Qualification method of lung vessel based on lobe

Abstract

Disclosed is a method for the quantification of pulmonary vessels by lobe, the method including extracting, at extraction unit, pulmonary vessels based on a medical image, locating, at analysis unit, voxels of pulmonary vessels with respect to the surface of a lobe, and quantifying, at calculation unit, the extracted pulmonary vessels.

Claims

1. A method for the quantification of pulmonary vessels by lobe, the method comprising: extracting, at extraction unit, pulmonary vessels based on a medical image; locating, at analysis unit, voxels of pulmonary vessels with respect to the surface of a lobe; and quantifying, at calculation unit, the extracted pulmonary vessels, wherein the locating, at the analysis unit, voxels of pulmonary vessels with respect to the surface of the lobe further comprises: forming at least one offset surface as a set of voxels at a predefined distance inwardly from the surface of the lobe; and after forming the at least one offset surface, locating voxels that correspond to intersections between the extracted pulmonary vessels and the at least one offset surface, and wherein the quantifying, at the calculation unit, the extracted pulmonary vessels further comprises: quantifying the extracted pulmonary vessels based on the intersections between the extracted pulmonary vessels and the at least one offset surface such that a distribution of pulmonary vessels is measured to correspond to an anatomical structure of pulmonary vessels.

2. The method according to claim 1, wherein locating, at analysis unit, voxels of pulmonary vessels with respect to the surface of a lobe further comprises: locating an outer lobe face; and locating an inner lobe face, with the outer lobe face and the inner lobe face comprising the surface of a lobe.

3. The method according to claim 2, wherein locating an inner lobe face involves locating a first inner face, a second inner face, a third inner face and a fourth inner face, with the first inner face being present between a first lobe on the upper part of right lung and a second lobe in the middle part of right lung, with the second inner face being present between the second lobe and a third lobe on the lower part of right lung, with the third inner face being present the third lobe and a fourth lobe on the upper part of left lung, with the fourth inner face being present between the fourth lobe and a fifth lobe on the lower part of left lung.

4. The method according to claim 1, wherein, in forming the at least one offset surface, the at least one offset surface is formed such that the at least one offset surface includes inner surface of the lobe which is provided inside the surface of the lobe.

5. The method according to claim 1, wherein the extraction unit extracts all blood vessels on the medical image.

6. The method according to claim 1, wherein, in extracting, at extraction unit, pulmonary vessels based on a medical image, the medical image is obtained by extracting voxels of blood vessels.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) For a more complete understanding of the present disclosure and the advantages thereof, reference is now made to the following description taken in conjunction with the accompanying drawings in which like reference numerals indicate like features and wherein:

(2) FIG. 1 and FIGS. 2A, 2B, 2C, 2D and 2E illustrate an exemplary method of the pulmonary vessel extraction for automatically detecting a lesion in a thoracic CT image, which is presented in registered Korean Patent Publication No. 10-2011-0129239;

(3) FIG. 3A and FIG. 3B illustrate a relationship between lungs and lobes;

(4) FIG. 4 is a flow chart describing a method for the quantification of pulmonary vessels by lobe;

(5) FIG. 5 shows an example of extracted vessels according to the present disclosure;

(6) FIG. 6A and FIG. 6B describe offset surfaces according to the present disclosure;

(7) FIG. 7A and FIG. 7B describe a procedure of locating the surface of a lobe, according to the present disclosure; and

(8) FIGS. 8A and 8B and FIGS. 9A and 9B illustrate benefits of locating pulmonary vessels with respect to the surface of a lobe, according to the present disclosure.

DETAILED DESCRIPTION

(9) The present disclosure will now be described in detail with reference to the accompanying drawing(s).

(10) FIG. 3(A) and FIG. 3(B) illustrate a relationship between lungs and lobes;

(11) In particular, FIG. 3(A) shows the entire lung 100, and FIG. 3(B) shows bronchi 151, 153, and the entire lung 100.

(12) The lung 100 has a right lung 110 and a left lung 130, and a trachea 150 between the right lung 110 and the left lung 130. During inhalation, air flows into the lungs through the trachea 150; during exhalation, air flows out of the lungs. For the lung 100, the right lung 110 has three lobes (lung lobes) 111, 113, 115, and the left lung 130 has two lobes 131, 133.

