[go: up one dir, main page]

WO2009001378A1 - Procédé de correction morphologique d'une segmentation de nodules pulmonaires juxta-vasculaires dans des images de tomodensitométrie - Google Patents

Procédé de correction morphologique d'une segmentation de nodules pulmonaires juxta-vasculaires dans des images de tomodensitométrie Download PDF

Info

Publication number
WO2009001378A1
WO2009001378A1 PCT/IT2007/000453 IT2007000453W WO2009001378A1 WO 2009001378 A1 WO2009001378 A1 WO 2009001378A1 IT 2007000453 W IT2007000453 W IT 2007000453W WO 2009001378 A1 WO2009001378 A1 WO 2009001378A1
Authority
WO
WIPO (PCT)
Prior art keywords
voxels
nodule
increase
fields
form region
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/IT2007/000453
Other languages
English (en)
Inventor
Stefano Diciotti
Simone Lombardo
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
IM3D SpA
Original Assignee
IM3D SpA
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by IM3D SpA filed Critical IM3D SpA
Priority to PCT/IT2007/000453 priority Critical patent/WO2009001378A1/fr
Publication of WO2009001378A1 publication Critical patent/WO2009001378A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/155—Segmentation; Edge detection involving morphological operators
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30061—Lung
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30061—Lung
    • G06T2207/30064—Lung nodule

