract

Dental caries is one of the oral diseases which are a major health problem for many people across the globe. It can lead to pain, discomfort, disfigurement, and even death in some cases. Dental caries is caused by the infection of the calcified tissue of the teeth. They can be prevented easily by early diagnosis and treated in the early stages. The development of a reliable model for the diagnosis and classification of dental caries can lead to effective and timely treatment. The G.V Black Classification system of dental caries is one of the systems which is widely accepted worldwide. It classifies caries into six classes based on the location of caries. This paper proposes a novel deep convolution layer network (CNN) with a Long Short-Term Memory (LSTM) model for the detection and diagnosis of dental caries on periapical dental images. The proposed model utilizes a convolutional neural network for extracting the features and Long Short term memory (LSTM) for conducting short-term and long-term dependencies. The main objective of this study is to detect dental caries and classify them into various classes based on G.V Black Classification. The periapical dental images are pre-processed and are fed as input to deep convolutional neural networks. The deep convolutional neural network classifies the input into various classes. The proposed algorithm is optimized using the Dragonfly optimization algorithm and gave an accuracy of 96%. Experiments are conducted to evaluate and compare the proposed model with the recent state-of-art deep learning models. This study justifies that a deep convolutional neural network is one of the most efficient ways to detect and classify dental caries into various G.V black classes. The achieved accuracy of the proposed optimal CNN-LSTM model for G.V black classification proves its efficacy as compared to the classification accuracy achieved by widely used pre-trained CNN models i.e. Alexnet (accuracy: 93%) and GoogleNet (accuracy: 94%) on the same database. The performance of the proposed CNN-LSTM model is further strengthened by comparing the results with the CNN model, 2 layer LSTM model and CNN-LSTM model without dragonfly optimization. The proposed optimal CNN-LSTM model shows the best performance with 96% accuracy and helps in dental image classification as the second opinion to the medical expert.