CN102663446A - Building method of bag-of-word model of medical focus image - Google Patents
Building method of bag-of-word model of medical focus image Download PDFInfo
- Publication number
- CN102663446A CN102663446A CN2012101232473A CN201210123247A CN102663446A CN 102663446 A CN102663446 A CN 102663446A CN 2012101232473 A CN2012101232473 A CN 2012101232473A CN 201210123247 A CN201210123247 A CN 201210123247A CN 102663446 A CN102663446 A CN 102663446A
- Authority
- CN
- China
- Prior art keywords
- focus
- image
- bag
- medical
- pixel points
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000003902 lesion Effects 0.000 claims description 51
- 239000013598 vector Substances 0.000 claims description 49
- 230000009466 transformation Effects 0.000 claims description 13
- 238000010276 construction Methods 0.000 claims description 12
- 238000009499 grossing Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 4
- 238000003064 k means clustering Methods 0.000 claims description 3
- 238000003759 clinical diagnosis Methods 0.000 abstract description 4
- 230000002349 favourable effect Effects 0.000 abstract 1
- 239000011159 matrix material Substances 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 206010018338 Glioma Diseases 0.000 description 4
- 208000007913 Pituitary Neoplasms Diseases 0.000 description 4
- 206010027191 meningioma Diseases 0.000 description 4
- 238000000691 measurement method Methods 0.000 description 3
- 210000001519 tissue Anatomy 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 208000003174 Brain Neoplasms Diseases 0.000 description 1
- 208000032612 Glial tumor Diseases 0.000 description 1
- 206010030113 Oedema Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 210000004004 carotid artery internal Anatomy 0.000 description 1
- 210000001175 cerebrospinal fluid Anatomy 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 210000004884 grey matter Anatomy 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 208000010916 pituitary tumor Diseases 0.000 description 1
- 210000003625 skull Anatomy 0.000 description 1
- 210000003718 sphenoid sinus Anatomy 0.000 description 1
- 238000011426 transformation method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Landscapes
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention relates to recognition of medical focus images, in particular to a building method of a bag-of-word model of a medical focus image. The method includes that first the medical focus image is divided into a focus area and a focus boundary area, then bags of words of the focus area and the focus boundary area are obtained respectively, and finally the bag-of-word model combining the bag of words of the focus area and the bag of words of the focus boundary area is built. Relative space position information of focus area local characteristics is added on the bag-of-word model built through the building method compared with a common bag-of-word model, and the building method is favorable for improving accuracy of clinical diagnosis.
Description
Technical Field
The invention relates to recognizing medical focus images, in particular to a word bag model for recognizing medical focus images.
Background
The basic idea of Content-Based image retrieval (CBIR) is to extract visual features of an image for image representation, and retrieve the image using the image representation. CBIR technology can provide support and help for managing image data, clinical diagnosis, medical teaching, and the like. In particular, the retrieval of similar lesions in medical images may improve the reliability of clinical diagnosis and the integrity of relevant information.
Generally, the CBIR system stores an image to be queried and an image representation of the image to be queried in a database, and when a user provides a query image, the CBIR system extracts visual features of the image for image representation, compares the visual features with the image representation of the image to be queried in the database, and returns an image similar to the query features. The visual features commonly used for image representation in CBIR systems are: color, texture, shape, edges, etc. The texture features are one of the important features of the image, and can describe the features of smoothness, sparseness, regularity and the like of the image. In medical images, since the texture features of the extracted images can reflect detail information hidden in image structures that cannot be observed by human eyes, the texture features are widely used in image expression of medical image retrieval. Several conventional methods for extracting texture features include: gray level co-occurrence matrices, wavelet transforms, Gabor (Gabor) transforms, etc. The gray level co-occurrence matrix represents the spatial dependence of the pixel gray level in the uniform texture mode, but the gray level co-occurrence matrix does not completely capture the image characteristics of local gray level, and the effect of extracting the texture features by the method is not ideal for larger local parts. The wavelet transform can provide a clear mathematical framework for multi-scale requirements, and therefore, the wavelet transform can effectively extract multi-scale features of texture images. However, in the wavelet transform, the problem of filter bank selection has not been solved yet, resulting in an influence on texture analysis. The gamma transformation is most matched with human vision and plays an important role in image analysis, but the size of a transformation window of the gamma transformation is fixed, and the gamma transformation is not sensitive to fine transformation of textures in direction and frequency and cannot meet the requirements of people in practical application. The bag of Words (BoW) model captures subtle changes and overall features of an image by extracting local detail features of the whole image and further performing quantitative expression on the image.
