Virtual planning of liver resections: image processing, visualization and volumetric evaluation

https://doi.org/10.1016/S1386-5056(98)00162-2Get rights and content

Abstract

Operability of a liver tumor depends on its three dimensional relation to the intrahepatic vascular trees as well as the volume ratio of healthy to tumorous tissue. Precise operation planning is complicated by anatomic variability and distortion of the vascular trees by the tumor or preceding liver resections. We have developed a computer based 3D virtual operation planning system which is ready to go in routine use. The main task of a system in this domain is a quantifiable patient selection by exact prediction of post-operative liver function. It provides the means to measure absolute and relative volumes of the organ structures and resected parenchyma. Another important step in the pre-operative phase is to visualize the relation between the tumor, the liver and the vessel trees for each patient. The new 3D operation planning system offers quantifiable liver resection proposals based on individualized liver anatomy. The results are presented as 3D movies or as interactive visualizations as well as in quantitative reports.

Introduction

Selection of the patients to be surgically treated out of the continuum of metastatic tumor progression is the crucial step in liver resection planning. The surgical strategy is dependent on the localization of the liver tumor in relation to the three liver vascular trees. In detail three cornerstones have to be considered in liver surgery:

  • 1.

    The liver tumor has to be resected completely within sound liver parenchyma.

  • 2.

    Segments or even subsegments of the liver which might get devascularized by liver resection have to be removed as well.

  • 3.

    Liver function is dependent on sufficient functionally active liver volume.

In patients with multiple liver lesions operation planning is getting difficult as resection planes may cut in a complicated way through the vascular trees leading to devascularized (sub-)segments which have to be removed as well. Distorted anatomy after preceding liver resection and consecutive hypertrophy of the remaining liver parenchyma demands high skills in imagination and operation planning is not quantifiable.

A prediction of absolute or relative rest liver volume has been very difficult and has widely been substituted by simple imagination and rough estimation. Up to now a computer based calculation was not practicable due to missing analysis systems dedicated to tumor resections. A variety of efforts has been undertaken in the past to establish a computer based 3D presentation system for the liver anatomy of individual patients [1], [2]. To our knowledge none of these is available in clinical routine.

The knowledge of the exact three-dimensional relation of the tumor and the vascular trees within the liver decides on the preoperative selection of patients for hepatic resection. Despite low perioperative mortality there is still a high complication rate. Therefore objective and quantifiable selection criteria should be mandatory.

The aim of the developments was to establish a computer based, quantitative, three-dimensional operation planning system in liver surgery. An individualized operation strategy is facilitated which takes into account the anatomic variability of liver architecture in the individual patient. The classical approximation of the liver anatomy is given by the Couinaud model which divides the liver into eight segments [3]. These segments shown in Fig. 1 represent areas that are vascularly independent.

A precise operation planning system has to be based on the exact knowledge of the individual liver anatomy. Patients with multiple liver metastases may benefit by minimizing resection volume without devascularizing sound parenchyma.

The three major goals have been:

  • To recognize and visualize parenchyma, vessels and tumor.

  • To calculate the absolute and relative volumes and thus obtain quantified measurements.

  • To calculate a resection proposal, which may be modified by the user interactively.

These developments might optimize patient selection as operation planning is getting quantifiable and objective. Complication rates might decrease due to an improved patient selection and by a highly individualized operation planning and resection. Besides the advantages for the patient there might be a reduction of the expenditure for intensive care and hospital stay (Fig. 2).

Section snippets

Imaging

CT scans taken from routine diagnostics are used as the input for image processing. According to the actual problem two different methods of contrast agent based techniques may be used:

  • 1.

    Bolus tracking: contrast agent injection using a high pressure pump is triggered by a computer so that the contrast medium reaches the liver just in time when the spiral-CT-scan starts sampling the data from the liver. The cross-section pictures are reconstructed in 2 mm slices.

  • 2.

    CTAP (arterial portal CT): An

Discussion

Despite the fact that 3D rendering of CT-scans of the liver are carried out for several years no virtual liver resection planning system has been available for use in clinical practice. Quality of patient selection is dependent on a precise quantifiable tumor resection planning. Security margins around the tumor, recognition of vascularly dependent liver segments and sufficient functionally active remaining liver volume are the cornerstones of resection planning.

We have developed a computer

Acknowledgements

This work is supported by the Tumorzentrum Heidelberg/Mannheim and the University of Heidelberg.

References (15)

There are more references available in the full text version of this article.

Cited by (69)

  • Automatic CT liver Couinaud segmentation based on key bifurcation detection with attentive residual hourglass-based cascaded network

    2022, Computers in Biology and Medicine
    Citation Excerpt :

    Some traditional methods [16] utilize the morphological anatomy of the external appearance of the liver to divide the 3D liver surface into different units, but they usually ignore the internal features of vessels that are very important in liver surgery. In recent years, based on the hepatic vascular system, some computer-assisted Couinaud segmentation approaches have been developed, which are generally classified into two categories 1) portal vein based methods [6,17–19]: each segment is defined by the topology of its autonomous portal vessel supply tree; 2) plane based methods [5,12,14,16,20–22]: according to the main stems of the portal and venous trees, planes that best fit each of the three branches of the hepatic veins and the plane that best fits the portal vein are constructed to define the subdivision of the liver into different functional units. A comparison of these two types of methods can be found in Ref. [23].

  • Fully-automated functional region annotation of liver via a 2.5D class-aware deep neural network with spatial adaptation

    2021, Computer Methods and Programs in Biomedicine
    Citation Excerpt :

    In order to improve annotation efficiency, computer-aided annotation has been developed. Computer-aided annotation includes the following two categories: anatomical landmarks-based methods [2,4–7,14] and vascular tree-based methods [3,8–13,16,17]. The anatomical landmarks-based method is easy to operate, but it has difficulty in selecting marker points while neglecting the influence of vascular changes on plane estimation.

  • How many CT detector rows are necessary to perform adequate three dimensional visualization?

    2010, European Journal of Radiology
    Citation Excerpt :

    In fact, the diagnostic quality of these MDCT was shown in studies, some of them performed in a randomized fashion [8–10]. We and others were dealing with three dimensional (3D) visualization of the liver anatomy based on CT data for many years [11–15]. The main goal of this research was to increase anatomical and surgical understanding for challenging surgical procedures such as liver transplantations [11,16–19].

View all citing articles on Scopus
View full text