Automatic teeth segmentation on panoramic X-rays using deep neural networks

Abstract

In order to build an intelligent dental care process that both facilitates the treatment and improves the diagnosis, an accurate tooth segmentation and recognition on panoramic X-ray images might prove helpful. Although many studies have been conducted on teeth segmentation, few methods allow to perform tooth recognition and numbering at the same time. The existing methods allowing both those processes rely on instance segmentation architectures. To fill some gaps in the area of dental image segmentation, we propose a novel approach of automatic joint teeth segmentation and numbering using the pioneer U-Net model. We are first to employ the conventional U-Net model and show its limitations to provide accurate segmentation, being affected by noisy pixels outside the teeth region and by missing teeth in the X-ray images. To overcome this problem and reduce the misclassifications, we use a bounding box prior at the level of the skip connections. Such an approach helps guiding the network to better locate the teeth, and hence improves the segmentation. To validate the effectiveness of the method, we have conducted two experiments on the DNS Panoramic Dataset: a first one using manual bounding boxes and another one relying on a preliminary step of object detection. The implemented networks were evaluated using the Dice coefficient index and our results showed that consideration of location information onto the skip connections improves the performances of the semantic segmentation by 5% to 10% in average Dice accuracy depending on the quality of the bounding box labels.

Detection and Numbering of Teeth in Panoramic Radiographic Images Using Deep Neural Networks

“…This value is the highest accuracy reported in studies. Among the studies that used U-Net, study of Nader et al with 543 samples has a Dice value of 90%, and in the study of Koch et al with 1500 samples, the Dice value is reported as 93.6% [12,14]. According to the network settings and its hyperparameters, in our proposed method, even though the number of samples is 527, the Dice value is 95.3%.…”

“…The accuracy of the identification has reached 99.7%. In Nader et al's study, tooth segmentation was investigated on the panoramic radiographic images by using the U-Net network [14]. The results of that study have been reported to be about 90% according to Dice.…”

Multiclass Segmentation Using Teeth Attention Modules for Dental X-Ray Images

“…This section presents a comprehensive analysis and discussion of the performance of the proposed model for dental segmentation. We compared our segmentation model to several well-established segmentation models, including the traditional U-Net [20], Attention U-Net [12], ResNet-50 Attention U-Net [39], Swin U-Net [40], and a Modified U-Net [10], which is identical to BB-Unet [11].The performance of the proposed model was evaluated using multiple critical metrics, including ACC, Dice Similarity Coefficient (DSC), JI, precision, recall, and specificity. It is noted that we aim to segment each tooth in an X-ray image into 32 categories based on the World Dental Federation (FDI) notation [42], where each tooth is categorized into #11 to #18, #21 to #28, #31 to #38 and #41 to #48.…”

“…Our proposed method required approximately 60 minutes for training to achieve the segmentation performance in Table II. Other methods achieve DSC values of 0.7602, 0.7846, 0.7875, 0.6348, and 0.9004 and run-times of 45, 48-, 48-, 68-, and 64-minutes training times for [20], [12], [39], [40] and [10], respectively. These values are obtained from the above-mentioned experimental setup.…”

“…Fan et al [13] successfully segmented lesions in colonoscopy images using boundary box filters. By focusing on specific areas of interest in medical images, Nader et al [10] demonstrated that boundary box techniques can enhance segmentation precision. In dental imaging, boundary box techniques have been used to segment teeth and other dental structures by identifying the regions of interest on panoramic radiography and dental cone-beam computed tomography (CBCT) [2].…”