However, when extracting local features, the general bag-of-words model does not take into account the relative spatial position relationship between the features, so that the generated bag-of-words model lacks effective spatial information, and the description capability of the bag-of-words model is reduced. To solve this problem, researchers have proposed a method of combining the spatial pyramid model and the bag-of-words model for classification of natural images (S.Lazebnik, C.Schmid, and J.Ponce, "Beyond bands of features: spatial pyramids for acquiring natural scene targets," IEEE Conference on Computer Vision and Pattern recognition, pp.2169-2178, 2006). The method is characterized in that an image is subdivided into a plurality of sub-regions in a spatial pyramid model extraction mode, then a bag-of-words model of each sub-region is constructed, and finally the bag-of-words models of all the sub-regions are combined into a large bag-of-words model. In addition, researchers have proposed adding spatial coordinates of features to feature vectors of bag-of-words models to improve classification and retrieval performance of Medical X-ray films (U.Avni, H.Greenspan, E.Konen, M.Sharon, and J.Goldberger, "X-ray classification and retrieval on the organ and clinical level, using batch-based services," IEEE Transactions on Medical Imaging, vol.30, pp.733-746, 2011). Because the medical focus image is generally a gray image, the texture information in the focus tissue is different from the adjacent normal tissue, and the texture information of different types of focuses is also different, and in addition, the adjacent tissue structures of the focuses are also different. For example, in brain tumor images, gliomas often have a ring of edema around their margins; meningiomas are adjacent to the skull, gray matter and cerebrospinal fluid; pituitary tumors often appear in the sphenoid sinuses, the visual cross and in the vicinity of the internal carotid artery. Therefore, although the above prior art can well express the texture features in the focus region, the accuracy of distinguishing different types of focuses is not ideal due to the lack of the focus boundary region and the gray scale change information of the focus region transitioning to the focus boundary region.
Disclosure of Invention
The invention aims to solve the technical problem of providing a construction method of a bag-of-words model of a medical focus image, wherein the bag-of-words model obtained by the method can reflect the spatial position of a focus area and is beneficial to improving the accuracy of clinical diagnosis.
The object of the invention can be achieved by the following technical measures:
a method for constructing a bag-of-words model of a medical focus image comprises the following steps:
(1) reading medical focus images with focus outlines drawn in a database, wherein the read medical focus images contain at least 50 images of each focus type, and each medical focus image is processed as follows:
(1.1) firstly, performing one-dimensional Gaussian smoothing on a focus contour line, then respectively taking pixel points of which the number is the same as that of pixel points which are taken from the focus contour line to the inside and the outside of a focus along the normal line of the focus contour as starting points, then respectively taking the pixel points taken from the normal line direction of each focus contour as columns, arranging the pixel points clockwise or anticlockwise, taking the pixel points at the same position in the normal line direction of each focus contour as lines, and arranging the pixel points in the sequence from the inside of the focus contour to the outside of the focus contour or from the outside of the focus contour to the inside of the focus contour to obtain a transformation image of a boundary region of the medical focus image; then, taking pixel points at intervals of 0-5 pixel point distances in a transformed image as centers to expand outwards, obtaining a series of small squares with the pixel point array size of 5 × 5, 7 × 7 or 9 × 9, turning each row of pixel points in the small squares by 90 degrees anticlockwise, then arranging the pixel points in the small squares into a row of queues from top to bottom, then respectively replacing the pixel points in the transformed image in each row of queues with corresponding gray values, and respectively giving zero to the gray values of the pixel points outside the transformed image to obtain a gray value vector A1 of the focus boundary region;
(1.2) in a focus area, extending outwards by taking pixel points at the distance of 0-5 pixel points as centers from pixel points at the edge of a focus contour line, acquiring a series of small squares with the pixel point array size of 5 × 5, 7 × 7 or 9 × 9, rotating each row of pixel points in the small squares by 90 degrees anticlockwise, arranging the pixel points in a row of queues from top to bottom, respectively replacing the pixel points in the focus contour line in each row of queues with corresponding gray values, and respectively giving zero to the gray values of the pixel points outside the focus contour line to obtain a gray value vector A2 of the focus area;
