CN115081476A - Method and device for bone recognition for bone tissue grinding - Google Patents
Method and device for bone recognition for bone tissue grinding Download PDFInfo
- Publication number
- CN115081476A CN115081476A CN202210668751.5A CN202210668751A CN115081476A CN 115081476 A CN115081476 A CN 115081476A CN 202210668751 A CN202210668751 A CN 202210668751A CN 115081476 A CN115081476 A CN 115081476A
- Authority
- CN
- China
- Prior art keywords
- grinding
- grinding force
- force signal
- bone
- signal
- 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
- 238000000227 grinding Methods 0.000 title claims abstract description 248
- 210000000988 bone and bone Anatomy 0.000 title claims abstract description 147
- 238000000034 method Methods 0.000 title claims abstract description 81
- 238000003062 neural network model Methods 0.000 claims abstract description 28
- 230000009467 reduction Effects 0.000 claims abstract description 17
- 230000001054 cortical effect Effects 0.000 claims abstract description 16
- 230000002829 reductive effect Effects 0.000 claims abstract description 15
- 230000009466 transformation Effects 0.000 claims abstract description 8
- 239000010410 layer Substances 0.000 claims description 41
- 230000006870 function Effects 0.000 claims description 35
- 230000008569 process Effects 0.000 claims description 33
- 238000001914 filtration Methods 0.000 claims description 15
- 238000007781 pre-processing Methods 0.000 claims description 12
- 239000002356 single layer Substances 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 14
- 239000000126 substance Substances 0.000 description 12
- 238000013528 artificial neural network Methods 0.000 description 10
- 238000012549 training Methods 0.000 description 5
- 238000000354 decomposition reaction Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000338 in vitro Methods 0.000 description 2
- 238000002684 laminectomy Methods 0.000 description 2
- 230000000670 limiting effect Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000001356 surgical procedure Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000006837 decompression Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000001747 exhibiting effect Effects 0.000 description 1
- 235000015220 hamburgers Nutrition 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000003801 milling Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000036961 partial effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B90/00—Instruments, implements or accessories specially adapted for surgery or diagnosis and not covered by any of the groups A61B1/00 - A61B50/00, e.g. for luxation treatment or for protecting wound edges
- A61B90/08—Accessories or related features not otherwise provided for
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Surgery (AREA)
- Biomedical Technology (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Pathology (AREA)
- Artificial Intelligence (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
The invention provides a method and a device for identifying bone for grinding bone tissue, wherein the method comprises the following steps: acquiring a grinding force signal and a related signal in real time when the bone tissue is ground, wherein the related signal comprises at least one of grinding speed, grinding depth and ultrasonic knife power; performing wavelet transformation noise reduction on the grinding force signal, and then extracting the noise-reduced grinding force signal in the vertical direction as a target grinding force signal; and inputting the target grinding force signal and the related signal into a trained BP neural network model for bone identification to determine whether the inner cortical bone is accessed. The method provided by the invention solves the problem that the bone identification accuracy is not high during grinding in the prior art.
Description
Technical Field
The invention relates to the technical field of bone tissue grinding, in particular to a method and a device for identifying bone for bone tissue grinding.
Background
With the rapid development of science and technology and medical science, more and more computer-assisted medical technologies and devices are applied to the medical field. For example, in a surgical operation, the computer-assisted medical equipment can replace a doctor to perform the operation to a certain extent, so that the operation failure caused by operation fatigue of the doctor due to a long-time operation is avoided, and the operation risk is reduced.
The vertebral plate decompression surgery in the existing surgical operation mostly utilizes a cervical vertebra grinding robot to assist the surgery, and the cervical vertebra grinding robot can plan a grinding path and detect the grinding depth position, for example, Chinese patent application with application publication number CN106725711A discloses a bone grinding robot, a vertebral plate grinding robot control system and a method. However, there are many high-frequency noises during force signal acquisition, and these noises easily cause misjudgment of the acquired force signal when compared with a preset value, which results in inaccurate bone identification during grinding, and further fails to ensure the safety of the operation.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the technical problems in the related art.
Therefore, the first purpose of the present invention is to provide a method for bone identification for bone tissue grinding, so as to solve the problem of low accuracy of bone identification during grinding in the prior art.
A second object of the invention is to provide a device for bone recognition for bone tissue grinding.
A third object of the invention is to propose an electronic device.
A fourth object of the present invention is to propose a non-transitory computer-readable storage medium.
