TWI428843B - System and method for determining the scoliosis - Google Patents

Description

Spine scoliosis determination system and determination method

The invention relates to a scoliosis determination system and a determination method, which have the advantages and functions of providing a neural network to quickly predict classification, computer-aided calculation to improve accuracy, and simultaneously provide scoliosis data and suggested treatment methods.

At present, most medical institutions use it to determine whether a patient has a scoliosis. Most of the doctors directly interpret the patient's X-ray film. In this way, if there are more doctors, sometimes there may be judgment errors, and each The doctor's judgment is different. Some may think that the patient's scoliosis is quite serious, and some may think it is okay. There are sometimes no standards for visual judgment, which may allow the patient to open the knife without surgery, or delay the patient. There is no auxiliary device for correcting or reducing the time for the doctor to accurately determine.

In addition, artificial neural networks (ANNs) are systems derived from biological neuron models that have the ability to process data. They mimic human thinking and simplify a complex problem. Simple configuration, combined with a large amount of data in parallel, learning and training capabilities, fault tolerance and fast, can solve a variety of problems, such as control, identification and prediction, and many professional research and high-tech products. It has been widely used in many publications and publicly available patents (for example, the Republic of China Patent Publication No. 200822903 "Smart Medical Execution System"). However, it has not been used to immediately determine whether a patient has a scoliosis after analysis through X-ray images.

Therefore, it is necessary to develop new technologies to solve the above shortcomings and problems.

The object of the present invention is to provide a scoliosis determination system and a determination method, which have the functions of providing a neural network to quickly predict classification, computer-aided calculation to improve accuracy, and simultaneously provide scoliosis data and suggested treatment methods. In particular, the problems to be solved by the present invention include: no problems such as a system that can quickly assist a doctor in determining a scoliosis condition and providing a suggestion.

The technical means for solving the above problems is to provide a scoliosis determination system and a determination method, the system part of which comprises: an image capture device for capturing a to-be-determined spinal image for a predetermined object; a vertebra having a plurality of head-to-tail joints; each of the vertebrae has a transverse protrusion extending from the head to the tail to the two sides, and each of the two transverse ends has a target, and a virtual axis between the two coordinates, and each The virtual axis has a virtual center point; and a reference line is positioned on the two virtual center points on the front and the last vertebrae; a central control unit is coupled to the image capturing device, and includes: a processing device Reading the to-be-determined spinal image, and connecting the virtual center points of the predetermined number of vertebrae to the reference line respectively to calculate the angle between them; a type of neural network device is pre-reading the complex array before being officially used. The patient's scoliosis data, and the neural network learning training program is performed to obtain the weighted value and partial weight formula of the neural network. When the neural network device is in formal use, the plurality of angles outputted by the processing device are read, and the weighting value of the neural network and the partial weight formula are substituted for calculation, thereby determining the spinal image to be determined. Classify and produce results, along with recommendations for improving lateral curvature.

Some of its methods include: Preparation steps; two. Capture image steps; three. Positioning step; four. Calculation step; five. Classification and suggestion steps; and six. Complete the steps.

The above objects and advantages of the present invention will be readily understood from the following detailed description of the preferred embodiments illustrated herein.

The invention will be described in detail in the following examples in conjunction with the drawings:

Referring to the first, second, third and fourth figures, the present invention is a scoliosis determination system and a determination method. The system portion includes: an image capture device 10 for aligning a predetermined object Taking the determined spinal image 91; the to-be-determined spinal image 91 has a plurality of vertebrae 911 that meet end to end; each of the vertebras 911 extends from the head to the tail to a lateral extent 912, respectively, to the two transverse protrusions 912 The ends have a standard L ( X _L , Y _L ), R ( X _R , Y _R ), and have a virtual axis D between the two coordinates L ( X _L , Y _L ), R ( X _R , Y _R ), And each virtual axis D has a virtual center point C ( X , Y ); and two virtual center points C ₁ ( X ₁ , Y ₁ ), C _n ( X _n , on the foremost and last vertebra 911 Y _n ) is located with a reference line L; a central control unit 20 is coupled to the image capturing device 10 and includes: a processing device 21 for reading the to-be-determined spinal image 91 and a predetermined number of vertebrae 911 virtual center point C (X, Y) are respectively connected to the reference line L is calculated respectively as shown in the included angle therebetween (e.g., fifth and sixth FIG, e.g. _{△ θ 1, △ θ 2,} △ θ 3); a neural network apparatus 22, before you use system, previously read scoliosis patient information of the plural groups of 9OA been determined, and neural network learning and training Program operations (see eighth and ninth figures) to obtain weighted values and partial weight formulas for such neural networks; when such neural network devices 22 are in active use, they are read into the output of the processing device 21 The multiple angles (the eighth figure is △ θ ₁ , ..., Δθ ₂₄ as an example, which can actually be input to the Δθ _n ), and the weighting value and the partial weight formula of the neural network are substituted. An operation is performed to classify the to-be-determined spinal image 91 and produce a result, while providing a suggestion that can improve the lateral bending condition.

