CN101989351A - Suspected lung nodule image enhancement directional scale filtering method - Google Patents
Suspected lung nodule image enhancement directional scale filtering method Download PDFInfo
- Publication number
- CN101989351A CN101989351A CN 201010545654 CN201010545654A CN101989351A CN 101989351 A CN101989351 A CN 101989351A CN 201010545654 CN201010545654 CN 201010545654 CN 201010545654 A CN201010545654 A CN 201010545654A CN 101989351 A CN101989351 A CN 101989351A
- Authority
- CN
- China
- Prior art keywords
- image
- lung
- picture
- rabat
- original
- 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.)
- Pending
Links
Images
Landscapes
- Apparatus For Radiation Diagnosis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a suspected lung nodule image enhancement directional scale filtering method, which comprises the following steps: inputting an original image, and calculating a peak value image and a valley value image of the original X-ray chest radiography respectively by utilizing a top-hat transform operator and a bottom-hat transform operator in gray morphology; carrying out directional enhancement on a lung similar-round focus image in the X-ray chest peak value image by utilizing a directional scale Laplace Gauss function as a matched filter, introducing a visual correction factor in image enhancement operation, and carrying out visual correction on a convolution image of a directional scale Laplace Gauss operator; and adding the original X-ray chest image in the image obtained from the step 2, and then subtracting the valley value image obtained from the step 1 to obtain an output image. The beneficiary effect of the invention is that the grey contrast between the lung similar-round structural focus in the X-ray chest image and the periphery background tissue of the lung similar-round structural focus is enhanced obviously, the texture details in the X-ray chest are maintained, and the visual effect of the X-ray chest is improved preferably.
Description
Technical field
The invention belongs to the automatic analyzing and processing technical field of medical image, be specifically related to a kind of based on the doubtful tubercle figure of the improved lung of mathematical morphology image intensifying direction scale filter method.
Background technology
The mortality ratio of lung cancer occupies due to world's malignant tumour first of the death toll, and patient's 5 years survival rates and neoplasm staging are closely related, and 5 years total survival rates of patients with lung cancer are about about 15%, yet 5 years survival rates of I phase patient can reach 70%
[1], therefore detect and diagnose in early days survival rate to have great importance in early days to improving patients with lung cancer.
Lung cancer is with the performance of the form of pleurotome point in early days.(Helical Computed Tomography Scans is two kinds of major technique means that lung's node medical image detects with X ray rabat (Chest Radiography) HCTS) to Spiral CT scan
[2]In lung's node medical image early detection, though the detection performance of CT scan technology obviously is better than X ray rabat technology, but because the ability of X ray rabat aspect some illnesss not under a cloud of detection, particularly compared to CT scan, x-ray dose that it is lower and comparatively cheap financial cost make the X ray rabat become the most preferential and the most common technological means in the node medical image detection of present lung
[3]Yet it may be noted that, because the X ray chest film picture similarly is a kind of radiophotography image, various tissue images are stack mutually in X rabat image, add that each breast tissue is also bigger to the absorptivity difference of X-ray, for example at the bone of the existing strong absorption X-ray in lung district, blood vessel, two lobes of the lung that seldom absorb X-ray etc. are also arranged, x-ray image is difficult to all correct exposures of each several part to chest, therefore cause the X rabat to have lower image contrast and more noise and relatively poor video vision, smaller again usually owing to lung's node image simultaneously, thus make the clinical detection of X rabat lung node concerning the doctor, remain a challenging job
[4]
As a kind of basic image preprocessing means, the figure image intensifying is for improving X rabat image quality in images and visual effect, and reducing the difficulty of reading sheet has vital role.In recent years, people have done a large amount of work in this regard, propose a lot of effectively image enchancing methods, and are applied in the X rabat image enhancement processing
[5]-[12], as X rabat image enchancing method based on histogram equalization
[8]-[9], based on the figure image intensifying of Laplce's template
[10]-[11]And based on the image enchancing method of rough set
[12]Deng.Yet said method all is to strengthen at view picture X rabat, do not consider especially certain specific lesion region or pathological tissues in the X rabat, strengthen as lung's node, and mainly be detection for the detection of lung cancer clinically to lung's node, so how research strengthens and improve X rabat lung node image quality, the grey-scale contrast that particularly strengthens lung's node and its surrounding tissue image has even more important meaning for the early detection result who improves lung cancer.
In document [16], based on observation analysis to X rabat and X rabat lung node imaging characteristic, proposed a kind of based on vision and multiple dimensioned LoG operator X rabat lung node image Enhancement Method, its method is: by at X, the Y direction is got different scale-value respectively and is come traditional Laplce's Gauss operator is improved, make X, the difference that Y direction upper frequency passband becomes, thus consistent with the visually-perceptible of human eye and meet human-eye visual characteristic.Experiment shows: this method can effectively strengthen the lung tubercle, and non-lung tubercle is weakened, and the image after the enhancing relatively meets the physiological property of X rabat own.But this method strengthens DeGrain for the contrast of trickle focus structure and its surrounding structure in the X rabat.Analyzing its reason is: because X rabat image resolution ratio is low excessively, noise is too serious, and for some trickle institutional framework and focus, its structure and shape are difficult to differentiate and cause.
