Search

International audience; 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.

International audience; Convolutional neural networks (CNN) have established state-of-the-art performance in computer vision tasks such as object detection and segmentation. One of the major remaining challenge concerns their ability to capture consistent spatial attributes, especially in medical image segmentation. A way to address this issue is through integrating localization prior into system architecture. The CoordConv layers are extensions of convolutional neural network wherein convolution is conditioned on spatial coordinates. This paper investigates CoordConv as a proficient substitute to convolutional layers for organ segmentation in both fully and weakly supervised settings. Experiments are conducted on two public datasets, SegTHOR, which focuses on the segmentation of thoracic organs at risk in computed tomography (CT) images, and ACDC, which addresses ventricular endocardium segmentation of the heart in MR images. We show that if CoordConv does not significantly increase the accuracy with respect to standard convolution, it may interestingly increase model convergence at almost no additional computational cost.

International audience; Medical image segmentation is the process of anatomically isolating organs for analysis and treatment. Leading works within this domain emerged with the well-known U-Net. Despite its success, recent works have shown the limitations of U-Net to conduct segmentation given image particularities such as noise, corruption or lack of contrast. Prior knowledge integration allows to overcome segmentation ambiguities. This paper introduces BB-UNet (Bounding Box U-Net), a deep learning model that integrates location as well as shape prior onto model training. The proposed model is inspired by U-Net and incorporates priors through a novel convolutional layer introduced at the level of skip connections. The proposed architecture helps in presenting attention kernels onto the neural training in order to guide the model on where to look for the organs. Moreover, it fine-tunes the encoder layers based on positional constraints. The proposed model is exploited within two main paradigms: as a solo model given a fully supervised framework and as an ancillary model, in a weakly supervised setting. In the current experiments, manual bounding boxes are fed at inference and as such BB-Unet is exploited in a semi-automatic setting; however, BB-Unet has the potential of being part of a fully automated process, if it relies on a preliminary step of object detection. To validate the performance of the proposed model, experiments are conducted on two public datasets: the SegTHOR dataset which focuses on the segmentation of thoracic organs at risk in computed tomography (CT) images, and the Cardiac dataset which is a mono-modal MRI dataset released as part of the Decathlon challenge and dedicated to segmentation of the left atrium. Results show that the proposed method outperforms state-of-the-art methods in fully supervised learning frameworks and registers relevant results given the weakly supervised domain.