(1.3) specifying gray value vectors A1 of the lesion boundary regions of all the medical lesion images obtained in the step (1.1) as one type, specifying gray value vectors A2 of the lesion regions of all the medical lesion images obtained in the step (1.2) as another type, and then respectively carrying out k-means clustering on the gray value vectors A2 to obtain a lesion region dictionary and a lesion boundary region dictionary which comprise words formed by clustering centers;
(2) taking any medical focus image in the database, and respectively constructing a word bag B2 in a focus area and a word bag B1 in a focus boundary area; wherein,
(2.1) the construction method of the bag of words B2 in the focus area comprises the following steps: expanding outwards by taking each pixel point in the focus area as the center, acquiring a series of small squares with the same size as the small squares in the step (1.2), then obtaining a gray value vector A2 ' of secondary operation of the focus area of the medical focus image by the same method in the step (1.2), then respectively calculating the Euclidean distance between each gray value vector A2 ' and each word in a focus area dictionary according to the sequence of the line queue, and quantizing each gray value vector A2 ' to the word with the minimum Euclidean distance to obtain the frequency of each word in the focus area dictionary, wherein the one-dimensional vector formed by the frequency number is the bag B2 of the focus area of the medical focus image;
(2.2) the construction method of the bag of words B1 in the lesion boundary region comprises the following steps: processing the focus boundary region of the medical focus image according to the same method of the step (1.1) to obtain a transformation image of the boundary region, expanding outwards by taking each pixel point in the transformed image as a center to obtain a series of small squares with the same size as the small squares in the step (1.1), obtaining a gray value vector A1' of secondary operation of a focus boundary region of the medical focus image by the same method in the step (1.1), then respectively calculating the Euclidean distance between each gray value vector A1' and each word in the lesion boundary region dictionary according to the sequence of the line queue, and quantizes each gray value vector A1' to the word with the minimum Euclidean distance to the word to obtain the frequency of each word in the focus boundary region dictionary, the one-dimensional vector formed by the frequency number is a word bag B1 of the lesion boundary area of the medical lesion image;
(2.3) arranging two vectors representing the bag-of-words B1 and the bag-of-words B2 end to end into a vector to obtain a bag-of-words model of the medical lesion image.
The construction method of the invention is characterized in that the size of the small square is preferably 7 x 7, the k value of the k-means cluster can be obtained according to the number of the images, and the principle of obtaining the k value is that the more the number of the images is, the larger the k value is; the smaller the number of images, the smaller the k value is obtained, and generally 1000 is preferable.
The construction method is suitable for constructing the bag-of-word models of MRI, CT and ultrasonic images.
Compared with the prior art, the method has the following beneficial effects:
because the image database is composed of images with well-delineated lesion outlines, the lesion images are divided into two parts, namely a lesion area and a lesion boundary area according to the characteristics of the lesion images, word bags of the lesion area and the lesion boundary area are respectively obtained, and then a word bag model combining the word bags of the lesion area and the word bags of the lesion boundary area is constructed.
Drawings
FIG. 1 is a flow chart of a method of constructing a bag-of-words model based on medical lesion images according to the present invention;
FIG. 2 is a recall-precision curve, wherein the curve numbered I is the recall-precision curve obtained using the bag-of-words model of the present invention; the curve numbered II is a recall-precision curve obtained by constructing a word bag B2 in a focus area; the curve numbered as III is a recall-precision curve obtained by a gray level co-occurrence matrix method; the curve numbered IV is a recall-precision curve obtained by a wavelet transform method; the curve with the number V is a recall-precision curve obtained by using a gamma transformation method;
FIG. 3 is a recall-precision curve, wherein the curve numbered I is the recall-precision curve obtained using the bag-of-words model of the present invention; the curve numbered II is a recall-precision curve obtained by constructing a bag-of-words model by adopting a space pyramid method; the curve numbered III is a recall-precision curve obtained by constructing a bag-of-words model by adding a spatial coordinate of a feature to a feature vector.
Detailed Description
Example 1 (construction bag of words model)
The database used in this example is 800 each of T1 weighted enhanced MRI images of the brain with meningiomas, gliomas, and pituitary tumors, each MRI image having delineated the lesion. The method for constructing the bag-of-words model of any MRI image in the database is described in detail below with reference to fig. 1.