Mathematical morphology is that a kind of morphological feature with image is a research object, to the mathematical tool that image is analyzed, its basic thought is to go to measure and extract corresponding form in the image to reach the purpose to graphical analysis and identification with the structural element with certain form
[17]-[18]The most basic morphology operations has: burn into expands, open and close.Wherein, the effect of corrosion is to eliminate frontier point, thereby dwindles target area, increase hole, elimination noise spot; The effect of expanding is opposite with corrosion, background dot can be merged in the target area, increases the target area, eliminates hole, forms connected region; Opening operation is the merging of corrosion after expansion computing earlier, establishes
ABe input picture
, BBe structural element
,Image then
BTo image
ACan be described as opening operation:
, opening operation smoothly larger object, place to go wisp, separate object at the fine rule place.Closed operation is the dual operations of opening operation, is the merging of first expansion post-etching computing, establishes
ABe input picture
, BBe structural element
,Image then
BTo image
ACan be described as do closed operation:
, closed operation can be removed the aperture on the object, the fusion gap.
From a width of cloth original image, deduct it is done the image that obtains behind the opening operation, can obtain some important information, as high curvature point, this method is to asking dark pixel aggregation in brighter background, or asks bright pixel aggregation very effective in darker background.Ash value morphology Top-Hat (high cap) can satisfy the requirement of above-mentioned two aspects respectively with Bottom-Hat (low cap) conversion, can be used for gray scale peak value and valley littler than structural element in the detected image signal.
If
ABe input picture
, BBe structural element
,Image then
BTo image
ACan be described as cap transformation
, image
BTo image
ADoing low cap conversion can be described as
By the introduction of above gray scale morphology fundamental operation as can be seen, image is carried out the shadow information that the cap transformation computing can keep the gray scale peak value enhancing image of target, and low cap conversion can be obtained the valley in the image, boundary between the outstanding interconnective target, so being used in combination of high and low cap conversion, display foreground and background gray scale further are stretched, highlight some targets and the details darker, play the effect that target strengthens than background.
Summary of the invention
The purpose of this invention is to provide the doubtful tubercle figure of a kind of lung image intensifying direction scale filter method, solved prior art and be difficult to problem that doubtful lung node image in the X rabat is strengthened.
The technical solution adopted in the present invention is that the doubtful tubercle figure of a kind of lung image intensifying direction scale filter method comprises following operation steps:
Step 1,
The input original image utilizes the high cap in the gray scale morphology, low cap transformation operator to calculate the peak image and the valley image of original X rabat respectively;
Step 2,
Utilizing direction yardstick Laplce Gaussian function as matched filter similar round lesion image in the lung in the X rabat peak image to be carried out orientation strengthens, in figure image intensifying computing, introduce the vision correcting factor, direction yardstick Laplce Gauss operator convolved image is carried out vision correcting;
Step 3,
The image that step 2 is obtained adds that original X chest film picture picture deducts the valley image that step 1 obtains again, can obtain output image.
Wherein, in the step 1, calculate the peak image of original X rabat and the method for valley image and be:
Wherein,
Wherein,
The expression input picture,
Represent selected structural element,
Represent the peak image corresponding with input picture,
Represent the valley image corresponding, the gray-scale value of x and y difference remarked pixel point horizontal direction and vertical direction with input picture; Described structural element is that radius is formed multiple dimensioned disc structure at interval [5 15] a plurality of disc structure elements.
Wherein, in the step 2, utilize direction yardstick Laplce Gaussian function similar round lesion image in the lung in the X rabat peak image to be carried out orientation and strengthen, be shown below as matched filter:
In the formula,
Expression utilizes direction yardstick Laplce Gaussian function that similar round lesion image in the lung in the X rabat peak image is carried out directed enhanced results image, the gray-scale value of x and y difference remarked pixel point horizontal direction and vertical direction;
Described direction yardstick Laplce Gaussian function is:
In the formula,
Represent that respectively Laplce's Gaussian function is in x and y direction scale factor, be used to control the filter scale of Laplce's Gaussian function in x and y direction, n is the size of Laplce's Gaussian function convolution kernel, the gray-scale value of x, y difference remarked pixel point horizontal direction and vertical direction;
The method of introducing the vision correcting factor in figure image intensifying computing is:
In the formula,
Expression vision correcting image, β is the vision correcting factor, and the β value is in [1.5 2.5] scope, and the β value preferentially gets 2.
Wherein, in the step 3, the image that step 2 is obtained adds that the method that original X chest film picture picture deducts the valley image that step 1 obtains again is:
In the formula,
Expression strengthens output image as a result.
The invention has the beneficial effects as follows, not only make in the X chest film picture picture in the lung class circle structure focus with its on every side the intensity contrast of background tissues obviously strengthen, and kept grain details in the X rabat, improved the visual effect of X chest film picture picture preferably.More helping the doctor detects accurately to lung's node in the X rabat.
