KR100715763B1 - Method for evaluation of skeletal malformation using atlas matching method - Google Patents

Abstract

The present invention relates to a bone age evaluation method using the atlas matching method, the registration step of extracting the standard bone image by bone age and the standard bone image for each area of the hand bone image and then registering in the database; An extraction step of extracting a bone image for each region of the hand bone from a digital X-ray image of the hand bone of the subject; An image registration step of measuring a difference between the extracted bone image of each region of the hand bone and a standard bone image of each region registered in the database through computer image processing; A similarity measuring step of measuring similarity between the bone image of each region and the standard bone image of each region of the image-matched subject; And an allocating step of allocating bone age using the measured similarity and a pre-trained classifier. According to the bone age evaluation method using the disclosed atlas matching method, the difference between the bone images of the extracted hand bones and the bone images of the standard hand bones by area registered in the database can be objectively and reproducibly used by measuring the similarity by image registration. Age can be assessed.

Atlas, Bone Age, Image, Similarity, Hand Bone, Database

Description

Method for evaluation of skeletal malformation using Atlas matching method}

1 is a flow chart according to the bone age evaluation method using the atlas matching method according to the present invention,

2 is a diagram illustrating a human hand bone structure.

The present invention relates to a method for assessing bone age, and more specifically, to evaluate bone age in an objective and reproducible manner, and to allow a computer to perform the process of the atlas matching method, so that there is less rejection and it is possible to settle in the clinic quickly. The present invention relates to a bone age estimation method using atlas matching.

In general, assessment of bone age is one of the most frequently treated subjects in pediatrics. The child's left hand is taken with X-rays, and then the image is evaluated for bone age. Bone age assessment is an important indicator of whether a child is developing normally. If there is a big difference between the actual age and the measured bone age, it can be determined that there is an abnormal component of bone growth, such as diabetes, genetic disease, etc., and appropriate treatment. In addition, there are atlas matching methods and TW2 matching methods for assessing bone age in the X-ray image of the left hand.

First, the atlas matching method is a method proposed by Greulich and Pyle of the United States and evaluated with a book containing the radiographic images of the atlas pattern group, which summarizes the radiographic images of the left hand according to age and gender. Doctors directly assess the bone age by looking directly at a book from the Atlas pattern group with the image that most closely resembles the overall shape of the child's X-ray image.

Currently 76% of physicians who evaluate bone age use this method. However, the atlas matching method is very likely to cause errors in the process of assessing bone age because the variation in bone growth occurs for each doctor's skill or people.

Another method of assessing bone age is the TW2 method proposed by Tanner and Whitehouse of England. This method compares each bone of the left hand with the images in a book published in the TW2 group, without looking at the similarity of the whole bone image as in the atlas matching method.

In atlas matching, similarity is analyzed in the whole bone image, whereas in TW2 method, each bone is separated and analyzed for more accurate bone age than in atlas matching method. However, due to the disadvantage that it takes more time to evaluate bone age than the atlas method, it is not currently used much.

The bone age estimation method using the atlas or TW2 has a disadvantage in that reliability of the bone age is low and it takes a long time to estimate the bone age because the reading specialist evaluates the bone age by visually checking the similarity of the images.

In 1972, Stanly M.Garn published a paper entitled 'Metacarpophalangeal Length in the Evaluation of Skeletal Malformation'. However, this method simply attempted to assess bone age by measuring only the absolute length of the bone and could be used only in limited numbers in the Caucasian middle class.

Accordingly, the need for a quick and accurate bone age estimation method is increasing by overcoming such a conventional bone age estimation method and implementing the bone age estimation method programmatically.

In order to overcome the problems of the prior art, the present invention provides an objective and reproducible bone age difference between the bone images of each region of the extracted hand bone and the standard bone images of each region registered in the database by measuring similarity by image registration. It is an object of the present invention to provide a bone age evaluation method using atlas matching method, which has the advantage of being able to evaluate, less objection and settle in the clinic quickly.

Other objects and advantages of the invention will be described below and will be appreciated by the embodiments of the invention. Furthermore, the objects and advantages of the present invention can be realized by means and combinations indicated in the claims.

In order to achieve the above object, the present invention provides a bone age evaluation method using the atlas matching method, registration step of extracting the standard bone image by bone age and the standard bone image for each area of the hand bone image and then registering in the database; An extraction step of extracting a bone image for each region of the hand bone from a digital X-ray image of the hand bone of the subject; An image registration step of measuring a difference between the extracted bone image of each region of the hand bone and a standard bone image of each region registered in the database through computer image processing; A similarity measuring step of measuring similarity between the bone image of each region of the image-matched subject and the standard bone image of the region; And an allocating step of allocating bone age using the measured similarity and a pre-trained classifier.

Here, the registration step and the extraction step, it is preferable to extract one or more bone images selected from the group consisting of resin bone, hand bone, ulna and radius.

In addition, the image registration step is preferably performed by a combination of a rigid registration step, a non-rigid registration step.