In order to achieve the above object, a first aspect of the present invention provides a method for bone identification for bone tissue grinding, including: acquiring a grinding force signal and a related signal in real time when the bone tissue is ground, wherein the related signal comprises at least one of grinding speed, grinding depth and ultrasonic knife power; performing wavelet transformation noise reduction on the grinding force signal, and then extracting the noise-reduced grinding force signal in the vertical direction as a target grinding force signal; and inputting the target grinding force signal and the correlation signal into a trained BP neural network model for bone identification to determine whether to enter inner cortical bone.
According to the method provided by the embodiment of the invention, the influence of the grinding depth on the vertical force during grinding is considered to be small, and the peak value change trend can obviously reflect the force characteristics of different bone substances, so that the grinding force in the vertical direction is collected, the vertical grinding force is used as input data of the model, and the BP neural network model is used for bone substance identification, so that the bone substance identification can be more accurately carried out, and the problem of low accuracy of the bone substance identification during grinding in the prior art is solved.
A method for bone identification for bone tissue grinding according to an embodiment of the first aspect of the present invention further comprises said wavelet transform de-noising comprises decomposing said grinding force signal using wavelet basis functions to obtain an approximation component and a detail component; selecting a threshold value, and then filtering the detail component by using a soft threshold value function; the approximation component and the filtered detail component are reconstructed to obtain a noise reduced grinding force signal. Thus, signal spikes and abrupt signals can be better protected.
A method for bone identification for bone tissue grinding in accordance with an embodiment of the first aspect of the present invention further includes using Daubechies series as wavelet basis functions. This makes it possible to make the subsequently reconstructed signal smoother.
A method for bone identification for bone tissue grinding according to an embodiment of the first aspect of the present invention further includes preprocessing a grinding force signal after noise reduction in a vertical direction, where the preprocessing includes: sequencing all the grinding force signals after noise reduction of the single-layer bone tissue; respectively filtering data in a preset proportion before and after the sequence; and calculating the average value of the residual data as a target grinding force signal of the layer of bone tissue. Thus, the fluctuation of the force signal generated by the disturbance outside the system during the grinding process can be removed.
The method for identifying bone in bone tissue grinding according to the embodiment of the first aspect of the present invention further includes normalizing the target grinding force signal and the correlation signal before inputting the target grinding force signal and the correlation signal into the trained BP neural network model. Thus, the order of magnitude difference between various data can be avoided.
In order to achieve the above object, a second embodiment of the present invention provides an apparatus for bone identification for bone tissue grinding, comprising: the acquisition module is used for acquiring a grinding force signal and a related signal in the process of grinding bone tissues in real time, wherein the related signal comprises at least one of grinding speed, grinding depth and ultrasonic knife power; the de-noising module is used for receiving the grinding force signal output by the acquisition module, performing wavelet transformation de-noising on the grinding force signal, and extracting the de-noised grinding force signal in the vertical direction as a target grinding force signal; and the identification module is used for receiving the target grinding force signal output by the denoising module, receiving the correlation signal output by the acquisition module, and inputting the target grinding force signal and the correlation signal into a trained BP neural network model for bone identification to determine whether the bone enters the inner cortical bone.
According to the device provided by the embodiment of the invention, the influence of the grinding depth on the vertical force during grinding is considered to be small, and the peak value change trend can obviously reflect the force characteristics of different bone substances, so that the grinding force in the vertical direction is acquired through the acquisition module, the vertical grinding force is used as input data of the model in the identification module, and the bone substance identification is carried out by using the BP neural network model, so that the bone substance identification can be carried out more accurately, and the problem of low accuracy of the bone substance identification during grinding in the prior art is solved.
An apparatus for bone identification for bone tissue grinding according to an embodiment of the second aspect of the present invention is further included in the denoising module, and the wavelet transform denoising includes: decomposing the grinding force signal by using Daubechies series wavelet basis functions to obtain an approximate component and a detail component; selecting a threshold value, and then filtering the detail component by using a soft threshold value function; the approximation component and the filtered detail component are reconstructed to obtain a noise reduced grinding force signal. Thereby, signal spikes and abrupt signals can be better protected and subsequently reconstructed signals can be made smoother.