Description of drawings
Fig. 1 is the darker circle in brighter outside, inside that pleurotome point showed on the X ray rabat;
Fig. 2 is the three-dimensional probability distribution map of Laplce's Gaussian function;
Fig. 3 is the three-dimensional probability distribution reflection of a Laplce's Gaussian function cross-sectional view;
Fig. 4 is the process flow diagram of the doubtful tubercle image enchancing method of lung of the present invention;
Fig. 5 utilizes document [15] filtering method to X rabat lung doubtful tubercle image enhanced results, wherein, figure a1, a2, a3 are original X chest film picture picture, and scheming b1 is at the image of scheming after a1 strengthens, and b2 is that image, the b3 after strengthening at figure a2 is the image after strengthening at figure a3;
Fig. 6 utilizes document [16] method to X rabat lung doubtful tubercle image enhanced results, wherein, figure c1, c2, c3 are original X chest film picture picture, and figure d1 is that image, the d2 after strengthening at figure c1 is the image after strengthening at figure c3 for the image after strengthening at figure c2, d3;
Fig. 7 is the inventive method looks like to carry out gained in the processing procedure to original X chest film picture a image, wherein, figure e1 is original X chest film picture picture, figure f is corresponding peak image, figure g is corresponding valley image, figure h utilizes direction yardstick Laplce Gaussian function as matched filter similar round lesion image in the lung in the X rabat peak image to be carried out the directed image that strengthens, and e2 is an output image;
Fig. 8 utilizes the X chest film picture picture that the inventive method strengthened and the comparison diagram of original X chest film picture picture, and wherein, i1 is original X chest film picture picture, and j1 is the X chest film picture picture that utilizes after the inventive method is handled as i1 original X chest film picture; I2 is original X chest film picture picture, and j2 is the X chest film picture picture that utilizes after the inventive method is handled as i2 original X chest film picture; I3 is original X chest film picture picture, and j3 is the X chest film picture picture that utilizes after the inventive method is handled as i3 original X chest film picture;
Fig. 9 is original X chest film picture picture, the image after the described method of document [16] strengthens, the image after the Top-Bottom-hat method strengthens and the comparison diagram between the image after the inventive method strengthens; Wherein, l1 is original X chest film picture picture, m1 is the image after the described method of document [16] is handled original X chest film picture l1, and n1 is the image after the Top-Bottom-hat method is handled original X chest film picture l1, and o1 is the image after the inventive method is handled original X chest film picture l1; L2 is original X chest film picture picture, and m2 is the image after the described method of document [16] is handled original X chest film picture l2, and n2 is the image after the Top-Bottom-hat method is handled original X chest film picture l2, and o2 is the image after the inventive method is handled original X chest film picture l2; L3 is original X chest film picture picture, and m3 is the image after the described method of document [16] is handled original X chest film picture l3, and n3 is the image after the Top-Bottom-hat method is handled original X chest film picture l3, and o3 is the image after the inventive method is handled original X chest film picture l3.
Embodiment
The present invention is described in detail below in conjunction with the drawings and specific embodiments.
Clinical observation
[13]Show, typical lung node is rendered as a spheroid in shape usually approx, its density and water are worked as, and are a bit larger tham its lung liquid on every side, so pleurotome point generally shows as the different big or small darker circles (as shown in Figure 1) in brighter outside, inside on the X ray rabat
Laplce's Gaussian function can be described as with mathematical formulae:
In the formula,
Be the standard deviation of Gaussian function, be used to control the yardstick of filtering; The gray-scale value of x and y difference remarked pixel point horizontal direction and vertical direction; E represents the end of natural logarithm; π is a circular constant.This function is a class circle distribution function in the same way
[14]When
Value is 1 o'clock, and the three-dimensional probability distribution of Laplce's Gaussian function is videoed as shown in Figure 2, and three-dimensional probability distribution reflection sectional view as shown in Figure 3.Laplce's Gaussian function has a high central cynapse (as shown in Figure 2) that can excite high grey value profile to shift as can be seen, it makes the three-dimensional probability distribution reflection of Laplce's Gaussian function sectional view present a border circular areas with two obvious light and shade contrast differences, i.e. Ming Liang middle section (high gray-scale value) and relative darker peripheral region (as shown in Figure 3).Lung's node pixel probability distribution is very alike in probability distribution reflection characteristics that Laplce's Gaussian function is such and the X rabat, and promptly middle section is bright, the peripheral region is darker relatively.Therefore Laplce's Gaussian function and X chest film picture are looked like to carry out convolution, Laplce's Gaussian function can be as a probe, realizes light and shade contrast difference inside and outside the detection of similar round structural images in the X rabat and the calculating region-of-interest.