In practice, the image capturing device 10 includes at least an X-ray illuminating portion 11 for illuminating an X-ray 111 with the predetermined object 90 (ie, the patient); and an image receiving portion 12 for receiving the predetermined object. The to-be-determined spinal image 91 generated by the X-ray 111 is irradiated to the central control unit 20.

The central control unit 20 further includes a display unit 23 for displaying the determination The result of the classification of the side curvature of the spinal image 91 to be determined.

The plurality of vertebrae 911 includes at least seven cervical vertebrae, twelve thoracic vertebrae and five lumbar vertebrae ( C ₁ ( X ₁ , Y ₁ ), . . . , C ₂₂ ( X ₂₂ , Y ₂₂ ), C ₂₃ ( X ₂₃ , Y ₂₃ ), C _n ( X _n , Y _n )).

As shown in the seventh, eighth and ninth figures, the part of the determination method of the present invention includes the following steps: 1. Preparing step 71: preparing an image capturing device 10 and a central control unit 20 as shown in the first, second, third and fourth figures; the central control unit 20 comprises a processing device 21 and a neural network device 22; two. Capture image step 72: the image capturing device 10 captures a to-be-determined spinal image 91 for a predetermined object 90; the to-be-determined spinal image 91 has a plurality of vertebrae 911 that meet end to end; each vertebra 911 A transverse protrusion 912 extends from the head to the tail to the two sides, and has a standard L ( X _L , Y _L ), R ( X _R , Y _R ) at the ends of the two transverse protrusions 912, and is at two coordinates L ( X _L , Y _L ), R ( X _R , Y _R ) has a virtual axis D, and each virtual axis D has a virtual center point C ( X , Y ); Positioning step 73: positioning a reference straight line L with two virtual center points C ₁ ( X ₁ , Y ₁ ), C _n ( X _n , Y _n ) on the foremost and last vertebra 911; Calculation step 74: the virtual center point C ( X , Y ) of the predetermined number of vertebrae 911 is connected to the reference line L by the processing device 21 to calculate an angle therebetween (see the fifth and sixth figures, for example, Δθ ₁ , Δθ ₂ , Δθ ₃ ); Classification and Suggestion Step 75: Before the formal use of the neural network device 22, the patient's scoliosis data 90A is read in advance and the neural network learning training program is performed (see the eighth and The ninth figure) is used to obtain the weighting value and the partial weight formula of the neural network; when the neural network device 22 is in formal use, the multiple angles output by the processing device 21 are read (eighth) The figure shows Δθ ₁ , ..., Δθ ₂₄ as an example. Actually, it can be input to the Δθ _n ), and the weighting value of the neural network and the partial weight formula are substituted for calculation. The spinal image 91 to be determined is classified and produces a result, and a suggestion scheme for improving the lateral bending condition is provided; Step 76 is completed: the determination of the spine image 91 to be determined for the predetermined object 90 is completed, and the side bend classification of the spine is determined to provide a suggestion scheme.

The embodiments of the related apparatus or structure and the part of the simple replacement structure in the present determination method are detailed in the system section, and should be directly referred to.

The mode of operation of the present invention is as follows: Before formal use, it is necessary to first analyze the scoliosis data of a plurality of identified patients (for example, the scoliosis data of one hundred patients shown in Table 1; The degree of side curve, the classification 2 represents moderate side curve, and the classification 3 represents severe side curve. The input type neural network device 22 performs learning operations through the neural network learning training program (as shown in the ninth figure):

After that, the weighted value and partial weight formula of the neural network can be obtained (the principle of such a neural network and the weighted value and the partial weight formula are well-known techniques, and will not be described). When it is officially used, it can be calculated and classified by simply entering the new patient's spine data.