The device for identifying bone in bone tissue grinding according to the embodiment of the second aspect of the invention further comprises a denoising module, wherein the preprocessing is required to be performed on the grinding force signal after the noise reduction in the vertical direction, and the preprocessing comprises the following steps: sequencing all the grinding force signals after noise reduction of the single-layer bone tissue; respectively filtering data of a preset proportion before and after the sequence; and calculating the average value of the residual data as a target grinding force signal of the layer of bone tissue. Thus, the fluctuation of the force signal generated by the disturbance outside the system during the grinding process can be removed.
To achieve the above object, a third aspect of the present invention provides an electronic device, including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method for bone identification for bone tissue grinding as embodied in the first aspect of the invention.
In order to achieve the above object, a fourth aspect of the present invention provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform a method for bone identification for bone tissue grinding according to the first aspect of the present invention.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The foregoing and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is a schematic flow chart of a method for identifying bone material for bone tissue grinding according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a wavelet transform denoising method according to an embodiment of the present invention;
FIG. 3 is a signal spectrum diagram before and after wavelet transform denoising provided by the embodiment of the present invention;
FIG. 4(a) is a time domain diagram of a grinding force signal in a vertical direction before and after wavelet transform denoising according to an embodiment of the present invention;
FIG. 4(b) is a time domain diagram of a grinding force signal in a horizontal direction before and after wavelet transform denoising according to an embodiment of the present invention;
FIG. 4(c) is a time domain diagram of a grinding force signal in a vertical feed direction before and after wavelet transform denoising according to an embodiment of the present invention;
FIG. 5 is a partial enlarged view of FIG. 4 (a);
FIG. 6 is a flow chart illustrating a method for preprocessing a grinding force signal according to an embodiment of the present invention;
FIG. 7 is a BP neural network model provided by an embodiment of the present invention;
FIG. 8 is a schematic illustration showing the results of an ex vivo bone laminectomy trial provided by an embodiment of the present invention;
fig. 9 is a schematic structural diagram of an apparatus for bone identification for bone tissue grinding according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention. Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing steps of a custom logic function or process, and alternate implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present invention.
A method and apparatus for bone identification for bone tissue grinding according to an embodiment of the present invention will be described with reference to the accompanying drawings.
Fig. 1 is a schematic flow chart of a method for identifying bone material for bone tissue grinding according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a wavelet transform denoising method according to an embodiment of the present invention.
The embodiment of the invention provides a bone identification method for bone tissue grinding, which aims to solve the problem of low bone identification accuracy in grinding in the prior art, and as shown in fig. 1, the bone identification method for bone tissue grinding (which may be referred to as a bone identification method for short) comprises the following steps:
and step S10, acquiring a grinding force signal and a related signal in real time when the bone tissue is ground, wherein the related signal comprises at least one of grinding speed, grinding depth and ultrasonic blade power.
In the present embodiment, the grinding force signal in step S10 is acquired by a force sensor. Specifically, a grinding force signal fed back when the ultrasonic osteotome grinds the vertebral plate is collected through a six-axis force sensor and then transmitted to an upper computer through a collection card. Wherein, the frequency of the grinding force signal is 0-100Hz, and the sampling frequency of the acquisition card is 2 kHz.
In the present embodiment, the related signal in step S10 refers to factors that affect the acquired grinding force signal, such as grinding speed, grinding depth, ultrasonic blade power, and the like.
And step S11, performing wavelet transformation noise reduction on the grinding force signal, and extracting the grinding force signal subjected to noise reduction in the vertical direction as a target grinding force signal.
In step S11, in consideration of the difference between the sampling frequency of the acquisition card and the frequency of the grinding force signal, the original grinding force signal transmitted by the acquisition card has high-frequency noise, which affects the recognition accuracy of the subsequent model, and therefore, the acquired original grinding force signal needs to be filtered. In the filtering process, the useless noise frequency band can be effectively extracted by considering the wavelet transform for processing, the influence on the time domain characteristics of effective signals is small, and signal peaks and abrupt signals can be well protected. Therefore, the present embodiment performs noise reduction processing on the original grinding force signal by using a noise reduction method based on wavelet transform threshold processing (wavelet transform noise reduction for short).
In the present embodiment, as shown in fig. 2, the wavelet transform denoising step in step S11 includes decomposing the grinding force signal using wavelet basis functions to obtain an approximate component and a detail component (step S110); selecting a threshold value, and then filtering the detail component by using a soft threshold value function (step S111); the approximation component and the filtered detail component are reconstructed to obtain a noise-reduced grinding force signal (step S112).