(1) Reading the MRI images in the database, and performing the following processing on each MRI image:
(1.1) firstly, performing one-dimensional Gaussian smoothing on a focus contour line, respectively taking 15 pixel points from the focus contour line to the inside and the outside of the focus along the focus contour normal line as a starting point, respectively, then respectively taking the pixel points taken in the direction of each focus contour normal line as a column, arranging the pixel points clockwise or anticlockwise, taking the pixel points at the same position in the direction of each focus contour normal line as a line, and arranging the pixel points in the sequence from the inside of the focus contour to the outside of the focus contour or from the outside of the focus contour to the inside of the focus contour to obtain a transformation image of a boundary region of the medical focus image; then, extending outwards by taking pixel points with the distance of 5 pixel points in the transformed image as centers to obtain a series of small squares with the pixel point array size of 7 multiplied by 7, rotating each row of pixel points in the small squares by 90 degrees anticlockwise, then arranging the pixel points in the transformed image into a row of queues from top to bottom, then respectively replacing the pixel points in each row of queues in the transformed image with corresponding gray values, and respectively giving zero to the gray values of the pixel points outside the transformed image to obtain a gray value vector A1 of the focus boundary area. The purpose of the one-dimensional Gaussian smoothing is to avoid the change of the normal direction of the focus contour line caused by the interference of noise on the focus contour line, and a Gaussian kernel is used for smoothing the focus contour line. The one-dimensional gaussian smoothing method is as follows:
one-dimensional Gaussian kernel of
Where σ denotes the standard deviation and X denotes the position of any point in space. Setting the coordinate of the point on the focus contour line of a focus as b (x, y), and the coordinate of the point on the focus contour line after Gaussian kernel smoothing as
B(x′，y′)＝b(x′，y′)*G1D(X；σ)′，
Where σ and X are the same as σ and X in the Gaussian kernel expression. Thus, the angle between the normal direction of the lesion contour line and the tangential direction of the lesion contour line is
Using the angle θ, the coordinates of the point corresponding to the normal line of each lesion contour can be obtained
x1＝x+l×cosθ，
y1＝y+l×sinθ，
Where l is the distance from the point taken in the direction of the normal to the lesion contour to the point on the lesion contour. Due to the coordinates (x)1，y1) It is not necessarily on a certain pixel point on the image, therefore, the linear interpolation is used to obtain the pixel point corresponding to the point taken in the normal direction of each focus contour.
(1.2) in the focus area, extending outwards by taking pixel points at the distance of 5 pixel points as centers from pixel points at the edge of the focus contour line to obtain a series of small squares with the pixel point array size of 7 multiplied by 7, turning each row of pixel points in the small squares by 90 degrees anticlockwise, then arranging the pixel points in each row of queues from top to bottom into a row of queues, then respectively replacing the pixel points in the focus contour line in each row of queues with corresponding gray values, and respectively giving zero to the gray values of the pixel points outside the focus contour line to obtain a gray value vector A2 of the focus area.
(1.3) defining the gray value vectors A1 of the lesion boundary regions of all the medical lesion images obtained in the step (1.1) as one type, defining the gray value vectors A2 of the lesion regions of all the medical lesion images obtained in the step (1.2) as another type, and then respectively carrying out k-means clustering on the gray value vectors A2 to obtain a lesion region dictionary and a lesion boundary region dictionary which are formed by clustering centers and respectively comprise 1000 words.
(2) Taking any medical focus image in the database, and respectively constructing a word bag B2 in a focus area and a word bag B1 in a focus boundary area; wherein,
(2.1) the construction method of the bag of words B2 in the focus area comprises the following steps: and (3) expanding outwards by taking each pixel point in the focus area as the center to obtain a series of small squares with the same size as the small squares in the step (1.2), obtaining a gray value vector A2 ' of secondary operation of the focus area of the medical focus image by the same method in the step (1.2), respectively calculating the Euclidean distance between each gray value vector A2 ' and each word in a focus area dictionary according to the sequence of the line queue, and quantizing each gray value vector A2 ' to the word with the minimum Euclidean distance to obtain the frequency of each word in the focus area dictionary, wherein a one-dimensional vector formed by the frequency number is the bag B2 of the focus area of the medical focus image.