Assuming that it is now officially used, the image capturing device 10 first captures the to-be-determined spinal image 91 of a new patient (ie, the predetermined object 90); and uses the two virtual center points C ₁ ( X ₁ on the foremost and last vertebra 911). , Y ₁ ), C _n ( X _n , Y _n ), a reference straight line L is positioned; the determined spinal image 91 is output to the processing device 21 of the central control unit 20; and the predetermined number of vertebrae are processed by the processing device 21 The virtual center point C ( X , Y ) of 911 is connected to the reference line L, respectively, and the angle between them is calculated: Δθ ₁ (assumed to be +8 degrees), Δθ ₂ (assumed to be +11 degrees), Δθ ₃ ( The assumption is -21 degrees). . . . Δθ _n (assumed to be -16 degrees) (Refer to the fifth and sixth figures, assuming that Δθ ₁ is the fourth lumbar vertebrae C ₁ ( X ₁ , Y ₁ ) (the two coordinates at both ends are L ₁ ( X _{L 1} , Y _{L 1} ) and R ₁ ( X _{R 1} , Y _{R 1} )), Δθ ₂ are the third lumbar vertebrae C ₂ ( X ₂ , Y ₂ ) (the two ends of which are coordinates L ₂ ( X _{L 2 )} , Y _{L 2} ) and R ₂ ( X _{R 2} , Y _{R 2} )), Δθ ₃ are the second lumbar vertebrae C ₃ ( X ₃ , Y ₃ ) (the two ends of which are coordinates L ₃ ( X _{L 3} , Y _{L 3} ) and R ₃ ( X _{R 3} , Y _{R 3} )).

After the plurality of angles calculated by the processing device 21 are input into the neural network weighting value and the partial weight formula, the side bending condition classification of the to-be-determined spinal image 91 can be determined (for example, classified into classification 1, classification 2, or classification). c) and provide a proposal for an improved classification of the side bend condition. E.g:

Category 1: Pay attention to posture correction and exercise therapy, and check once every six months.

Category 2: The treatment and exercise therapy should be used for the correction of the back frame. Exercise therapy can delay the deterioration of the scoliosis, which usually takes 2 to 3 years.

Category 3: These patients are prone to accelerate deformation and affect cardiopulmonary function, so surgery is recommended.

The foregoing is only an example, when the neural network device 22 of the present invention is used in different places (for example, from a health clinic to a teaching hospital), and it is assumed that the data of the patients with the scoliosis are read and the neural network learning is performed. After the training program is operated, another group of neural network weighting values and partial weight formulas may be obtained. At this time, the new patient's spine data is input into the operation, and the classification may be different, and the actual settings are used.

The advantages and effects of the present invention can be summarized as follows:

[1] A neural network is available to quickly predict classification. The invention provides a neural network device-like neural network learning training program and a neural network weighting value and a partial weight formula, which can quickly treat the determined spinal image for predictive classification operation, and greatly reduce the time for the doctor to interpret the X-ray film. .

[2] Computer-aided computing improves accuracy. The invention uses a simple X-ray device After taking the patient's chest X-ray film and inputting the X-ray film data into the central control unit, the computer can accurately calculate the auxiliary information of the patient's scoliosis, so that the doctor can quickly and accurately determine the spine state state.

[3] Simultaneously provide scoliosis data and recommended treatment methods. The invention simultaneously provides the patient's scoliosis data and suggested treatment methods, and assists the doctor to quickly determine the patient's spinal condition and treatment.

The present invention has been described in detail with reference to the preferred embodiments of the present invention, without departing from the spirit and scope of the invention.

From the above detailed description, those skilled in the art can understand that the present invention can achieve the foregoing objects, and has been in compliance with the provisions of the patent law, and has filed an application for an invention patent.

10‧‧‧Image capture device

11‧‧‧X-ray irradiation department

111‧‧‧X-ray

12‧‧‧Image Receiving Department

20‧‧‧Central Control Department

21‧‧‧Processing device

22‧‧‧ class neural network device

23‧‧‧ Display Department

71‧‧‧Preparation steps

72‧‧‧ Capture image steps

73‧‧‧ Positioning steps

74‧‧‧ Calculation steps

75‧‧‧Classification and recommended steps

76‧‧‧Complete steps

90‧‧‧Predetermined objects

90A‧‧‧Spine data

91‧‧‧Determined spinal image

911‧‧‧ vertebrae

912‧‧‧ transverse

D‧‧‧ virtual axis

△θ ₁ , Δθ ₂ , Δθ ₃ , Δθ ₂₄ ‧‧‧ angle

L‧‧‧ benchmark line

C ( X , Y ), C ₁ ( X ₁ , Y ₁ ), C ₂ ( X ₂ , Y ₂ ), C ₃ ( X ₃ , Y ₃ ), C ₂₂ ( X ₂₂ , Y ₂₂ ), C ₂₃ ( X ₂₃ , Y ₂₃ ), C _n ( X _n , Y _n )‧‧‧ virtual center point

L ( X _L , Y _L ), L ₁ ( X _{L 1} , Y _{L 1} ), L ₂ ( X _{L 2} , Y _{L 2} ), L ₃ ( X _{L 3} , Y _{L 3} ), R ( X _R , Y _R ), R ₁ ( X _{R 1} , Y _{R 1} ), R ₂ ( X _{R 2} , Y _{R 2} ), R ₃ ( X _{R 3} , Y _{R 3} ) ‧‧‧ coordinates

The first figure is a schematic diagram of the practical application of the present invention

The second figure is a block diagram of the structure of the present invention.