In the present embodiment, the wavelet basis function in step S110 adopts Daubechies series wavelet, which is abbreviated as "dbN" wavelet system, and for example, "db 4" may be used as the wavelet function. Due to the fact that the series of wavelets have good regularity, subsequently reconstructed signals can be smoother. The approximate component in step S110 refers to a component with a lower frequency, which is a portion that needs to be reserved; the detail component is a high-frequency component and is a portion to be filtered out.
In the present embodiment, step S110 considers the nyquist sampling theorem (i.e. the sampling frequency is at least 2 times of the signal frequency), and multiple high-frequency noise signals are superimposed in the signal due to line interference, so that the original grinding force signal is subjected to multi-layer wavelet decomposition to filter out the required effective frequency band.
In this embodiment, the threshold selection in step S111 needs to be performedTo utilize F n ＝f n +e n In the formula F n For mixed signals of length N, f n And e n Respectively representing the required grinding force signal and a Gaussian white noise signal which follows standard normal distribution and is obtained by mixing a signal F n A threshold value of a wavelet domain capable of removing noise is evaluated. Specifically, the present embodiment selects Stein unbiased risk threshold (SURE) as the threshold selection method: f is to be n Taking an absolute value of each element F in the signal, sorting the absolute values from large to small, and then squaring each element to obtain a new element sequence, namely: s (i) ═ sort (| F |) 2 If the threshold is taken as the square root of the i-th element of s (i), the risk of the threshold is:
according to the obtained risk curve Risk (k), the value corresponding to the minimum risk point is recorded as k min Then the rigrsure threshold is defined as
In the present embodiment, after the threshold of the white gaussian noise in the wavelet domain is determined in step S111, the detail component after wavelet decomposition in step S110 is filtered by using a soft threshold function. Therefore, the signal can be smoother, local jitter of the signal caused by denoising by using a hard threshold function is avoided, and then the step S112 is performed to reconstruct the approximate component and the filtered detail component to obtain the grinding force signal after denoising.
In the embodiment, the number of layers of the wavelet decomposition is continuously adjusted, and the frequency domain characteristics of the signal are analyzed. When the number of wavelet decomposition layers is 5, noise in a frequency band substantially outside 0-100Hz can be effectively filtered. Fig. 3 is a signal spectrum diagram before and after wavelet transform denoising according to an embodiment of the present invention, where the abscissa is frequency, the ordinate is amplitude, a blue signal is a signal wave before wavelet transform denoising, and an orange signal is a signal wave after wavelet transform denoising, as shown in fig. 3, noise greater than 100Hz after wavelet transform denoising is almost filtered out.
In step S11, a grinding force signal in the vertical direction is extracted from the grinding force signal after noise reduction in consideration of the characteristics of the grinding force signal after noise reduction.
Specifically, taking the acquisition and filtering of grinding force signals in the vertical direction, the horizontal direction and the vertical feeding direction of a group of grinding processes when the ultrasonic knife grinds the vertebral plate as an example, time domain graphs of the grinding force signals before and after the wavelet transformation denoising shown in fig. 4(a) -4 (c) are obtained. The vertical direction is a direction downward along the grinding tool axis, i.e. a direction along the grinding tool axis toward the grinding area, the horizontal direction is a feeding direction, the feeding direction is perpendicular to the vertical direction, and the vertical feeding direction is perpendicular to the feeding direction and perpendicular to the vertical direction.
FIG. 4(a) is a time domain diagram of a grinding force signal in a vertical direction before and after wavelet transform denoising according to an embodiment of the present invention; FIG. 4(b) is a time domain diagram of a grinding force signal in a horizontal direction before and after wavelet transform denoising according to an embodiment of the present invention; fig. 4(c) is a time domain diagram of a grinding force signal in a vertical feeding direction before and after wavelet transform denoising according to the embodiment of the present invention. The abscissa of fig. 4(a) -4 (c) is time, and the ordinate is the pressure value (i.e., grinding Force value) of the grinding Force signal in the corresponding direction, wherein the green waveform is the Original grinding Force signal (Original Force), and the blue waveform is the filtered (i.e., noise-reduced) grinding Force signal (Wavelet filter), and it can be seen by comparing the green and blue signal waveforms in fig. 4(a) -4 (c), the filtered signal characteristics are completely retained, and the noise signal mixed in the Original grinding Force signal is filtered.