In document [15], the method of utilizing the LoG filter operator to strengthen X rabat lung node image is inquired into and is studied, experimental result (as shown in Figure 5) shows that the method that proposes in the document [15] is feasible, but utilize this method when the lung's node to the X rabat strengthens, cause that easily the part grain details of lung's image is weakened.At this problem, in document [16], a kind of improving one's methods proposed, promptly to the LoG operator at X, the Y direction is got different standard deviation values respectively and is made X, the difference that Y direction upper frequency passband becomes, thereby can strengthen at X rabat lung node image different directions texture, this improved LoG operator can be described with following mathematic(al) representation:
In the formula,
Represent that respectively Laplce's Gaussian function is in x and y direction scale factor, be used to control the filter scale of Laplce's Gaussian function in x and y direction, n is the size of Laplce's Gaussian function convolution kernel, x, y are consistent with implication in the LoG expression formula, respectively the gray-scale value of remarked pixel point horizontal direction and vertical direction.According to document [21], standard deviation
Must be similar to respectively
, wherein
When
＞
The time, transverse on the x axle, minor axis on the y axle, this operator convolved image, it strengthens effect to image level and is far longer than vertical enhancing effect; When
＜
The time, this operator strengthens effect to vertical direction and is higher than horizontal effect far away; When
=
The time, ellipse becomes standard round, and it is identical that strengthen effect to level and vertical direction this moment, at this moment just become traditional Laplce's Gaussian function.
Though document [16] method has certain improvement to the texture information of X rabat lung node image, but the contrast for trickle focus structure and its surrounding structure in the X rabat strengthens effect and not obvious (as shown in Figure 6), its reason is because X rabat image resolution ratio is low excessively, noise is too serious, for some trickle institutional framework and focus, its structure and shape are difficult to differentiate and cause.
In order to realize further effectively strengthening to X rabat lung's node and its surrounding tissue gradation of image contrast, reduce the difficulty of radiation doctor reading X rabat, improve the accuracy that doctor and the node inspection of computer aided system lung are examined, the invention provides a kind of X rabat lung similar round lesion image Enhancement Method---promptly in conjunction with the X rabat lung similar round lesion image Enhancement Method of height cap conversion of gray scale morphology and direction yardstick Laplce Gaussian filter, as Fig. 4 and shown in Figure 7, comprise following operation steps:
Step 1, input original image (e1 among Fig. 7) utilizes high cap in the gray scale morphology, low cap transformation operator to calculate the peak image and the valley image of original X rabat respectively, promptly
Wherein,
Here
The expression input picture,
Represent selected structural element,
Represent the peak image corresponding with input picture,
Represent the valley image corresponding, the gray-scale value of x and y difference remarked pixel point horizontal direction and vertical direction with input picture;
Utilize the Top-Hat operator to calculate X rabat peak image, can from darker background, obtain brighter pixel aggregation (being the doubtful image of lung's node).Carry out in the Top-Hat conversion process at image, the selection of structural element is extremely important, and the information of wanting usually will strengthen in the combining image is determined its shape and size.In the X rabat, the pleurotome point typically refers to diameter less than similar round focus in 3 centimetres the lung.Therefore, the structural element that the present invention chooses is a disc structure, when considering to use Top-Hat conversion extraction X rabat peak image in addition, the radius of disc structure can not be less than the image radius of similar round focus in the lung, can guarantee not lose the information of similar round focus in the lung, so the present invention selects for use radius to form multiple dimensioned disc structure at interval [5 15] a plurality of disc structure elements, realizes the calculating to X rabat peak image and valley image, shown in Fig. 7 f and Fig. 7 g;
Step 2 is utilized direction yardstick Laplce Gaussian function as matched filter similar round lesion image in the lung in the X rabat peak image to be carried out orientation and is strengthened, and is shown below:
In the formula,
Expression utilizes direction yardstick Laplce Gaussian function that similar round lesion image in the lung in the X rabat peak image is carried out directed enhanced results image, x, and the implication of y is with mentioned above consistent;
After the X chest film picture picture process Top-Hat conversion process, though background information is inhibited, lung's information is enhanced, and the contrast of similar round focus and background is still lower in the lung, adopts matched filter that similar round focus in the lung is surveyed enhancing for this reason.The essence of matched filtering be to design one with image in region of interest or the similar wave filter of target object, can obtain to wish the image information that obtains through matched filtering.Based on above-mentioned analysis, and the directivity of consideration lung images texture, the present invention adopts the direction yardstick Laplce Gaussian function of document [16] proposition as matched filter, realizes the orientation of similar round lesion image in the lung in the X rabat peak image is strengthened.Among the present invention, according to experiment, the nuclear size of LoG filter operator is set at 15 * 15, and the scale size of horizontal direction and vertical direction value respectively is
=5 Hes
=6.Fig. 7 h has shown the output result who utilizes after this improvement LoG operator strengthens input picture;
Because the smoothing effect of Gaussian function in the LoG operator, after adopting the LoG operator that the X rabat is carried out enhancement process, X rabat part detailed information can be weakened, thereby cause filtered image to seem point fuzziness is arranged a little.At the problems referred to above, consider human-eye visual characteristic, the present invention introduces a vision correcting factor-beta in above-mentioned improvement operator, promptly
In the formula,
Expression vision correcting image.