(13) Although the right lung 110 and the left lung 130 of the whole lung 100 seem to be a single mass, respectively, they are divided into five lobes 111, 113, 115, 131, 133 in total. The lobes are divided by bronchus 151, 153, and pulmonary vessels (not shown) are located around the bronchi 151 and 153. Therefore, in practice, it is accurate to extract pulmonary vessels with respect to the corresponding lobe 111, 113, 115, 131, 133, while the lobes 111, 113, 115, 131, 133 are divided by bronchus 151, 153. In addition, the lung 100 can be segmented into a greater number of lobes according to the division of pulmonary vessels or bronchi 151 and 153.

(14) Referring to FIG. 3(B), bronchi 151-1, 151-3, 151-5, 153-1, 153-3 are also distributed to the respective lobes. In other words, the lobes can be divided by bronchus 151-1, 151-3, 151-5, 153-1, 153-3 or pulmonary vessels surrounding the corresponding bronchus 151-1, 151-3, 151-5, 153-1, 153-3.

(15) FIG. 4 is a flow chart describing a method for the quantification of pulmonary vessels by lobe.

(16) In the method for the quantification of pulmonary vessels by lobe, first of all, extraction unit extracts vessels based on a medical image (51). The vessels are extracted as a 3D set of voxels based on the medical image. On the medical image, the extraction unit can locate voxels of the vessels according to a HU (Hounsfield Unit) value thereof. For instance, the extraction unit can find voxels having at least −750 HU in order to find a target vessel. This is discussed below in reference to FIG. 5.

(17) Next, analysis unit locates voxels of pulmonary vessels with respect to the surface of a lobe (S2).

(18) The surface of a lobe has an outer lobe face and an inner lobe face. The outer lobe face corresponds to an exterior surface of the lung, and the inner lobe face refers to an interface formed between lobes.

(19) Moreover, the quantification method involves locating an outer lobe face as well as locating an inner face between lobes. Such an inner face between lobes can be predicted with respect to a bronchus or pulmonary vessel on the medical image. The medical image around this predicted inner face is then used for locating an inner face between lobes. More details on this will be provided below in reference to FIG. 7.

(20) Analysis unit forms an offset surface with respect to the surface of a lobe. The term ‘offset surface’ is defined as a set of voxels at a predefined distance inwardly from the surface of a lobe. Further, voxels of pulmonary vessels corresponding to the intersection between the offset surface and the extracted vessel are found. More details on this will be provided below in reference to FIG. 6.

(21) Next, calculation unit quantifies pulmonary vessels that were extracted (S3). In particular, calculation unit calculates a radial distance or diameter of pulmonary vessels using their voxels found. Moreover, calculation unit calculates an area ratio of pulmonary vessels with their radial distance thus obtained. Calculation unit may also calculate volume of the pulmonary vessels using their voxels.

(22) FIG. 5 shows an example of extracted vessels according to the present disclosure.

(23) Vessel extraction is done on the medical image. In particular, all vessels shown on the medical image are extracted and put together as data, thereby constructing 3D vessels. In case of extracting blood vessels from a thoracic medical image, cardiac vessels as well as pulmonary vessels can be extracted. In FIG. 5, only blood vessels having a predefined diameter or greater are shown, and any blood vessels having a diameter shorter than the predefined diameter are not shown even if they are extracted.

(24) FIG. 6(A) and FIG. 6(B) describe offset surfaces according to the present disclosure.

(25) An exemplary method of generating a Euclidean Distance field to form an offset surface will now be described.

(26) Based on a fact that pulmonary branch vessels are stretched out to an end region from the inside of the body, an assumption can be made that there would be similar sized vessels at the same distance from outer end boundary surfaces of lobes. Therefore, it is necessary to locate intersections between vessels excluding those in the mediastinal region and inner surfaces IA, IB, IC, ID, IE (hereinafter “I” (inner surface)) of the lung at a predefined distance from the outer end boundary surfaces of lobes, and to assess diameters of pulmonary vessels in those intersections.

