Hepatic vascular network segmentation for liver surgery planning
release_hjkwmjsv55egrnmgeh62ysssna
by
Olivér Benis, Csaba Benedek
2021
Abstract
Locating the blood vessels in the liver is a crucial task for surgery planning. Liver resection is performed to remove tumors from the organ or for transplantation. The location of the cut is selected with taking the position of the vessels into account to ensure intact blood flow after the surgery. Contrast enhanced computed tomography is suitable to be used for vessel network segmentation. Doing this process manually is time consuming and prone to errors, therefore in the recent years both traditional and deep learning based methods were proposed for automatic vessel extraction.<br> During my work I have developed a new method for automatic liver vessel network segmentation that is based on a marked point process (MPP) model and region growing algorithm. A synthetically generated training dataset was created consisting of artificial<br> CT images and the vascular network was annotated by the program during the generation process which means that neither real CT images nor manual segmentation was required. The model parameters can be determined fully automatically using a low number of<br> training images of the synthetic CT scan dataset. The low number of parameters give us the possibility to modify the model if adaptation is needed for a new dataset, with possibly different CT image characteristics. The method was tested on multiple CT image datasets both qualitatively and quantitatively and it was compared to other methods. Based on the experiments, the technique is showing promising results: while it is easily trainable it also comes near to the state-of-the-art performance. The adaptivity and robustness of the proposed 3D vessel network extraction algorithm was also demonstrated with testing its applicability for vessel segmentation in the lungs. As the liver resection surgery planning depends directly on the precisely delineated liver segments I have developed a method that is able to locate these segments using the liver vascular network data. Only minimal human interaction is needed for the algorithm to [...]
In text/plain
format
Archived Files and Locations
application/pdf 5.0 MB
file_uojnn6tggrhjbhz45vsx2pa37u
|
zenodo.org (repository) web.archive.org (webarchive) |
article-journal
Stage
published
Date 2021-12-10
access all versions, variants, and formats of this works (eg, pre-prints)
Datacite Metadata (via API)
Worldcat
wikidata.org
CORE.ac.uk
Semantic Scholar
Google Scholar