The third figure is a reference diagram of a neural network device such as the present invention.

The fourth picture is a schematic diagram of the normal spine

The fifth picture is a schematic diagram of the scoliosis

The sixth figure is a partial enlarged view of the fifth figure

The seventh figure is a flow chart of the determination method of the present invention

The eighth figure is a flow chart of the operation process of the present invention.

The ninth figure is a flow chart of the part of the process of the eighth figure

10. . . Image capture device

11. . . X-ray irradiation department

111. . . X-ray

12. . . Image receiving unit

20. . . Central control department

twenty one. . . Processing device

twenty two. . . Neural network device

twenty three. . . Display department

90. . . Scheduled object

91. . . Spine image to be determined

Claims (4)

A scoliosis judging system includes: an image capturing device for extracting a spinal image to be determined from a predetermined object; the spinal image to be determined having a plurality of vertebrae that meet end to end; each vertebra Each has a transverse protrusion extending from the head to the tail to the two sides, and has a mark at the ends of the two transverse protrusions, has a virtual axis between the two coordinates, and has a virtual center point on each virtual axis; and the front and the last The two virtual center points on the vertebrae are positioned to define a reference line; a central control unit is coupled to the image capturing device, and includes: a processing device for reading the to-be-determined spinal image and a predetermined number of vertebrae The virtual center points are respectively connected with the reference line to calculate the angle between them; a type of neural network device is pre-reading the scoliosis data of the patient who has been identified in the complex array before the formal use, and performing neural network learning. Training program operation to obtain weighted value and partial weight formula of the neural network; when the neural network device is officially used, the processing is read in A plurality of output angle is set, the weighting value is substituted into the formula weight value and the biasing class of neural network for computation, and thus can be determined that the spine and generating image classification results can be improved while providing proposals of scoliosis condition. The scoliosis determination system of claim 1, wherein the image capturing device comprises at least: an X-ray illuminating portion for illuminating the predetermined object with X-rays; and an image receiving portion for Receiving the predetermined object to illuminate the X-ray to be determined to be determined by the X-ray, and outputting the image to the central control unit; the central control unit further includes a display portion for displaying a result of determining the side-cut classification of the to-be-determined spinal image; The plurality of vertebrae includes at least seven cervical vertebrae, twelve thoracic vertebrae and five lumbar vertebrae. A method for determining a scoliosis includes the following steps: Preparation steps: preparing an image capturing device and a central control unit; The central control unit includes a processing device and a type of neural network device; Step of capturing an image: the image capturing device captures a to-be-determined spinal image for a predetermined object; the to-be-determined spinal image has a plurality of vertebrae that meet end to end; each vertebra is from the head to the tail Extending a transverse process respectively, having a mark at the ends of the two transverse processes, having a virtual axis between the two coordinates, and having a virtual center point on each virtual axis; Positioning step: positioning a reference line with two virtual center points on the foremost and last vertebrae; Calculating step: using the processing device to respectively connect a virtual center point of a predetermined number of vertebrae with a reference line to calculate an angle therebetween; Classification and suggestion steps: Before the formal use of the neural network device, the pre-reading of the scoliosis data of the patient in the complex array is performed, and the neural network learning training program is performed to obtain the neural network. The weighted value and the partial weight formula; when the neural network device is officially used, the multiple angles output by the processing device are read, and the weighting value and the partial weight formula of the neural network are substituted for the operation. In turn, the spinal image to be determined can be classified and the result is produced, and a suggestion scheme for improving the lateral bending condition is provided; Completion steps: completing the image of the spine to be determined for the predetermined object, and determining the side bend classification of the spine to provide a suggestion scheme. The method for judging the scoliosis according to the third aspect of the invention, wherein the image capturing device comprises at least: an X-ray illuminating portion for illuminating the predetermined object with X-rays; and an image receiving portion for Receiving the predetermined object to illuminate the X-ray to be determined to be determined by the X-ray, and outputting the image to the central control unit; the central control unit further includes a display portion for displaying a result of determining the side-cut classification of the to-be-determined spinal image; The plurality of vertebrae includes at least seven cervical vertebrae, twelve thoracic vertebrae and five lumbar vertebrae.