(27) In order to obtain such intersections, first of all, inner surfaces I should be gradually extracted. The inner surfaces I at a predefined distance from the outer end boundary surfaces of lobes become offset surfaces OA, OB, OC, OD, OE (hereinafter “O” (offset surfaces)) at that distance. Similar to face-based offset or vertex-based offset calculations, these offset surfaces O can be formed with surface data calculation schemes that are more efficient (time-wise) than volume-based calculation schemes.

(28) However, the aforementioned surface data calculation schemes do not properly work in the presence of local and overall interferences that often occur when the surfaces of lobes are being offset inwardly. Particularly in this exemplary embodiment, offset distances range from 5 mm to 30 mm, which are much larger than lengths of a surface extracted by marching cubes from CT images, making it more difficult to avoid the interferences. Hence, in this exemplary embodiment, offset surfaces O are formed with volume-based schemes which generate a Euclidean Distance field (see FIG. 6(A)).

(29) For instance, referring to FIG. 6(A), the lung is divided into lobes 111, 113, 115, 131, 133. The lobes 111, 113, 115, 131, 133 form at least one offset surface O inwardly from the surfaces of the lobes. Among others, an offset surface O formed inwardly of a lobe is called an inner surface I.

(30) FIG. 6(B) illustrates offset surfaces O formed on the lobes 111, 113, 115, 131, 133, respectively.

(31) After obtaining these inner surfaces I, i.e. offset surfaces O at a predefined distance, intersections between the offset surfaces O and small vessels are located. More details on this will be provided below in reference to FIG. 8 to FIG. 9.

(32) FIG. 7(A) and FIG. 7(B) describe a procedure of locating the surface of a lobe, according to the present disclosure.

(33) The surface 170 of a lobe has an outer lobe face 171 and an inner lobe face 173. For instance, the lung 100 includes a right lung 110 on the right hand side, and a left lung 130 on the left hand side. The right lung 110 has a first lobe 111 (see FIG. 3) on the upper part, a second lobe 113 (see FIG. 3) in the middle part, and a third lobe 115 (see FIG. 3) on the lower part. The left lung 130 has a fourth lobe 131 (see FIG. 3) on the upper part and a fifth lobe 133 (see FIG. 3) on the lower part. A first inner face 173-1 is present between the first lobe 111 and the second lobe 113. A second inner face 173-2 is present between the second lobe 113 and the third lobe 115. A third inner face 173-3 is present between the first lobe 111 and the third lobe 117. A fourth inner face 173-4 is present between the fourth lobe 131 and the fifth lobe 133.

(34) First of all, the outer lobe face 171 is located. The outer lobe face 171 corresponds to the surface of the lung. Therefore, the surface of the lung is located on the medical image. For instance, once the pulmonary area is determined as in FIG. 2, it is separated from the medical image. In this sense, FIG. 7(A) may be referred to as the surface of the lung or as the outer lobe face 171.

(35) Next, the inner lobe face 173 is located. As can be seen in FIG. 7(B), a fissure that forms an interface (hereinafter, inner face 173) between lobes on the medical image is very thin and therefore, very difficult to distinguish. For instance, ranges or scopes S are set with respect to the extracted pulmonary vessels or bronchi (151-1, 151-3, 151-5, 153-1, 153-3), and inner face 173 between the scope S is located on the medical image. Locating the inner face 173 can be done by artificial intelligence. That is, with the lobes being divided by the extracted pulmonary vessels or bronchi (151-1, 151-3, 151-5, 153-1, 153-3), it is possible to predict where the inner face 173 is possibly going to be located, based on thickness and length of the pulmonary vessels or bronchi (151-1, 151-3, 151-5, 153-1, 153-3). In this way, the scope is defined, and the inner face 173 is found.

(36) FIG. 8(A) to (B) and FIG. 9(A) to (B) illustrate benefits of locating pulmonary vessels with respect to the surface of a lobe, according to the present disclosure.

(37) In particular, FIG. 8(A) and FIG. 8(B) each show one inner surface out of many inner surfaces with respect to the surface of a lobe, and FIG. 9(A) and FIG. 9(B) each show one inner surface out of many inner surfaces with respect to the surface of the lung.

(38) The inner surfaces shown in FIG. 8(A) and FIG. 8(B) are formed with respect to the surface of a lobe as mentioned above, and pulmonary vessels are extracted at the intersections between these inner surfaces and the pulmonary vessels.