Further, the similarity measuring step is preferably performed by a combination of the similarity measuring step after the rigid matching step, the similarity measuring step after the non-rigid matching step.

In addition, the similarity measuring step after the non-rigid matching step preferably includes a total similarity measuring step or a similarity measuring step for each section.

In addition, the classifier is preferably any one selected from the group consisting of a neural network, a support vector machine and a linear classifier.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Prior to this, terms or words used in the present specification and claims should not be construed as being limited to the common or dictionary meanings, and the inventors should properly explain the concept of terms in order to explain their invention in the best way. Based on the principle that can be defined, it should be interpreted as meaning and concept corresponding to the technical idea of the present invention.

Therefore, the embodiments described herein are only the most preferred embodiments of the present invention and do not represent all of the technical idea of the present invention, and various equivalents and modifications that may substitute them at the time of the present application may be applied. It should be understood that there may be.

1 is a flow chart according to the bone age evaluation method using the atlas matching method according to the present invention. Referring to the drawing, in the standard bone bone image database registration step (S100) by bone age, first, digital X-ray hand bone images of a child determined to be following a normal growth process by experts are collected. The digital X-ray image of the hand bone is obtained with the same type of imaging device and shooting conditions in order to reduce the deviation by the imaging technique.

Next, a reading expert group skilled in reading the bone age of the hand bone for the images reads the bone age. After collecting a large number of bone age reading image data by bone age group, a plurality of representative images are selected by bone age group and registered in the database together with relevant information such as gender, residence area, reverse age, bone age, height and weight. do.

Next, in the standard bone image extraction and database registration step for each region of the hand bone image (S200), one or more bone images selected from the group consisting of a resin bone, a hand bone, an ulna and a radial bone are extracted.

First, the phalanges are the epiphyseal regions of the three phalanges, that is, the distal phalanges, the middle phalanges, and the proximal phalanges, as shown in FIG. 2 for three or more fingers. Extracts them automatically or manually.

Methods for automatic or manual extraction of the phalanx region of the resinous bone are known in the art (Pietka E, Gertych A, Pospiech S, Cao F, Huang HK, Gilsanz V. Computer-assisted bone age assessment: image preprocessing and epiphyseal / metaphyseal ROI extraction Removing the background from the X-ray image, detecting the tip and beginning of the finger, using the procedure described in IEEE Trans Med Imaging.2001 Aug; 20 (8): 715-29). Extracting the resin bone from the finger and extracting the bone plate area from the resin bone. At this time, when the angle of the finger is not vertical, the image of the finger portion is aligned and the image of the phalangeal plate area is extracted by aligning the axial direction of the finger vertically by automatic or manual method.

Next, the carpal bone is extracted by the manual or automatic method and registered as a single region image. At this time, a method for automatically or manually extracting the handbone is known literature (Pietka E, Kaabi L, Kuo ML, Huang HK. Feature Extraction in Carpal-Bone Analysis.IEEE TRANSACTIONS ON MEDICAL IMAGING. VOL. 12, NO.I, MARCH 1993) using the method described in the above method, and after extracting all the resinous bones in the remaining area of the metacarpal bone (metacarpal bone) is removed in a similar manner again.

The ulna and radius areas then extract the ulna and radius of the forearm by manual or automatic methods and register each area. In addition, the length of the ulna and the radius may vary from case to case, depending on the shooting conditions.

Digital X-ray image acquisition step (S300) of the subject's hand acquires a digital X-ray image of the hand bone for the subject to determine the bone age. In order to reduce the deviation caused by the change of shooting technology, keep the device and shooting condition of the same model as the standard acquisition method. use.

The image correction processing technique is classified into a method of image processing for correcting characteristics when a photographing apparatus is different and a method of correcting when a pixel size is different between photographing apparatuses.

First, when the imaging apparatus is different, the image processing method for correcting the characteristic is that when the high frequency component attenuation characteristics (MTF) are different between the imaging apparatuses, the image having the less attenuation of the high frequency component is applied to the image having the higher attenuation. Fit. Next, by applying a high frequency attenuation filter designed based on the high frequency attenuation characteristics of the two-side photographing apparatus previously applied to the image, the image characteristics are corrected to have the same high frequency attenuation pattern.

The correction method in the case of different pixel sizes between imaging apparatuses first corrects the high-frequency attenuation characteristics by the same method as the image processing method in the case where the imaging apparatuses are different, and then resamples the image of the smaller pixel size to obtain the pixel. Correct so that the size of is equal to the larger one. In this case, it is recommended to use a cubic interpolation technique as an interpolation technique for resampling. Calculation of the cubic interpolation technique can be performed using a formula described in known literature (Gonzalez RC, Woods RE. Digital Image Processing. P 300, Addison-Wesley Publishing, 1993).

Extraction of bone image by region of the subject's hand (S400) is to extract the main region of the hand bone from the digital X-ray image of the examinee's hand bone obtained by the digital X-ray image acquisition step (S300) of the subject's hand by an automatic or manual method do. The extraction method uses the same method as the standard bone image extraction and database registration step (S200) for each region of the hand bone image.