Here an important human-eye visual characteristic of our foundation is that human eye is to more responsive at the noise of detail section in the noise ratio of the mild part of image.Netravali and Prasada
[19]The observability that proposes noise increases along with the spatial variations rate of pixel and reduces, and according to above principle, can make the local contrast of image increase more greatly at detail section, and is smaller in mild part.Introduce this factor and act on the X chest film picture picture that improved Laplce's Gauss operator convolution obtains, can make X rabat and lung's node after the enhancing have visual effect preferably.According to experiment and in conjunction with document [19], it is 2 that the present invention chooses the β value;
Step 3 in order to realize the overall enhanced to X chest film picture picture, improves original X chest film picture as the contrast between similar round focus in the lung and its background, and the image that step 2 is obtained adds that original X chest film picture picture deducts the valley image that step 1 obtains again, promptly
In the formula,
Expression strengthens output image as a result, shown in figure e2.
Based on JRST medical science X rabat image data base [20], we verify the validity of the inventive method:
JRST medical science X rabat image data base is the medical image databases by the open issue of Japanese radiation technique association exploitation.In experiment, 29 normal and 23 improper postero-anterior position X rabats that include 143 false nodes and 33 nodes from this database, have been chosen.Having or not by CT examination of node confirmed in the rabat.All experiments all are at a PVI 800, the 1.8MB internal memory, and operating system is on the PC machine of Windows XP, finishes with the matlab7.0 programming.
Fig. 8 has provided three experimental results of checking the inventive method validity on JRST medical science X rabat image data base.As seen from Figure 8, in original X chest film picture picture (figure i1, i2, i3), the contrast in the lung between similar round lesion image and its background structure is very low, almost is submerged in its background the non-constant of identifiability; And through in the X chest film picture picture after the inventive method enhancing (figure j1, j2, j3), contrast in the lung between similar round lesion image and its background structure is significantly improved, and can be easy to observe similar round lesion image in the brilliant white lung the X chest film picture picture after strengthening.Illustrate that the inventive method can effectively strengthen the contrast between interior similar round lesion image of X rabat lung and the structures surrounding thereof.
In order further to verify the validity of the inventive method, also contrast, chosen three groups of results as shown in Figure 9 with the inventive method and based on the X rabat lung node image enhancement algorithms of tradition height cap conversion and the direction yardstick LoG operator X rabat algorithm for image enhancement of the middle proposition of document [16]:
First group: l1 is original X chest film picture picture, m1 is the image after the described method of document [16] is handled original X chest film picture l1, n1 is the image after the Top-Bottom-hat method is handled original X chest film picture l1, and o1 is the image after the inventive method is handled original X chest film picture l1;
Second group: l2 is original X chest film picture picture, m2 is the image after the described method of document [16] is handled original X chest film picture l2, n2 is the image after the Top-Bottom-hat method is handled original X chest film picture l2, and o2 is the image after the inventive method is handled original X chest film picture l2;
The 3rd group: l3 is original X chest film picture picture, m3 is the image after the described method of document [16] is handled original X chest film picture l3, n3 is the image after the Top-Bottom-hat method is handled original X chest film picture l3, and o3 is the image after the inventive method is handled original X chest film picture l3;
As seen from Figure 9, at original input picture (figure l1, l2, l3), class circle structure almost is submerged in the X rabat background image in the lung, difficultly observes clearly class circle structural images in the lung very soon in X rabat lung background image.And, can in X rabat lung images, observe clearly class circle structure (brilliant white class circle structure) in the lung easily at the X chest film picture picture that strengthens through the inventive method (figure o1, o2, o3) with through X chest film picture picture (figure m1, m2, m3) that document [16] method strengthens and in the X chest film picture picture (figure n1, n2, n3) of height cap conversion process.Further relatively, at the X chest film picture picture that strengthens through document [16] method (figure m1, m2, m3) in, not only can observe class circle structural images clearly, also can observe non-clearly class circle structural images, as rib etc., class circle structure and the mutual juxtaposition of non-class circle structure, but (scheme m1 through the X chest film picture picture that document [16] method strengthens, m2, m3) detail textures information is not as X chest film picture picture (the figure n1 through height cap conversion process, n2, n3) and through the X chest film picture picture that the inventive method strengthens (scheme o1, o2, o3) be kept perfectly, its visual effect is also not as X chest film picture picture (the figure n1 through height cap conversion process, n2, n3) and through the X chest film picture picture that the inventive method strengthens (scheme o1, o2, o3).Compare through the X chest film picture picture (figure n1, n2, n3) of height cap conversion process and the X chest film picture picture (figure o1, o2, o3) that strengthens through the inventive method, intensity contrast difference between background is bigger on every side for class circle structure and its in the lung in the X chest film picture picture (figure o1, o2, o3) that the inventive method strengthens, and detail textures is more clear.The The above results explanation, compared with prior art, this paper method has better intensity contrast for class circle structure focus and its surrounding tissue in the X rabat lung and strengthens effect.
In addition, also adopt signal to noise ratio (S/N ratio) that above-mentioned three kinds of algorithms are compared result such as table 1; Signal to noise ratio (S/N ratio) is defined as:
In the formula,
Be the image average,
Be standard deviation.By the signal to noise ratio (S/N ratio) definition as can be known, signal to noise ratio (S/N ratio) is big more, and then picture quality is good more, and key diagram image intensifying effect is good more.As can be seen from Table 1, the signal to noise ratio (S/N ratio) of document [16] algorithm is 0.5850, and the signal to noise ratio (S/N ratio) of height cap conversion is 1.7852, and the signal to noise ratio (S/N ratio) maximum of the inventive method is 2.0898; This result shows once more: and the prior art ratio, the inventive method has better enhancing effect to X chest film picture picture and lung's node image.
