Computer-assisted liver tumor surgery using a novel semiautomatic and a hybrid semiautomatic segmentation algorithm

Abstract

We developed a medical image segmentation and preoperative planning application which implements a semiautomatic and a hybrid semiautomatic liver segmentation algorithm. The aim of this study was to evaluate the feasibility of computer-assisted liver tumor surgery using these algorithms which are based on thresholding by pixel intensity value from initial seed points. A random sample of 12 patients undergoing elective high-risk hepatectomies at our institution was prospectively selected to undergo computer-assisted surgery using our algorithms (June 2013-July 2014). Quantitative and qualitative evaluation was performed. The average computer analysis time (segmentation, resection planning, volumetry, visualization) was 45 min/dataset. The runtime for the semiautomatic algorithm was <0.2 s/slice. Liver volumetric segmentation using the hybrid method was achieved in 12.9 s/dataset (SD ± 6.14). Mean similarity index was 96.2 % (SD ± 1.6). The future liver remnant volume calculated by the application showed a correlation of 0.99 to that calculated using manual boundary tracing. The 3D liver models and the virtual liver resections had an acceptable coincidence with the real intraoperative findings. The patient-specific 3D models produced using our semiautomatic and hybrid semiautomatic segmentation algorithms proved to be accurate for the preoperative planning in liver tumor surgery and effectively enhanced the intraoperative medical image guidance.

3D Liver Tumor Segmentation in CT Images Using Improved Fuzzy C-Means and Graph Cuts

“…Many different approaches have been proposed to improve the tumor segmentation performance. Recent publications mainly included semiautomatic and automatic tumor segmentation methods based on region growing or thresholding [ 6 – 8 ], clustering [ 9 – 11 ], level set [ 12 , 13 ], graph cuts [ 14 , 15 ], and machine learning [ 16 – 19 ].…”

Up-to-date intraoperative computer assisted solutions for liver surgery

“…Modern software solutions and systems have been developed that allow manual, automatic or semiautomatic segmentation of medical liver images with acceptable results[16,20-22]. Our team developed the PROMETHEIA system which uses a hybrid semiautomatic segmentation method that allow of a rapid and easy segmentation that can be performed even by a surgeon[7,23]. The patient-specific 3D models produced using our algorithms proved to be accurate and effectively enhanced the intraoperative medical image guidance.…”

“…Since most of the intraoperative discovered liver tumours have a diameter of less than 1.5 cm, they can be incorporated manually in 3D liver models as simple spheres. We have implemented this approach in our computer analysis system for liver surgery and test it in clinical practice[18,23]. Our results showed that the incorporation of newly found tumours was possible, safe and accurate and the intraoperative plan had to be changed in 21% of the cases.…”

Intra-Operative Modeling of the Left Atrium: A Simulation Approach Using Poisson Surface Reconstruction

“…In this eld, surface representation and tessellation 3 [51] are the predominant uses for Bézier surfaces.…”

“…To be sure, the major bottleneck is segmentation (currently is subject to extensive research), which automation is still considered as a challenging task [51], thus requiring some degree of human interaction. While segmentation falls out of the scope of this work, the integration ability of PSR and how this can increase the adoption of 3D patient-speci c models with the help of state-of-the-art tools, like MeVis Distant Services [51,52,53] or Fuji lm Synapse VINCENT [54], is a relevant question. PSR can be seamlessly integrated in any medical platform, provided that the platform is able to obtain segmented images (Figure 6.1).…”