Next, the image registration step (S500), the position and direction through the computer image processing to easily measure the difference between the extracted bone image of the area of the examinee's hand and the bone image of each region of the standard person registered in advance Is the process of performing an alignment or proper deformation.

The image registration step S500 is performed by a combination of a rigid registration step and a non-rigid registration step. First, in the rigid registration step, the entire image is moved while the relative positions of the pixels in the image are not changed to match the two images.

Next, change the position and direction of the bone images extracted from each part of the subject's hand and match them with the bone images of the standard subjects to match the most image features. In this case, when the deviation of the area occurs between the two images, the matching degree is not calculated for the out of area, but only for the overlayed area.

In the meantime, the non-rigid registration step first performs a rigid registration step and adds an appropriate modification to the bone image for each region of the matched subject to increase the degree of registration with the standard bone image. At this time, the deformation condition is made within the range of parameter values defined by preliminary experiments so as to have a degree of flexibility for matching between bone images. Next, a deformation map is generated when the non-rigid registration step is completed, and the deformation map includes detailed information that each pixel of the bone image of the examinee needs to be modified in order to match with the standard image.

The similarity measuring step (S600) is performed by a combination of the similarity measuring step after the rigid matching step and the similarity measuring step after the non-rigid matching step. First, the similarity measuring step after the rigid registration calculates the similarity between the bone image of each region of the subject and the bone image of each region of the standard image data set after the rigid registration is performed. As a method for measuring the similarity between the two images, methods such as mean square error, correlation, and mutual information are used.

Next, the similarity measurement step after non-rigid registration is calculated in two ways: the total similarity and the similarity for each section. First, after the total similarity is modified by the non-rigid registration method for the bone image of the subject, the similarity is calculated between the bone images of the standard image data set in the same manner as in the similarity measurement step after the rigid registration.

In addition, the similarity for each section is obtained by dividing the deformation map generated by applying deformation to the bone image of the subject for each section, and then obtaining the average direction and the average amount of deformation of the deformation vector for each section, respectively.

Finally, bone age assignment step S700 allocates bone age using similarity values measured in the similarity measurement step S600 and a classifier trained in advance. As the classifier, any classifier that can be trained in advance may be used. For example, any one selected from the group consisting of a neural network, a support vector machine, and a linear classifier is preferable.

Here, the pre-training process is performed using a case extracted partially from the standard bone image database, and the addition of the case enables continuous training and performance improvement. Therefore, when the similarity measurement after the rigid registration, the total similarity after the non-rigid registration and the similarity value for each section are measured for each bone region of the hand bone, the optimal bone age is assigned by the classifier by inputting the measured value to the trained classifier.

According to the bone age evaluation method using the atlas matching method of the present invention as described above, by using a standard image database of each region of the hand bone and similarity measurement by image registration provides a method that can objectively and reproducibly evaluate bone age do.

In addition, it is possible to exclude the dependence of readers in the conventional subjective bone age assessment method of the doctor can be used to evaluate the standard bone age anywhere in the country, the accuracy and reading guidelines are regularly updated and supplied standard bone age image database There is an effect that can be kept up to date by.

In addition, since the computer can perform the process of the conventional atlas matching method has a low rejection and has the advantage that can be settled in the clinic within a short time, and when combined with the digital X-ray imaging apparatus of the present invention can be immediately after shooting It can be used as an intelligent X-ray imaging device dedicated to bone age measurement that can measure bone age.

Claims (6)

In the bone age evaluation method using the atlas matching method, A registration step of extracting a standard bone image for each bone age and a standard bone image for each region of the hand bone image and then registering the same in a database; An extraction step of extracting a bone image for each region of the hand bone from a digital X-ray image of the hand bone of the subject; An image registration step of measuring a difference between the extracted bone image of each region of the hand bone and a standard bone image of each region registered in the database through computer image processing; A similarity measuring step of measuring similarity between the bone image of each region of the image-matched subject and the standard bone image of the region; And And an allocating step of allocating bone age using the measured similarity and a pre-trained classifier. The method of claim 1, The registration step and extraction step, bone age evaluation method, characterized in that for extracting one or more bone images selected from the group consisting of resin bone, hand bone, ulna and radius. The method according to claim 1 or 2, The image matching step is bone age evaluation method, characterized in that performed by a combination of a rigid registration step, a non-rigid registration step. The method of claim 3, wherein The similarity measuring step is bone age evaluation method characterized in that the combination of the similarity measurement step after the rigid registration step, the similarity measurement step after the non-rigid registration step. The method of claim 4, wherein The similarity measuring step after the non-rigid registration step is a bone age evaluation method, characterized in that it comprises a total similarity measuring step or a similarity measuring step for each section. The method of claim 1, The classifier is a bone age evaluation method, characterized in that any one selected from the group consisting of a neural network, a support vector machine and a linear classifier.

Images