Table 1, the average signal-to-noise ratio of each algorithm on the JRST image library
| Document [16] algorithm | Top-Bottom-Hat | Algorithm of the present invention | |
| SNR | 0.5850 | 1.7852 | 2.0898 |
List of references
[1]?Ahmedin.?Jemal,?Rebecca?Siegel,?Elizabeth?Wand,?Taylor?Murray,?Jiayuan?Xu?and?Michael?J.?Thun:?Cancer?statistics,?2007.?CA?Cancer?J.?Clin.?2007,?57:?43-66.
[2]?Bram?van?Ginneken,?Bart?M.?ter?Haar?Romeny,?and?Max?A.?Viergever:?Computer-Aided?Diagnosis?in?Chest?Radiography:?A?Survey.?IEEE?Transactions?on?medical?imaging,?2001,?20(12):?1228-1241.
[3]?K.?Doi,?“Computer-aided?diagnosis?in?medical?imaging:?historical?review,?current?status?and?future?potential,”?Computerized?Medical?Imaging?and?Graphics,?2007,?31(4-5):?198-211.
[4]?Quekel?L?G,?Kessels?A?G,?Goei?R,?et?al.?Miss?rate?of?lung?cancer?on?the?chest?radiograph?in?clinical?practice?.Chest.?1999,?115?:720-724
[5]?I.C.?Mchta,?Z.J.?Khan?and?R.R.?Khotpal:?Aanlysi?and?Review?of?Chest?Radiograph?enhancement?Techniques.?Information?Technology?Journal,2006,?5(3):577-582.
[6] He Qian, He Wen, Wang Keyang, etc. digital rabat multifrequency territory aftertreatment is to the influence of computer-aided detection system lung nodule detection. Chinese medicine image technology, 2007,23 (12): 1888-1891.
[7] Liu Hong, Shen Liming, the happy prestige of building: the research of x-ray image enhancement process. science and technology and engineering, 2007,7 (22): 5763-5766.
[8]?SM?Pizer,?EP?Amburn,?JD?Austin,?R?Cromartie?etc.?:?Adaptive?histogram?equalization?and?its?variations.?Computer?Vision,?Graphics,?and?Image?Processing.?1987(39):355-368.
[9]?Andrea?Polesel,?Giovanni?Ramponi?and?V.?John?Matthews,?Image?enhancement?via?Adaptive?Unsharp?Masking,?IEEE?Trans.?on?Image?Processing,?2000,9(3):505-510
[10]?Martin?Stahl,?Til?Aach?and?Sabine?Dippel:?Digital?radiography?enhancement?by?nonlinear?multiscale?processing.?Med.?Phys.?2000,27(1):?56-65.
[11] Sun Zengguo, Han Chongzhao: based on the figure image intensifying of Laplacian operator. computer utility and research, 2007,24 (1): 222-223.
[12] Chen Zhencheng, woods is firm, Zhang Feng, etc.: based on the research of the chest x-ray digital image Enhancement Method of rough set theory. photon journal, 2002,31 (8): 1020-1023
[13]?Q.Li,?S.?Katsuragawa?and?K.?Doi,?Computer-aided?diagnostic?scheme?for?lung?nodule?detection?in?digital?chest?radiographs?by?use?of?a?multiple-template?matching?technique,?
Medical?Physics?28?(2001):?2070-2076.
[14]?Fisher,?R.,?Perkins,?S.,?Walker,?A.,?Wolfart:?“Laplacian/Laplacian?of?Gaussian”,?http://homepages.inf.ed.ac.uk/rbf/HIPR2/log.htm
[15]?Zhenghao?Shi,?Jun?Bai,?Lifeng?He,?Tsuyoshi?Nakamura,?Quanzhu?Yao,?Hidenori?Itoh:?A?Method?for?Enhancing?Lung?Nodules?in?Chest?Radiographs?by?Use?of?LoG?Filter.?In?Proc.?of?CISP’09,?2009(1):417-420.
[16]
Shi Zhenghao, the White Army, Yao Quanzhu, Zhou Peidong: in conjunction with the tubercle figure of the X rabat lung image intensifying of vision and direction yardstick LoG operator.Computer engineering and application, 46 (11): 151-154,2010.
[17]?Soille?P.?Morphological?image?analysis:principles?and?applications?.New?York:?Springer?Verlag?Inc,?1999,?:236-245?.
[18]?Park?H,Yoo?J.?Structuring?element?decomposition?for?efficient?implementation?of?morphological?filters?.IEEE?Image?Signal?Process.?2001,?148(1):31-35.
[19]?A.?N.?Netravali?and?B.?Prasada:?Adaptive?quantization?of?picture?signals?using?spatial?masking. Proc.?IEEE,?1997,65(4):?536-548.
[20]?http://www.jsrt.or.jp/english03.html.
