Preoperative Image Segmentation for Organ Visualization Using Augmented Reality Technology During Open Liver Surgery

With the emergence of Computed Tomography (CT) and Magnetic Resonance Imaging (MRI), three-dimensional images facilitate the generation of 3D models of a patient, providing a new practical and accurate assistance, particularly for surgical planning. These images can be manipulated to produce an accurate 3D representation of an organ. The reconstructed mesh can be used to generate and visualize a deformable model during surgical intervention using Augmented Reality (AR) technology. To obtain an efficient reconstruction, a segmentation of these medical images using deep learning architecture can be used to extract the target organ's properties. Many methods were proposed based on the captured pre-operative patient's CT scans. Generally, the segmentation process is done manually using image processing software. In this context several approaches were proposed, these methods are not efficient and need human interaction to select the segmentation area correctly. This work aims to develop a deep learning method using a Convolutional Neural Network (CNN) that captures the liver organ from a set of CT scans. Given preoperative patient-specific data (CT scans), the U-net architecture is implemented to detect the liver organ. As a result, the segmented 2D images are used to generate a 3D patient-specific liver model.