Deep Learning is providing new solutions for medical image segmentation problems and becomes a key point for future clinical application. The study of vascular structures is a challenging task due to the extremely small size of the vessel structure, low SNR, and varying contrast in medical image data. In this study, we present an end-to-end deep learning segmentation method relying on the integration of vessel enhancement filters inside a 3-D U-Net based architecture. In particular, the raw data used in the learning process -or used as input- is preprocessed using these filters. The use of vesselness filters can significantly improve the contrast in raw images; this step can also help to improve segmentation decision especially on bifurcations. 3-D U-Net, Dense U-Net and MultiRes U-Net are pitted against each other in the vessel segmentation task with the public IRCAD dataset. Considering the integration of vesselness filters, the model parameters were optimized in order to identify the optimal configuration for fully automatic segmentation of hepatic vessels. In addition the three architectures were tested on full 3-D images and slabs (stacks of 2-D slices). The results showed that the most accurate setup is the full 3-D process, which provides the highest Dice for most of the considered models. The 3-D Dense U-Net also gives the best results compared to other models with or without vessel enhancement filters.

Accurate liver vessel segmentation is of crucial importance for the clinical diagnosis and treatment of many hepatic diseases. Recent state-of-the-art methods for liver vessel reconstruction mostly utilize deep learning methods, namely, the U-Net model and its variants. However, to the best of our knowledge, no comparative evaluation has been proposed to compare these approaches in the liver vessel segmentation task. Moreover, most research works do not consider the liver volume segmentation as a preprocessing step, in order to keep only inner hepatic vessels, for Couinaud representation for instance. For these reasons, in this work, we propose using accurate Dense U-Net liver segmentation and conducting a comparison between 3D U-Net models inside the obtained volumes. More precisely, 3D U-Net, Dense U-Net, and MultiRes U-Net are pitted against each other in the vessel segmentation task on the IRCAD dataset. For each model, three alternative setups that allow adapting the selected CNN architectures to volumetric data are tested, namely, full 3D, slab-based, and box-based setups are considered. The results showed that the most accurate setup is the full 3D process, providing the highest Dice for most of the considered models. However, concerning the particular models, the slab-based MultiRes U-Net provided the best score. With our accurate vessel segmentations, several medical applications can be investigated, such as automatic and personalized Couinaud zoning of the liver.

La segmentation des vaisseaux hepatiques est une tâche tres difficile en raison de la petite taille et de la complexite du reseau vasculaire. Dans cet article, nous proposons une etude comparative originale de plusieurs modeles d’apprentissage profond combines a des filtres de rehaussement (algorithmes Jerman, Frangi, Sato et RORPO) comme etape de pretraitement. Les modeles 3-D U-Net, 3-D Dense U-Net et 3-D MultiRes U-Net ont ete testes sur des images tomodensitometriques pour extraire les reseaux vasculaires avec et sans pretraitement sur des volumes hepatiques complets et sur des slabs (groupes de coupes 2-D). Les modeles ont ete testes sur la base de donnees publique IRCAD et le 3-D Dense U-Net a obtenu le meilleur coefficient Dice sur les donnees pretraitees (avec le filtre Jerman en particulier), par rapport aux donnees brutes. En plus de cette analyse numerique, nous proposons une inspection visuelle des resultats de segmentation, qui confirment la precision de ce modele profond base sur des filtres de rehaussement vasculaire

Deep learning (DL) is one of the most important machine learning methods which has achieved great success in the field of medical image analysis. DL teaches a computer model how to perform classification tasks directly from images, but since the acquisition problems of these images this method has lost its effectiveness mainly for the segmentation of complex structures such as vessels that are hardly or not visible in the raw data. Nowadays, researchers are trying to find solutions to these kinds of problem since the information of the local appearance of pixel are not enough. To illustrate the limits of using standard DL models for vessel reconstruction, we first show a comparative study based on the IRCAD dataset. This experiment motivates our study, wherein we provide a review of DL models which covers liver vessel segmentation and medical image processing, in order to confirm if these problems can be solved by DL, and discusses a new approach to guide the experts who want to use these approaches in their work. A Systematic Literature Review (SLR) was carried out. More than 40 papers were founded by manual search in Elsevier, Springer and Science Direct, IEEE, 20 primary studies were finally included. According to the literature studies, we will define the most pertinent articles related to DL applications for complex structure reconstruction. Results: DL based topological signature methods have better results than classical topics and DL based pixelwise.