[21]?D.Blostein?and?N.Ahuja?.A?multiscale?region?detector.?Comput.?Vision?Graphics?Image?Process.45.1989:22-41
Claims (4)
1. the doubtful tubercle figure of a lung image intensifying direction scale filter method is characterized in that, comprises following operation steps:
Step 1,
The input original image utilizes the high cap in the gray scale morphology, low cap transformation operator to calculate the peak image and the valley image of original X rabat respectively;
Step 2,
Utilizing direction yardstick Laplce Gaussian function as matched filter similar round lesion image in the lung in the X rabat peak image to be carried out orientation strengthens, in figure image intensifying computing, introduce the vision correcting factor, direction yardstick Laplce Gauss operator convolved image is carried out vision correcting;
Step 3,
The image that step 2 is obtained adds that original X chest film picture picture deducts the valley image that step 1 obtains again, can obtain output image.
2. image enchancing method according to claim 1 is characterized in that: in the described step 1, calculate the peak image of original X rabat and the method for valley image and be:
Wherein,
Wherein,
The expression input picture,
Represent selected structural element,
Represent the peak image corresponding with input picture,
Represent the valley image corresponding, the gray-scale value of x and y difference remarked pixel point horizontal direction and vertical direction with input picture; Described structural element is that radius is formed multiple dimensioned disc structure at interval [5 15] a plurality of disc structure elements.
3. image enchancing method according to claim 2, it is characterized in that: in the described step 2, utilize direction yardstick Laplce Gaussian function similar round lesion image in the lung in the X rabat peak image to be carried out orientation and strengthen, be shown below as matched filter:
In the formula,
Expression utilizes direction yardstick Laplce Gaussian function that similar round lesion image in the lung in the X rabat peak image is carried out directed enhanced results image, the gray-scale value of x and y difference remarked pixel point horizontal direction and vertical direction;
Described direction yardstick Laplce Gaussian function is:
In the formula,
Represent that respectively Laplce's Gaussian function is in x and y direction scale factor, be used to control the filter scale of Laplce's Gaussian function in x and y direction, n is the size of Laplce's Gaussian function convolution kernel, the gray-scale value of x, y difference remarked pixel point horizontal direction and vertical direction;
The method of introducing the vision correcting factor in figure image intensifying computing is:
In the formula,
Expression vision correcting image, β is the vision correcting factor, and the β value is in [1.5 2.5] scope, and the β value preferentially gets 2.
4. image enchancing method according to claim 3 is characterized in that: in the described step 3, the image that step 2 is obtained adds that the method that original X chest film picture picture deducts the valley image that step 1 obtains again is:
In the formula,
Expression strengthens output image as a result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010545654 CN101989351A (en) | 2010-11-16 | 2010-11-16 | Suspected lung nodule image enhancement directional scale filtering method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010545654 CN101989351A (en) | 2010-11-16 | 2010-11-16 | Suspected lung nodule image enhancement directional scale filtering method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101989351A true CN101989351A (en) | 2011-03-23 |
Family
ID=43745878
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201010545654 Pending CN101989351A (en) | 2010-11-16 | 2010-11-16 | Suspected lung nodule image enhancement directional scale filtering method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101989351A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104809712A (en) * | 2015-05-15 | 2015-07-29 | 河海大学常州校区 | Rapid image repairing method based on rough set |
| CN104933682A (en) * | 2015-06-03 | 2015-09-23 | 浙江大学 | Integrated denoising method of inscription-like images |
| CN105765625A (en) * | 2013-12-04 | 2016-07-13 | 皇家飞利浦有限公司 | Fluorescence image processing apparatus and method |
| WO2017096946A1 (en) * | 2015-12-07 | 2017-06-15 | 乐视控股（北京）有限公司 | Method and device for locating high-frequency information of image |
| CN109238950A (en) * | 2018-09-06 | 2019-01-18 | 中国兵器工业第五九研究所 | Atmospheric corrosion of metal materials prediction technique based on qualitative analysis and quantitative forecast |
| CN111557676A (en) * | 2020-05-13 | 2020-08-21 | 山东省肿瘤防治研究院（山东省肿瘤医院） | System and equipment for dynamically adjusting target area position according to change of tumor in radiotherapy process |
| CN111583252A (en) * | 2020-05-15 | 2020-08-25 | 宁波威联生物科技有限公司 | Analysis evaluation device and evaluation system for lung tomography image based on electric field CT |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1912927A (en) * | 2006-08-25 | 2007-02-14 | 西安理工大学 | Semi-automatic partition method of lung CT image focus |
-
2010
- 2010-11-16 CN CN 201010545654 patent/CN101989351A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1912927A (en) * | 2006-08-25 | 2007-02-14 | 西安理工大学 | Semi-automatic partition method of lung CT image focus |