(2.2) the construction method of the bag of words B1 in the lesion boundary region comprises the following steps: processing the focus boundary region of the medical focus image according to the same method of the step (1.1) to obtain a transformation image of the boundary region, expanding outwards by taking each pixel point in the transformed image as a center to obtain a series of small squares with the same size as the small squares in the step (1.1), obtaining a gray value vector A1' of secondary operation of a focus boundary region of the medical focus image by the same method in the step (1.1), then respectively calculating the Euclidean distance between each gray value vector A1' and each word in the lesion boundary region dictionary according to the sequence of the line queue, and quantizes each gray value vector A1' to the word with the minimum Euclidean distance to the word to obtain the frequency of each word in the focus boundary region dictionary, the one-dimensional vector formed by the frequency number is the bag of words B1 in the lesion boundary region of the medical lesion image.
(2.3) arranging two vectors representing the bag-of-words B1 and the bag-of-words B2 end to end into a vector to obtain a bag-of-words model of the medical lesion image.
Example 2 (verification of Effect)
1. Building CBIR systems
(1) Adopting the construction method described in example 1 to construct a bag-of-word model of each image in the database described in example 1, and then using the obtained bag-of-word model of 2400 MRI images to construct a CBIR system 1 for inquiring meningiomas, gliomas and pituitary adenomas;
(2) similarly, 2400 MRI images are used, and the CBIR system 2, the CBIR system 3, the CBIR system 4 and the CBIR system 5 are respectively constructed by using the bag-of-words B2 of the lesion area as a bag-of-words model and texture features obtained by gray level co-occurrence matrix, wavelet transform and gamma transform.
(3) Similarly, the CBIR system 6 and the ICBIR system 7 are constructed by using the 2400 MRI images and respectively using a bag-of-words model constructed by a spatial pyramid method and a bag-of-words model constructed by a method of adding spatial coordinates of features to feature vectors.
2. Retrieval
(1) 100 pieces of T1 weighted enhanced MRI images of brain of meningioma, glioma and pituitary tumor are submitted to the CBIR system 1 constructed in the step 1 as query images, the system 1 firstly constructs a bag-of-word model of the query images according to the method described in the example 1, then compares the similarity of the query images with each image in the CBIR system by adopting a distance measurement method, and then arranges the images in the CBIR system according to the sequence of the similarity of the corresponding query images from high to low. The number of the returned images at one time is set to be 1, 2, … …, n and … … 2400 respectively, the precision ratio and the recall ratio of each image are obtained, and then the average value of the precision ratio and the recall ratio of 300 query images is calculated to obtain a curve with the number I in fig. 2 or fig. 3.
(2) And similarly, respectively submitting the 300 query images to the CBIR system 2, the CBIR system 3, the CBIR system 4 and the CBIR system 5 which are constructed in the step 1, wherein the system 2 constructs a word bag model of the query images according to the focus area word bag B2, the systems 3 to 5 respectively adopt gray level co-occurrence matrix, wavelet transformation and gamma transformation to extract texture features of the query images, respectively adopt a distance measurement method to respectively compare the similarity between the query images and each image in the CBIR systems 2 to 5, and then arrange the images in the CBIR systems 2 to 5 according to the sequence from high to low of the similarity of the corresponding query images. The number of the returned images at one time is set to be 1, 2, … …, n and … … 2400 respectively, the precision ratio and the recall ratio of each image are obtained, and then the mean values of the precision ratio and the recall ratio of 300 query images are calculated to obtain the curves numbered as II, III, IV and V in the graph shown in figure 2. FIG. 2 shows that the precision ratio obtained by using the bag-of-words model of the present invention is much higher than that obtained by other texture feature extraction methods in the relevant images of databases with 90% or less returned by each CBIR system.
(3) And similarly, the 300 query images are respectively submitted to the CBIR system 6 and the CBIR system 7 constructed in the step 1, the system 6 constructs bag-of-words models of the query images according to a space pyramid method, the system 7 constructs the bag-of-words models of the query images according to a method of adding space coordinates of features into feature vectors, the similarity of the query images and each image in the CBIR systems 6 and 7 is respectively compared by adopting a distance measurement method, and then the images in the CBIR systems 6 and 7 are arranged according to the sequence of the similarity of the corresponding query images from high to low. The number of the returned images at one time is set to be 1, 2, … …, n and … … 2400 respectively, the precision ratio and the recall ratio of each image are obtained, and then the average values of the precision ratio and the recall ratio of 300 query images are calculated to obtain curves numbered as II and III in FIG. 3. FIG. 3 shows that in the relevant images of databases with 90% or less returned by each CBIR system, the precision obtained by using the bag-of-words model of the present invention is higher than that obtained by using the other two bag-of-words methods with spatial information added.
