Department of Computer Engineering, University of Sharjah, Sharjah, United Arab Emirates.
Electrical and Electronics Engineering, University of Sharjah, Sharjah, United Arab Emirates.
Oral Radiol. 2024 Apr;40(2):165-177. doi: 10.1007/s11282-023-00717-3. Epub 2023 Dec 4.
Dental radiographs, particularly bitewing radiographs, are widely used in dental diagnosis and treatment Dental image segmentation is difficult for various reasons, such as intricate structures, low contrast, noise, roughness, and unclear borders, resulting in poor image quality. Recent developments in deep learning models have improved performance in analyzing dental images. In this research, our primary objective is to determine the most effective segmentation technique for bitewing radiographs based on different metrics: accuracy, training time, and the number of training parameters as a reflection of architectural cost.
In this research, we employ several deep learning models, namely Resnet-18, Resnet-50, Xception, Inception Resnet v2, and Mobilenetv2, to segment bitewing radiographs. The process begins by importing the radiographs into MATLAB®(MathWorks Inc), where the images are first improved, then segmented using the graph cut method based on regions to produce a binary mask that distinguishes the background from the original X-ray.
The deep learning models were trained on 298 and 99 radiograph training and validation sets and were evaluated using 99 images from the testing set. We also compare the segmentation model using several criteria, including accuracy, speed, and size, to determine which network is superior. Furthermore, we compare our findings with prior research to provide a comprehensive understanding of the advancements made in dental image segmentation. The accurate segmentation achieved was 93.67% and 94.42% by the Resnet-18 and Resnet-50 models, respectively.
This research advances dental image analysis and facilitates more accurate diagnoses and treatment planning by determining the best segmentation technique. The outcomes of this study can guide researchers and practitioners in selecting appropriate segmentation methods for practical dental image analysis.
牙科射线照相术，特别是咬翼射线照相术，在牙科诊断和治疗中被广泛应用。由于结构复杂、对比度低、噪声、粗糙度和边界不清晰等原因，牙科图像分割具有一定难度，导致图像质量较差。深度学习模型的最新发展提高了分析牙科图像的性能。本研究的主要目的是根据不同的度量标准，确定最有效的咬翼射线照相术分割技术：准确性、训练时间和训练参数的数量，以反映架构成本。
在本研究中，我们使用了几种深度学习模型，即 Resnet-18、Resnet-50、Xception、InceptionResnet v2 和 Mobilenetv2，对咬翼射线照相术进行分割。该过程首先将射线照相术导入 MATLAB®（MathWorks Inc），在该软件中，首先对图像进行改进，然后使用基于区域的图割方法对图像进行分割，生成一个二进制掩模，将背景与原始 X 射线区分开来。
使用 298 个和 99 个射线照相术训练和验证集对深度学习模型进行训练，并使用来自测试集的 99 个图像对其进行评估。我们还使用多种标准（包括准确性、速度和大小）比较了分割模型，以确定哪种网络更优。此外，我们将研究结果与先前的研究进行比较，以全面了解牙科图像分割的进展。Resnet-18 和 Resnet-50 模型的准确分割率分别为 93.67%和 94.42%。
本研究通过确定最佳分割技术，推进了牙科图像分析，有助于更准确的诊断和治疗计划。本研究的结果可以为研究人员和从业者在选择实际牙科图像分析的适当分割方法时提供指导。