Definitions

  • the present invention relates to a morphological correction method of a binary image generated by a computerised axial tomography (CAT) image.
  • CAT computerised axial tomography
  • the invention relates to a morphological correction method of a segmented image of a juxta-vascular pulmonary nodule obtained from a CAT image for characterisation of pulmonary nodules as defined in claim 1.
  • a pulmonary nodule in its first stages, from a radiological point of view, takes the form of a solitary, non-calcified nodule.
  • the repeated acquisition over a period of time of computerised tomography images makes it possible to observe the dimensional variations of a nodule; in fact, the growth is the most important indicator of malignity of a solitary pulmonary nodule with small dimensions, i.e. which has a mean diameter of 10mm or less.
  • the mean diameter of a nodule is defined as the mean of the maximum diameter and of the maximum diameter perpendicular to the maximum diameter, both measured on the computerised tomography image in which the nodule has the greatest transverse area.
  • these linear measurements do not take into consideration the real three-dimensional form of the nodule.
  • the volume of a nodule is typically determined after a stage of segmentation, hi the last few years, many algorithms have been developed for the segmentation of a pulmonary nodule, for example by means of the application of threshold methods (such as those described in J.P. Ko et al., "Small pulmonary nodules: volume measurement at chest CT- phantom study", Radiology, vol. 228, n°8, pp. 864-870, 2003).
  • the traditional threshold methods are suitable for distinguishing nodules which are well-defined and distinct from the pulmonary parenchyma in a computerised tomography image.
  • nodules which are in contact with other anatomical structures such as, for example, the blood vessels or the pleura
  • these nodules constitute most of the nodules which are identified in CAT examinations.
  • the nodules which are connected to a blood vessel are known in the literature as juxta- vascular nodules, and the nodules which are adjacent to the pleural wall are known as juxta-pleural nodules.
  • An object of the present invention is to propose an innovative method for morphological correction of a segmented image of pulmonary nodules of the juxta-vascular type, which can carry out the correction locally, whilst maintaining the natural irregularities of the morphology of the nodule unaltered in the segmentation result.
  • figure 1 is a schematic representation of an image of a pulmonary nodule and of the first map of the distances
  • figure 2 is a schematic representation of an image of the second map of the distances of the nodule
  • figure 3 is a representation of a tree of the connected regions
  • figure 4 is a graph of the development of the number of voxels, according to the distance from the region R 0
  • figure 5 is a graph similar to the graph in figure 4, for another path.
  • the method according to the invention makes it possible to carry out a three-dimensional analysis and a procedure of automatic correction of the segmentation of a juxta-vascular pulmonary nodule.
  • An initial segmentation is considered, which is carried out on the basis of an image generated by means of computerised tomography.
  • the voxel which is approximately the centre of the nodule is identified.
  • This operation can be carried out by means of the instructions given by an expert operator, or by means of use of a known computer-aided detection system.
  • a method based on multi-scale Gaussian Laplacian filters such as those which are present in S. Diciotti et al., "3D segmentation algorithm of small lung nodules in spiral CT images”. IEEE Transactions on Information Technology in Biomedicine;- S. Diciotti et al., "Multi-scale neural network system for lung nodules detection in spiral CT images: preliminary results", in Mediterranean Conference on Medical and Biological Engineering, IFMBE, 2004).
  • a volume of interest is extracted, which includes the pulmonary nodule itself in its entirety; taking into consideration the analysis of nodules with a mean diameter of between 3 and 10mm, the volume of interest is preferably a cube with sides of 25mm.
  • an anisotropic diffusive filter such as that described in P. Perona, J. Malik., "Scale-space and edge detection using anisotropic diffusion", IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 12, pp. 629-639, 1990
  • a fixed- threshold segmentation operation is carried out, preferably with a threshold value of -500 HU. It is extracted, from the resulting image, the region of voxels which have the greatest threshold value, and which is connected to the voxel previously indicated as the centre of the nodule. This last region constitutes the result of the initial segmentation of the nodule.
  • Figure 1 shows in a simplified manner a binary image of a section of the segmentation of a nodule 1, comprising in a known manner a plurality of voxels, some of which are represented in figure 1 as minute squares indicated as a whole by the reference 2.
  • the real voxels in the CAT images of the thorax typically have a parallelepiped form, with two sides of approximately 0,6mm and one side of approximately lmm, but the voxels 2 considered, on which the method according to the invention is to be applied, are obtained from the real ones by means of an over-sampling operation, which leads to obtain cubic voxels with sides of approximately 0,3mm.
  • a map of the geodetic distances MD 1 is constructed, with allocation to each voxel 2 of a value equal to the distance of the latter from an outline 3 of the nodule 1, considering as positive distances those which face towards the interior of the nodule 1.
  • First lines 4 exemplify the development of the values of the map of the distances MD 1 .
  • Each line 4 is constituted by the voxels 2 which have the same value of distance from the outline 3.
  • the maximum value of this map MD 1 represents the voxel 2 which is furthest from the outline 3, and is considered as the "seed" 5 for the subsequent stages of processing of the image.
  • a region which consists of all the voxels 2 which have distance values between the maximum value (corresponding to the "seed” 5) and the maximum value reduced by a predetermined quantity T, as described hereinafter. It is therefore obtained a form region R 0 6, defined by an outline 8, which represents the global form of the nodule 1.
  • the quantity T is such that the form region R 0 6 is large enough to be able still to describe the global form of the nodule 1, and at the same time small enough not to contain the irregularities produced on the form of the nodule 1 by the blood vessels.
  • the value T can be selected as 1/6 the diameter of the smallest nodule from amongst those concerned.
  • the value of T is preferably 0,5mm.
  • Figure 2 shows the outline 3 of the nodule 1, the form region Ro 6, and an example of the second map of distances MD 2 represented by means of second lines 7, each consisting of the voxels 2 which have the same value of distance from the outline 8 of the form region Ro 6.
  • These second lines 7 define a set of regions of increase 7a, 7b, ...., 7n, the form of which is determined by the form region R 0 6.
  • the n-th region is the region which is furthest from the form region R 0 6.
  • regions of increase 7a, 7b, ..., 7n are defined connected regions 9, sub-sets of regions of increase 7a, 7b, ..., 7n, the regions constituted by voxels 2 which are connected to one another.
  • the first distance d m corresponding to a non-connected region 7m is determined, in which the voxels 2 are no longer all connected, but two or more distinct connected regions 9 are present which, when joined to one another, form the non-connected region 7m.
  • the regions of increase 7a, 7b, ..., 7n of equidistant voxels 2 corresponding to the outline 3 of the nodule 1 can be fragmented, i.e. they can consist of more than one connected region 9.
  • the fragmentation creates paths through the regions of increase 7a, 7b, ..., 7n, as indicated by the arrows A, B, C and D in figure 2.
  • Each connected region 9 belongs to at least one path A, B, C, D which connects the connected region 9 to the form region R 0 6, via other regions of voxels 2 which are connected closer to the said form region R 0 6.
  • the length L of the path A, B, C, D is defined as the number of regions of increase 7a, 7b, ..., 7n and of connected regions 9 which belong to the same path A, B, C, D.
  • a tree of the connected regions 50 is constructed, as represented in figure 3, in which each level represents each region of increase 7a, 7b, ..., 7n, and thus corresponds to a value of the distance of the voxels 2 belonging to each region of increase 7a, 7b, ... , 7n from the form region Ro 6.
  • the number of blocks 10 represents the number of regions of increase 7a, 7b, ... , 7n and of connected regions 9 belonging to this level.
  • a region of increase 7a, 7b, ..., 7n or a connected region 9 is considered to be a candidate for removal from the initial segmentation (together with all the regions of increase 7a, 7b, ..., 7n or with the connected regions 9 at a greater distance connected by means of a path A, B, C, D) if the decrease in the number of voxels 2 in relation to the region of increase 7a, 7b, ..., 7n or to the preceding connected region 9 to which it is connected, is greater than a certain predetermined value, which is preferably equal to 40%.
  • the second condition concerns the length of the residual path from the region of increase 7a, 7b, ..., 7n or from the connected region 9 in which the first condition has occurred, as far as the outline 3 of the nodule 1.
  • the mean dimension of the nodule 1 is defined as the sum of two factors.
  • the first factor is the dimension of the form region R 0 6. This is calculated by carrying the value of the quantity T over to the unit of measurement of the voxels 2 expressed in millimetres, i.e. by dividing the value of the quantity T by the length of one side of the voxels 2 of the image. If this value is less than one, the dimension of the form region R 0 6 is set as equal to one voxel, otherwise the dimension of the form region Ro 6 is set as equal to the numerical value derived from the above-described division.
  • the second factor is provided by the number of regions of increase 7a, 7b, ..., 7n contained between the form region R 0 6 and the first non-connected region 7m.
  • the value selected for the length of the residual path is 100% of the mean dimension.
  • the length of the residual path as far as the outline 3 of the nodule 1 is defined as the number of regions of increase 7a, 7b, ..., 7n or of connected regions 9 which connect the region of increase 7a, 7b, ..., 7n or the connected region 9 in which the first condition has occurred to the outline 3 of the nodule 1, including this region.
  • Figure 4 represents a graph of the development of the number of voxels 2 according to the distance from the form region R 0 6; as can be seen, this graph has an ascending leg which ends in a maximum point 101, and there is then a descending leg 102. On this descending leg 102 there is a point of intersection 103 where both the conditions occur.
  • Figure 5 represents a similar graph for another path, in which there is an ascending leg which ends in a maximum point 101, and then there is a descending leg 102. In this case, only the first condition has occurred and not the second, so no cut of the regions is applied.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