Claims (3)
1. A method for constructing a bag-of-words model of a medical focus image comprises the following steps:
(1) reading medical focus images with focus outlines drawn in a database, wherein the read medical focus images contain at least 50 images of each focus type, and each medical focus image is processed as follows:
(1.1) firstly, performing one-dimensional Gaussian smoothing on a focus contour line, then respectively taking pixel points of which the number is the same as that of pixel points which are taken from the focus contour line to the inside and the outside of a focus along the normal line of the focus contour as starting points, then respectively taking the pixel points taken from the normal line direction of each focus contour as columns, arranging the pixel points clockwise or anticlockwise, taking the pixel points at the same position in the normal line direction of each focus contour as lines, and arranging the pixel points in the sequence from the inside of the focus contour to the outside of the focus contour or from the outside of the focus contour to the inside of the focus contour to obtain a transformation image of a boundary region of the medical focus image; then, taking pixel points at intervals of 0-5 pixel point distances in a transformed image as centers to expand outwards, obtaining a series of small squares with the pixel point array size of 5 × 5, 7 × 7 or 9 × 9, turning each row of pixel points in the small squares by 90 degrees anticlockwise, then arranging the pixel points in the small squares into a row of queues from top to bottom, then respectively replacing the pixel points in the transformed image in each row of queues with corresponding gray values, and respectively giving zero to the gray values of the pixel points outside the transformed image to obtain a gray value vector A1 of the focus boundary region;
(1.2) in a focus area, extending outwards by taking pixel points at the distance of 0-5 pixel points as centers from pixel points at the edge of a focus contour line, acquiring a series of small squares with the pixel point array size of 5 × 5, 7 × 7 or 9 × 9, rotating each row of pixel points in the small squares by 90 degrees anticlockwise, arranging the pixel points in a row of queues from top to bottom, respectively replacing the pixel points in the focus contour line in each row of queues with corresponding gray values, and respectively giving zero to the gray values of the pixel points outside the focus contour line to obtain a gray value vector A2 of the focus area;
(1.3) specifying gray value vectors A1 of the lesion boundary regions of all the medical lesion images obtained in the step (1.1) as one type, specifying gray value vectors A2 of the lesion regions of all the medical lesion images obtained in the step (1.2) as another type, and then respectively carrying out k-means clustering on the gray value vectors A2 to obtain a lesion region dictionary and a lesion boundary region dictionary which comprise words formed by clustering centers;
(2) taking any medical focus image in the database, and respectively constructing a word bag B2 in a focus area and a word bag B1 in a focus boundary area; wherein,
(2.1) the construction method of the bag of words B2 in the focus area comprises the following steps: expanding outwards by taking each pixel point in the focus area as the center, acquiring a series of small squares with the same size as the small squares in the step (1.2), then obtaining a gray value vector A2 ' of secondary operation of the focus area of the medical focus image by the same method in the step (1.2), then respectively calculating the Euclidean distance between each gray value vector A2 ' and each word in a focus area dictionary according to the sequence of the line queue, and quantizing each gray value vector A2 ' to the word with the minimum Euclidean distance to obtain the frequency of each word in the focus area dictionary, wherein the one-dimensional vector formed by the frequency number is the bag B2 of the focus area of the medical focus image;
(2.2) the construction method of the bag of words B1 in the lesion boundary region comprises the following steps: processing the focus boundary region of the medical focus image according to the same method of the step (1.1) to obtain a transformation image of the boundary region, expanding outwards by taking each pixel point in the transformed image as a center to obtain a series of small squares with the same size as the small squares in the step (1.1), obtaining a gray value vector A1' of secondary operation of a focus boundary region of the medical focus image by the same method in the step (1.1), then respectively calculating the Euclidean distance between each gray value vector A1' and each word in the lesion boundary region dictionary according to the sequence of the line queue, and quantizes each gray value vector A1' to the word with the minimum Euclidean distance to the word to obtain the frequency of each word in the focus boundary region dictionary, the one-dimensional vector formed by the frequency number is a word bag B1 of the lesion boundary area of the medical lesion image;
(2.3) arranging two vectors representing the bag-of-words B1 and the bag-of-words B2 end to end into a vector to obtain a bag-of-words model of the medical lesion image.