(39) Similarly, the inner surfaces shown in FIG. 9(A) and FIG. 9(B) are formed with respect to the surface of the lung as mentioned above, and pulmonary vessels are extracted at the intersections between these inner surfaces and the pulmonary vessels.

(40) Because the inner surfaces IA, IB, IC, ID, IE in FIG. 8(A) and FIG. 8(B) are formed with respect to the surface of a lobe, the distribution of pulmonary vessels V can be measured to agree with anatomical structure of real blood vessels. Meanwhile, because the inner surfaces IA, IB, IC, ID, IE in FIG. 9(A) and FIG. 9(B) are formed with respect to the surface of the lung, the distribution of pulmonary vessels V can be measured regardless of anatomical structure of real blood vessels. In other words, with the inner surfaces IA, IB, IC, ID, IE (with respect to the surface of the lung) in FIG. 9(A) and FIG. 9(B), other vessel-free areas are also included in the analysis of the distribution of pulmonary vessels, negatively affecting reliability of the analysis results.

(41) Set out below are a series of clauses that disclose features of further aspects of the invention, which may be claims.

(42) (1) A method for the quantification of pulmonary vessels by lobe, the method including: extracting, at extraction unit, pulmonary vessels based on a medical image; locating, at analysis unit, voxels of pulmonary vessels with respect to the surface of a lobe; and quantifying, at calculation unit, the extracted pulmonary vessels.

(43) (2) There is also provided, the method of clause (1) wherein: locating, at analysis unit, voxels of pulmonary vessels with respect to the surface of a lobe includes forming offset surfaces at a predefined distance from the surface of a lobe.

(44) (3) There is also provided, the method of clause (2) wherein: locating, at analysis unit, voxels of pulmonary vessels with respect to the surface of a lobe further includes, after forming offset surfaces at a predefined distance from the surface of a lobe, locating voxels that correspond to intersections between the extracted pulmonary vessels and the offset surfaces.

(45) (4) There is also provided, the method of clause (1) wherein: locating, at analysis unit, voxels of pulmonary vessels with respect to the surface of a lobe includes locating an outer lobe face and locating an inner lobe face, with the outer lobe face and the inner lobe face comprising the surface of a lobe.

(46) (5) There is also provided, the method of clause (4) wherein: locating an inner lobe face involves locating a first inner face, a second inner face, a third inner face and a fourth inner face, with the first inner face being present between a first lobe on the upper part of the right lung and a second lobe in the middle part of the right lung, with the second inner face being present between the second lobe and a third lobe on the lower part of the right lung, with the third inner face being present the third lobe and a fourth lobe on the upper part of the left lung, with the fourth inner face being present between the fourth lobe and a fifth lobe on the lower part of the left lung.

(47) (6) There is also provided, the method of clause (2) wherein: in forming offset surfaces at a predefined distance from the surface of a lobe, the offset surfaces are formed inwardly from the surface of a lobe, and at least one offset surface is formed, with the offset surfaces including inner surfaces provided inside.

(48) (7) There is also provided, the method of clause (1) wherein: the extraction unit extracts all blood vessels on the medical image.

(49) (8) There is also provided, the method of clause (1) wherein: in extracting, at extraction unit, pulmonary vessels based on a medical image, the medical image is obtained by extracting voxels of blood vessels.

(50) In the method for the quantification of pulmonary vessels by lobe according to an exemplary embodiment of the present disclosure, offset surfaces are formed with respect to a lobe, allowing more accurate extraction of pulmonary vessels.

(51) In the method for the quantification of pulmonary vessels by lobe according to another exemplary embodiment of the present disclosure, the surface of a lobe is divided such that a greater number of pulmonary vessels can be located.

DRAWING REFERENCE NUMERALS

(52) S110: Gamma correction S120: 3D Otsu's binarization S130: Morphology S140: Determining pulmonary area S210: Determining of pulmonary vascular region S220: Thinning a vessel candidate S230: Post-processing thinned results S240: Smoothing thinned results S310: Determining first topology point S320: Constructing vascular topology S330: Post-processing vascular topology S340: Re-constructing vascular topology S410: Generating unit skeleton reference points S420: Selecting 3D model for analysis of thickness S430: Analyzing thickness of blood vessels S440: Re-classifying nodule candidates