期刊
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
卷 78, 期 -, 页码 -出版社
ELSEVIER SCI LTD
DOI: 10.1016/j.bspc.2022.103961
关键词
Dental Caries Segmentation and Detection; Fused Optimal Centroid K-means with K-Mediods Clustering; Meta-heuristic-based ResneXt-RNN; Hybrid Sea Lion-Squirrel Search Optimization
This paper presents a new segmentation model for dental caries using deep learning methods, as well as the superior performance of the well-trained M-ResneXt-RNN in caries detection.
The early identification of dental caries is necessary for corresponding treatments and the above reason, the bitewing radiography is utilized to offer initial caries detection. In clinical imaging, the usage of deep structured architectures with renowned neural network schemes helps process the vast amount of images, which has been researched actively in recent years and provided competitive performance. Therefore, deep learning approaches have attained remarkable diagnosis efficiency in the domain of radiology. Owing to this emerging intelligence, this paper aims to use the deep learning method for dental caries segmentation in an effective way. At first, contrast enhancement via Contrast Limited Adaptive Histogram Equalization (CLAHE), and noise filtering via bilateral filtering are performed under the pre-processing phase. Further, the segmentation of the caries is performed by the Fused Optimal Centroid K-means with K-Mediods Clustering (FOC-KKC), which will be enhanced by the Hybrid Sea Lion-Squirrel Search Optimization (HSLnSSO), inducing the best parameter opti-mization. Once the caries are segmented, the post-pre-processing of images is done by morphological operations. Finally, the detection of caries from the segmented image is employed by the meta-heuristic-based ResneXt with Recurrent Neural Network (RNN) (M-ResneXt-RNN), where the architecture modification is performed by the HSLnSSO algorithm. The new segmentation model and well-trained M-ResneXt-RNN for caries detection have exhibited superior performance when compared to the conventional techniques.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据