Classification of oral cysts is a crucial task as the similarity between cysts exists which requires a computer-aided diagnosis system. Panoramic dental image is one of the widely used images to identify dental cyst, periodontal bone defects, periapical lesions, and pathological jaw lesions. This article proposes a modified LeNet architecture in a convolutional neural network for classifying the oral cyst images and a morphology-based segmentation method for segmenting the cyst regions in the classified cyst images. A traditional data augmentation approach and a threefold cross-validation method are used to increase the number of input samples and evaluate the accurate results respectively. The proposed methodology is applied to the cyst images obtained from a dental hospital. This model achieves a classification rate of 99.63% for cyst classification and demonstrates a sensitivity of about 98.3% for cyst segmentation. The proposed work has been compared with state-of-the-art algorithms.

中文翻译：

---

使用卷积神经网络架构对全景牙科图像中囊肿进行分类和分割的性能分析

口腔囊肿的分类是一项至关重要的任务，因为囊肿之间存在相似性，这需要计算机辅助诊断系统。全景牙科图像是广泛用于识别牙囊肿、牙周骨缺损、根尖周病变和病理性颌骨病变的图像之一。本文提出了一种改进的卷积神经网络中的 LeNet 架构，用于对口腔囊肿图像进行分类，并提出一种基于形态学的分割方法，用于在分类的囊肿图像中分割囊肿区域。使用传统的数据增强方法和三重交叉验证方法分别增加输入样本的数量和评估准确结果。所提出的方法应用于从牙科医院获得的囊肿图像。该模型达到了 99 的分类率。63% 的囊肿分类和展示了约 98.3% 的囊肿分割敏感性。所提出的工作已与最先进的算法进行了比较。