L'invention concerne un procédé de correction morphologique consistant à allouer à chaque voxel (2) une image d'une première valeur de distance géodésique égale à la distance du voxel (2) à un contour (3) d'un nodule (1), construire une première carte des distances géodésiques, démarrer à partir du contour (3), constituer des premiers ensembles (4) comprenant des voxels (2) qui ont la même première valeur de distance, et sélectionner un voxel 'd'ensemencement' (5) qui a la première valeur de distance maximale. Ultérieurement, une région de forme de voxels (6) est augmentée, et il y a allocation aux voxels externes (2) d'une seconde valeur de distance qui est égale à la distance des voxels (2) à un contour (8) de la région de forme de voxels (6). Ultérieurement, il y a construction d'une seconde carte des distances géodésiques comprenant de seconds ensembles (7), comprenant des voxels (2) qui ont une même seconde valeur de distance prédéterminée, et ceci signifie qu'une pluralité de champs de voxels d'augmentation (7a, 7b,..., 7n) sont définis. Il est ensuite vérifié si ou non lesdits champs de voxels d'augmentation (7a, 7b,..., 7n) sont connexes, et il y a détermination du nodule présumé (1) en tant qu'ensemble de champs de voxels d'augmentation (7a, 7b,..., 7n) connexes. Enfin, il y a correction de l'image segmentée par élimination des régions périphériques qui ne sont pas incluses dans le nodule présumé.
PCT/IT2007/000453 2007-06-26 2007-06-26 Procédé de correction morphologique d'une segmentation de nodules pulmonaires juxta-vasculaires dans des images de tomodensitométrie Ceased WO2009001378A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/IT2007/000453 WO2009001378A1 (fr) 2007-06-26 2007-06-26 Procédé de correction morphologique d'une segmentation de nodules pulmonaires juxta-vasculaires dans des images de tomodensitométrie

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/IT2007/000453 WO2009001378A1 (fr) 2007-06-26 2007-06-26 Procédé de correction morphologique d'une segmentation de nodules pulmonaires juxta-vasculaires dans des images de tomodensitométrie

Publications (1)

Publication Number Publication Date
WO2009001378A1 true WO2009001378A1 (fr) 2008-12-31

Family

ID=39092303

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IT2007/000453 Ceased WO2009001378A1 (fr) 2007-06-26 2007-06-26 Procédé de correction morphologique d'une segmentation de nodules pulmonaires juxta-vasculaires dans des images de tomodensitométrie

Country Status (1)

Country Link
WO (1) WO2009001378A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107563384A (zh) * 2017-08-31 2018-01-09 江苏大学 基于广义Hough聚类的粘连猪的头尾识别方法
CN109410181B (zh) * 2018-09-30 2020-08-28 神州数码医疗科技股份有限公司 一种心脏图像分割方法及装置