2. The method as claimed in claim 1, wherein the size of the small square is 7 × 7 pixel point array.
3. The method of claim 1 or 2, wherein the k-value of the k-means cluster is 1000.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201210123247 CN102663446B (en) | 2012-04-24 | 2012-04-24 | Building method of bag-of-word model of medical focus image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201210123247 CN102663446B (en) | 2012-04-24 | 2012-04-24 | Building method of bag-of-word model of medical focus image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102663446A true CN102663446A (en) | 2012-09-12 |
| CN102663446B CN102663446B (en) | 2013-04-24 |
Family
ID=46772929
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201210123247 Expired - Fee Related CN102663446B (en) | 2012-04-24 | 2012-04-24 | Building method of bag-of-word model of medical focus image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102663446B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102968618A (en) * | 2012-10-24 | 2013-03-13 | 浙江鸿程计算机系统有限公司 | Static hand gesture recognition method fused with BoF model and spectral clustering algorithm |
| CN103310208A (en) * | 2013-07-10 | 2013-09-18 | 西安电子科技大学 | Identifiability face pose recognition method based on local geometrical visual phrase description |
| CN103345645A (en) * | 2013-06-27 | 2013-10-09 | 复旦大学 | Commodity image category forecasting method based on online shopping platform |
| CN103399870A (en) * | 2013-07-08 | 2013-11-20 | 华中科技大学 | Visual word bag feature weighting method and system based on classification drive |
| CN104598881A (en) * | 2015-01-12 | 2015-05-06 | 中国科学院信息工程研究所 | Feature compression and feature selection based skew scene character recognition method |
| CN105787494A (en) * | 2016-02-24 | 2016-07-20 | 中国科学院自动化研究所 | Brain disease identification method based on multistage partition bag-of-words model |
| CN106170799A (en) * | 2014-01-27 | 2016-11-30 | 皇家飞利浦有限公司 | From image zooming-out information and information is included in clinical report |
| CN108510497A (en) * | 2018-04-10 | 2018-09-07 | 四川和生视界医药技术开发有限公司 | The display methods and display device of retinal images lesion information |
| CN113506313A (en) * | 2021-07-07 | 2021-10-15 | 上海商汤智能科技有限公司 | Image processing method and related device, electronic equipment and storage medium |
| CN114119578A (en) * | 2021-12-01 | 2022-03-01 | 数坤(北京)网络科技股份有限公司 | Image processing method and device, computer equipment and storage medium |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107180422A (en) * | 2017-04-02 | 2017-09-19 | 南京汇川图像视觉技术有限公司 | A kind of labeling damage testing method based on bag of words feature |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102117337A (en) * | 2011-03-31 | 2011-07-06 | 西北工业大学 | Space information fused Bag of Words method for retrieving image |
| CN102129477A (en) * | 2011-04-23 | 2011-07-20 | 山东大学 | Multimode-combined image reordering method |
| US20110219360A1 (en) * | 2010-03-05 | 2011-09-08 | Microsoft Corporation | Software debugging recommendations |
-
2012
- 2012-04-24 CN CN 201210123247 patent/CN102663446B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110219360A1 (en) * | 2010-03-05 | 2011-09-08 | Microsoft Corporation | Software debugging recommendations |
| CN102117337A (en) * | 2011-03-31 | 2011-07-06 | 西北工业大学 | Space information fused Bag of Words method for retrieving image |
| CN102129477A (en) * | 2011-04-23 | 2011-07-20 | 山东大学 | Multimode-combined image reordering method |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102968618A (en) * | 2012-10-24 | 2013-03-13 | 浙江鸿程计算机系统有限公司 | Static hand gesture recognition method fused with BoF model and spectral clustering algorithm |
| CN103345645B (en) * | 2013-06-27 | 2016-09-28 | 复旦大学 | Commodity image class prediction method towards net purchase platform |
| CN103345645A (en) * | 2013-06-27 | 2013-10-09 | 复旦大学 | Commodity image category forecasting method based on online shopping platform |
| CN103399870A (en) * | 2013-07-08 | 2013-11-20 | 华中科技大学 | Visual word bag feature weighting method and system based on classification drive |
| CN103310208A (en) * | 2013-07-10 | 2013-09-18 | 西安电子科技大学 | Identifiability face pose recognition method based on local geometrical visual phrase description |
| CN103310208B (en) * | 2013-07-10 | 2016-05-11 | 西安电子科技大学 | The distinctive human face posture recognition methods of describing based on local geometric vision phrase |
| CN106170799B (en) * | 2014-01-27 | 2021-01-22 | 皇家飞利浦有限公司 | Extracting information from images and including information in clinical reports |