Non-Patent Citations (4)
| Title |
|---|
| 《Medical Physics》 20011031 Qiang Li et al Computer-aided diagnostic scheme for lung nodule detection in digital chest radiographs by use of a multiple-template matching technique 全文 1－4 第28卷, 第10期 2 * |
| 《PROCEEDINGS OF THE IEEE》 19770430 Arun N. Netravali et al Adaptive Quantization of Picture Signals Using Spatial Masking 全文 1－4 第65卷, 第4期 2 * |
| 《The 2nd International Congress on Image and Signal Processing》 20091019 Zhenghao Shi et al A Method for Enhancing Lung Nodules in Chest Radiographs by Use of LoG Filter 全文 1－4 , 2 * |
| 《中国优秀硕士学位论文全文数据库 信息科技辑》 20101115 白军 医学X胸片疑似肺部节点图像增强技术的研究 第37－38页4.1.3节，第40页图4－8、4.3.1节第2段最后一句，第41页1－2段、式（4.27）、4.3.2节第2段，第41－42页4.3.3节 1－4 , 第11期 2 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105765625A (en) * | 2013-12-04 | 2016-07-13 | 皇家飞利浦有限公司 | Fluorescence image processing apparatus and method |
| CN105765625B (en) * | 2013-12-04 | 2019-06-04 | 皇家飞利浦有限公司 | Fluorescence image processing device and method |
| CN104809712A (en) * | 2015-05-15 | 2015-07-29 | 河海大学常州校区 | Rapid image repairing method based on rough set |
| CN104809712B (en) * | 2015-05-15 | 2018-01-16 | 河海大学常州校区 | A kind of image fast repairing method based on rough set |
| CN104933682A (en) * | 2015-06-03 | 2015-09-23 | 浙江大学 | Integrated denoising method of inscription-like images |
| CN104933682B (en) * | 2015-06-03 | 2018-07-06 | 浙江大学 | A kind of integrated denoising method of upright stone tablet class image |
| WO2017096946A1 (en) * | 2015-12-07 | 2017-06-15 | 乐视控股（北京）有限公司 | Method and device for locating high-frequency information of image |
| CN109238950A (en) * | 2018-09-06 | 2019-01-18 | 中国兵器工业第五九研究所 | Atmospheric corrosion of metal materials prediction technique based on qualitative analysis and quantitative forecast |
| CN111557676A (en) * | 2020-05-13 | 2020-08-21 | 山东省肿瘤防治研究院（山东省肿瘤医院） | System and equipment for dynamically adjusting target area position according to change of tumor in radiotherapy process |
| CN111557676B (en) * | 2020-05-13 | 2023-12-19 | 山东省肿瘤防治研究院（山东省肿瘤医院） | System and equipment for dynamically adjusting target area position by tumor change in radiotherapy process |
| CN111583252A (en) * | 2020-05-15 | 2020-08-25 | 宁波威联生物科技有限公司 | Analysis evaluation device and evaluation system for lung tomography image based on electric field CT |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yasaka et al. | Deep learning with convolutional neural network in radiology | |
| CN101989351A (en) | Suspected lung nodule image enhancement directional scale filtering method | |
| Haidekker | Advanced biomedical image analysis | |
| US8165385B2 (en) | System and method for three-dimensional image rendering and analysis | |
| US8045770B2 (en) | System and method for three-dimensional image rendering and analysis | |
| JP5260892B2 (en) | Method of processing radiographic images in tomosynthesis for detection of radiological signs | |
| US8265390B2 (en) | Probabilistic segmentation in computer-aided detection | |
| US20190034750A1 (en) | Method, apparatus, device and storage medium for extracting a cardiovisceral vessel from a cta image | |
| CN107633514B (en) | Pulmonary nodule peripheral blood vessel quantitative evaluation system and method | |
| CN109300113A (en) | A kind of Lung neoplasm assisted detection system and method based on improvement Convex Hull Method | |
| Zaidi et al. | Novel quantitative PET techniques for clinical decision support in oncology | |
| Turco et al. | Interpretable machine learning for characterization of focal liver lesions by contrast-enhanced ultrasound | |
| Grégoire et al. | Is PET-based treatment planning the new standard in modern radiotherapy? The head and neck paradigm | |
| CN103679685B (en) | Image processing system and image processing method | |
| CN115100067A (en) | Image reconstruction method and device | |
| Li et al. | Multienergy cone-beam computed tomography reconstruction with a spatial spectral nonlocal means algorithm | |
| Duan et al. | A multiscale contrast enhancement for mammogram using dynamic unsharp masking in Laplacian pyramid | |
| Caballo et al. | Multi‐marker quantitative radiomics for mass characterization in dedicated breast CT imaging | |
| Xiao et al. | A cascade and heterogeneous neural network for CT pulmonary nodule detection and its evaluation on both phantom and patient data | |
| Wan et al. | Spatio‐temporal texture (SpTeT) for distinguishing vulnerable from stable atherosclerotic plaque on dynamic contrast enhancement (DCE) MRI in a rabbit model | |
| Li et al. | Multiscale registration of medical images based on edge preserving scale space with application in image-guided radiation therapy | |
| Oh et al. | Automated detection of bone metastatic changes using serial CT scans | |
| Zhang et al. | Review of value of CT texture analysis and machine learning in differentiating fat-poor renal angiomyolipoma from renal cell carcinoma | |
| Kanakatte et al. | A pilot study of automatic lung tumor segmentation from positron emission tomography images using standard uptake values | |
| Chen et al. | Hybrid Attention Fusion Segmentation Network for Diffuse Large B-cell Lymphoma in PET-CT |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20110323 |