bul 2004 
zh cu 
surement 
Sensing, 
4 — 269, 
i mmetric 
urnal of 
286-294, 
Invasive 
ournal of 
1998. 
yondence 
Journal, 
no, R., A 
| Light, 
al Stereo 
ormation 
ation of 
rnational 
Vol. 30, 
, ]SPRS 
NEW METHODS FOR LEAK DETECTION AND CONTOUR CORRECTION 
IN SEEDED REGION GROWING SEGMENTATION 
T. Heimann, M. Thorn, T. Kunert, H.-P. Meinzer 
German Cancer Research Center, Heidelberg, Germany 
Email: t.heimann@dkfz.de 
KEY WORDS: Medicine, Application, Segmentation, Algorithms, Correction 
ABSTRACT: 
Segmentation, i.e. the labelling of objects in image data, is a crucial step in many medical imaging processing tasks, e.g. operation 
planning, radio therapy or diagnostics. Seeded region growing is a basic yet effective method for semi-automatic segmentation. Its 
major drawback is that poor contrast at the organ edges may result in leaks, letting the area grow far beyond the region of interest. 
We developed a new method to reliably detect the origins of these leak regions and correct the segmentation results. Our approach is 
based on the observation that leaks are normally characterized by a narrow bottleneck connection to the seed area. We detect this 
bottleneck by specifying a single point somewhere within the erroneous area and tracing a path back to the seed point, moving along 
the skeleton of the segmented region. The point on the skeleton with the minimal distance to the contour marks the bottleneck. To 
cope with minor deviations from the desired result (not featuring a characteristic bottleneck), we optimized our freehand tool to 
minimize the necessary user interaction. Instead of separately cutting or merging parts of the segmentation, the new tool allows 
modifications by replacing entire parts of the contour in a single step. 
KURZFASSUNG: 
Segmentierung, d.h. die Markierung bestimmter Objekte in Bilddaten, ist ein kritischer Schritt bei vielen Aufgaben in der 
medizinischen Bildverarbeitung, wie Operationsplanung, Strahlentheraphie oder Diagnose. Das Regionenwachstumsverfahren ist 
eine grundlegende, aber wirkungsvolle Methode zur halbautomatischen Segmentierung. Der grófte Nachteil liegt darin, dass 
niedriger Kontrast an Organründern zum Auslaufen führen kann und die segmentierte Flüche weit über den interessierenden Bereich 
hinaus wüchst. Wir haben eine neue Methode entwickelt, um den Ursprung dieser ausgelaufenen Regionen zuverlássig zu finden und 
das Ergebnis zu korrigieren. Unser Ansatz basiert auf der Beobachtung, dass die ausgelaufene Region normalerweise nur eine 
schmale Verbindung zur ursprünglichen Flüche besitzt. Dieser Flaschenhals kann identifiziert warden, indem man einem beliebigen 
Punkt innerhalb der ausgelaufenen Flüche ein Pfad zum Ursprung entlang des Skeletts der Segmentierung berechnet wird. Der Punkt 
auf dem Skelett mit der niedrigsten Distanz zur Kontur markiert den Flaschenhals. Um kleinere Abweichungen vom gewünschten 
Segmentierungsergebnis zu korrigieren, haben wir unser Freihandwerkzeug optimiert, um die notwendige Benutzerinteraktion zu 
minimieren. Anstatt einzelne Teile der Region hinzuzufügen oder zu lóschen, ersetzt das neue Werkzeug komplette Abschnitte der 
Kontur in einem einzigen Schritt. 
1. INTRODUCTION for specialized applications, one of which is the segmentation 
of structures of interest that is performed with the Segmenta 
During the last years, a growing effort in medical imaging plugin (Kunert et al, 2004). Currently, this plug-in is mainly 
research has been put into the development of fully automatic used for segmenting CT images for liver operation planning 
methods for segmentation to be able to cope with the cver (Meinzer et al, 2002), a task which has been fully integrated 
growing amounts of data modern CT and MRT scanners are into the clinical workflow. The most time-consuming part of 
able to produce. While this trend is important and paving the the planning is the segmentation of liver tissue, which is 
way into the future, it is often overlooked that in clinical performed on a slice by slice basis using semi-automatic tools. 
practice, the methods of choice are still classic semi-automatic In order to optimize the liver operation planning workflow, a 
tools like freehand segmentation, snakes or region growing. selection of these tools was reengineered with a main focus on 
One reason for this situation might be that the automatic the concepts of usability and ergonomics (Nielsen, 1994). An 
methods are not yet robust enough for the clinical practice, important source of input for enhancements and simplifications 
another one that radiologists prefer more personal control and was a user survey that has been conducted with experts who 
flexibility when working with their images. But whatever the have been using the older version of the segmentation software 
reasons may be: To support the clinical workflow of today, it is extensively. 
essential to enhance and improve the tools which are actually In section 2 of this article, the new developments are presented 
used. from a technical point of view. Section 3 exposes the results of 
This article presents two examples of such enhancements that a comparison study between the old and the new versions of the 
have been developed for use in the surgery department of the tools and section 4 closes the article with a conclusion. 
university clinic of Heidelberg. Here, radiologists utilize the 
teleradiology and PACS system Chili (Engelmann ct al, 2000) 
for image access. This software implements a plug-in concept