| CN106170799A (en) * | 2014-01-27 | 2016-11-30 | 皇家飞利浦有限公司 | From image zooming-out information and information is included in clinical report |
| CN104598881B (en) * | 2015-01-12 | 2017-09-29 | 中国科学院信息工程研究所 | Feature based compresses the crooked scene character recognition method with feature selecting |
| CN104598881A (en) * | 2015-01-12 | 2015-05-06 | 中国科学院信息工程研究所 | Feature compression and feature selection based skew scene character recognition method |
| CN105787494A (en) * | 2016-02-24 | 2016-07-20 | 中国科学院自动化研究所 | Brain disease identification method based on multistage partition bag-of-words model |
| CN105787494B (en) * | 2016-02-24 | 2019-04-23 | 中国科学院自动化研究所 | Cerebral disease identification device based on multistage subregion bag of words |
| CN108510497A (en) * | 2018-04-10 | 2018-09-07 | 四川和生视界医药技术开发有限公司 | The display methods and display device of retinal images lesion information |
| CN108510497B (en) * | 2018-04-10 | 2022-04-26 | 四川和生视界医药技术开发有限公司 | Method and device for displaying focus information of retina image |
| CN113506313A (en) * | 2021-07-07 | 2021-10-15 | 上海商汤智能科技有限公司 | Image processing method and related device, electronic equipment and storage medium |
| CN114119578A (en) * | 2021-12-01 | 2022-03-01 | 数坤(北京)网络科技股份有限公司 | Image processing method and device, computer equipment and storage medium |
| CN114119578B (en) * | 2021-12-01 | 2022-07-08 | 数坤(北京)网络科技股份有限公司 | Image processing method and device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102663446B (en) | 2013-04-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102663446B (en) | Building method of bag-of-word model of medical focus image | |
| Jadoon et al. | Three‐class mammogram classification based on descriptive CNN features | |
| Nandi et al. | Principal component analysis in medical image processing: a study | |
| Huang et al. | Retrieval of brain tumors with region‐specific bag‐of‐visual‐words representations in contrast‐enhanced MRI images | |
| Oghli et al. | Automatic fetal biometry prediction using a novel deep convolutional network architecture | |
| Hu et al. | Automatic segmentation of left ventricle from cardiac MRI via deep learning and region constrained dynamic programming | |
| CN104881680A (en) | Alzheimer's disease and mild cognitive impairment identification method based on two-dimension features and three-dimension features | |
| CN103793908A (en) | Method for constructing prediction model of multifunctional veins based on brain nuclear magnetic resonance image | |
| CN104834931A (en) | Improved SIFT algorithm based on wavelet transformation | |
| WO2024011835A1 (en) | Image processing method and apparatus, device, and readable storage medium | |
| Chaki et al. | A beginner’s guide to image shape feature extraction techniques | |
| Carneiro et al. | Semantic-based indexing of fetal anatomies from 3-D ultrasound data using global/semi-local context and sequential sampling | |
| Lei et al. | Automatic recognition of fetal standard plane in ultrasound image | |
| Kumaraswamy et al. | Automatic prostate segmentation of magnetic resonance imaging using Res-Net | |
| CN118037791A (en) | Construction method and application of multi-mode three-dimensional medical image segmentation registration model | |
| Liu et al. | MfdcModel: a novel classification model for classification of benign and malignant breast tumors in ultrasound images | |
| Lei et al. | Multi-modal and multi-layout discriminative learning for placental maturity staging | |
| Perez et al. | Diffusion weighted imaging of prostate cancer: prediction of cancer using texture features from parametric maps of the monoexponential and kurtosis functions | |
| Mani | Deep learning models for semantic multi-modal medical image segmentation | |
| CN112183596B (en) | Linear segment matching method and system combining local grid constraint and geometric constraint | |
| Wang et al. | Automatic right ventricular segmentation for cine cardiac magnetic resonance images based on a new deep atlas network | |
| Zhang et al. | Medical Image Retrieval based on co-occurrence martix and edge histogram | |
| Tateyama et al. | 2D-PCA based statistical shape model from few medical samples | |
| Liu et al. | T-test based Alzheimer’s disease diagnosis with multi-feature in MRIs | |
| Cao et al. | Mental disease feature extraction with MRI by 3D convolutional neural network with multi-channel input |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20130424 Termination date: 20200424 |