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040105527A1 (en) * 2002-11-22 2004-06-03 Matthieu Ferrant Methods and apparatus for the classification of nodules
US20040252870A1 (en) * 2000-04-11 2004-12-16 Reeves Anthony P. System and method for three-dimensional image rendering and analysis
US6993174B2 (en) * 2001-09-07 2006-01-31 Siemens Corporate Research, Inc Real time interactive segmentation of pulmonary nodules with control parameters
US20070086637A1 (en) * 2005-10-07 2007-04-19 Siemens Corporate Research Inc Distance Transform Based Vessel Detection for Nodule Segmentation and Analysis

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040252870A1 (en) * 2000-04-11 2004-12-16 Reeves Anthony P. System and method for three-dimensional image rendering and analysis
US6993174B2 (en) * 2001-09-07 2006-01-31 Siemens Corporate Research, Inc Real time interactive segmentation of pulmonary nodules with control parameters
US20040105527A1 (en) * 2002-11-22 2004-06-03 Matthieu Ferrant Methods and apparatus for the classification of nodules
US20070086637A1 (en) * 2005-10-07 2007-04-19 Siemens Corporate Research Inc Distance Transform Based Vessel Detection for Nodule Segmentation and Analysis

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
PICOZZI G: "Operator-Dependent Reproducibility of Size Measurements of Small Phantoms and Lung Nodules Examined With Low-Dose Thin-Section Computed Tomography", INVESTIGATIVE RADIOLOGY, PHILADELPHIA, PA, US, vol. 41, no. 11, November 2006 (2006-11-01), pages 831 - 839, XP009095279 *
SCHILHAM ET AL: "A computer-aided diagnosis system for detection of lung nodules in chest radiographs with an evaluation on a public database", MEDICAL IMAGE ANALYSIS, OXFORD UNIVERSITY PRESS, OXOFRD, GB, vol. 10, no. 2, April 2006 (2006-04-01), pages 247 - 258, XP005276425, ISSN: 1361-8415 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107563384A (zh) * 2017-08-31 2018-01-09 江苏大学 基于广义Hough聚类的粘连猪的头尾识别方法
CN109410181B (zh) * 2018-09-30 2020-08-28 神州数码医疗科技股份有限公司 一种心脏图像分割方法及装置

Similar Documents

Publication Publication Date Title
Sun et al. Automated 3-D segmentation of lungs with lung cancer in CT data using a novel robust active shape model approach
JP4717935B2 (ja) 画像処理装置および方法並びにプログラム
Lee et al. Automated detection of pulmonary nodules in helical CT images based on an improved template-matching technique
Diciotti et al. Automated segmentation refinement of small lung nodules in CT scans by local shape analysis
US8687864B2 (en) Image diagnostic processing device and image diagnostic processing program
US7756316B2 (en) Method and system for automatic lung segmentation
Naqi et al. Lung nodule detection using polygon approximation and hybrid features from CT images
US7643663B2 (en) Volume measurement in 3D datasets
Alilou et al. A comprehensive framework for automatic detection of pulmonary nodules in lung CT images
EP1883047A2 (fr) Détection de limite de nodule
RU2449371C2 (ru) Устойчивая к ошибкам функциональная визуализация
US8165376B2 (en) System and method for automatic detection of rib metastasis in computed tomography volume
WO2010100858A1 (fr) Dispositif et procédé de traitement d'image et programme
EP3679514B1 (fr) Détermination des régions de tissu pulmonaire hyperdense dans une image d'un poumon
WO2012112627A4 (fr) Procédé et appareil de détection par tomographie à faisceau conique assistée par ordinateur basée sur des images de seins et diagnostic associé
US20180247407A1 (en) Method and apparatus for segmenting medical images
US7480401B2 (en) Method for local surface smoothing with application to chest wall nodule segmentation in lung CT data
US6744911B1 (en) Tomographic segmentation
Khordehchi et al. Automatic lung nodule detection based on statistical region merging and support vector machines
KR20070083388A (ko) 결절 검출 방법
US7787673B2 (en) Method and apparatus for airway detection and segmentation using 3D morphological operators
WO2018176319A1 (fr) Procédé et dispositif d'analyse d'image ultrasonore
KR101474162B1 (ko) 흉부 컴퓨터 단층 영상의 간유리음영 결절 자동 분할 시스템 및 그 분할 방법
Maitra et al. Automated digital mammogram segmentation for detection of abnormal masses using binary homogeneity enhancement algorithm
WO2009001378A1 (fr) Procédé de correction morphologique d'une segmentation de nodules pulmonaires juxta-vasculaires dans des images de tomodensitométrie

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 07805665

Country of ref document: EP

Kind code of ref document: A1

DPE1 Request for preliminary examination filed after expiration of 19th month from priority date (pct application filed from 20040101)
DPE1 Request for preliminary examination filed after expiration of 19th month from priority date (pct application filed from 20040101)
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 07805665

Country of ref document: EP

Kind code of ref document: A1