Based on the force value time domain diagram in the vertical direction shown in fig. 4(a), it can be seen that the force in the direction is less affected by the grinding depth, and the variation trend of the peak value can obviously reflect the force characteristics of different bone substances:
1) at 0s-9.8s, the grinding force values are relatively high, and it can be concluded from the structure of the vertebral plates of the spine and the grinding process that the grinding zone is in the Outer cortical bone part (Outer clinical FontWeight) during this period.
2) At 9.8s-13.5s and 27.7s-31.7s, the sharpening value is in the Transition region of greater and lesser values, which represents the ultrasonic blade bit at the interface of the outer cortical and cancellous bone (Transition region), thus exhibiting a force characteristic of transitional nature.
3) At 13.5s-27.7s, the grinding force value is relatively small and it can be concluded that the area ground during this time period is in the Cancellous bone portion (canalous bone layer).
4) At 31.7s-40s, the grinding force value again becomes a larger value, indicating that the area being ground at this time is in the Inner cortical bone portion (Inner cortical bone layer).
Based on the time domain graph of the force value in the horizontal direction shown in fig. 4(b), it can be seen that the force in the direction is obviously affected by the grinding depth, the peak value of the overall force has a tendency of rising layer by layer, and there is no obvious characteristic difference between different bone substances. The force value time domain graph in the vertical feeding direction shown in fig. 4(c) has no obvious characteristics and cannot be used for bone identification.
Amplifying and displaying a single-layer grinding force signal (red circle part) in a force value time domain diagram in the vertical direction shown in fig. 4(a) to obtain fig. 5, wherein fig. 5 shows an original grinding force signal (green wave) and a wavelet-transformed noise-reduced signal (blue wave) in single-layer grinding, and the whole Process of sinking-feeding-back grinding, wherein a peak value appearing from 0s to 1.1s represents the grinding force of the sinking Process (Downward Process); the highest peak occurring at 1.1s-2.45s represents the grinding force of the feed Process (Forward Process); the lowest peak occurring at 2.45s-3.6s represents the grinding force of the backgrinding Process (backing Process).
Through the analysis of the force value time domain diagrams in all directions, the grinding force signals in the vertical direction can more accurately show that the grinding force values of different bones are different in information. Therefore, in step S11, a grinding force signal in the vertical direction is extracted from the noise-reduced grinding force signal for subsequent processing.
Fig. 6 is a flowchart illustrating a method for preprocessing a grinding force signal according to an embodiment of the present invention. In step S11, the noise-reduced grinding force signal in the vertical direction is obtained and then pre-processed to obtain a target grinding force signal. As shown in fig. 6, the step of preprocessing includes: sequencing all the noise-reduced grinding force signals of the single-layer bone tissue (step S113); respectively filtering data in a preset proportion before and after the sequence (step S114); the average value of the remaining data is calculated as a target grinding force signal for the layer of bone tissue (step S115). Thus, the fluctuation of the force signal generated by the disturbance outside the system during the grinding process can be removed. All the grinding force signals in the step S113 are grinding force signals in the vertical direction in the feeding process, and the preset ratio in the step S114 is, for example, 20%, so that fluctuation of force signals caused by interference outside the system in the grinding process is removed by filtering out 20% of data before and after, for example, an ultrasonic tool bit touches the spinous process, a sensor acquires a maximum and minimum value, and an operator touches the spinous process by mistake in the grinding process.
And step S12, inputting the target grinding force signal and the related signal into a trained BP neural network model for bone identification to determine whether the inner cortical bone is entered.
In step S12, considering that the grinding force variation corresponding to different bones has regularity, which is hard to be directly defined by a general mathematical formula, and a model corresponding to the regularity needs to be obtained through training and verification of a large amount of experimental data, the present embodiment selects a neural network model as the recognition model, where the neural network model may select a BP neural network model.
First, a BP neural network model will be described. The input of the BP neural network model is a target grinding force signal and a related signal, and the output is bone information. The BP neural network model is divided into three layers: an input layer, a hidden layer, and an output layer. The BP neural network model output layer is provided with one output element, the input layer is provided with four input elements, and the hidden layer is
In the formula, h is the number of hidden layers, m is the number of input layers, n is the number of output layers, j is an arbitrary value between 1 and 10, three neurons in the hidden layers are determined according to the idea of minimizing the model, and finally the BP neural network model shown in FIG. 7 is constructed. Wherein X in the model shown in FIG. 7 1 ～X 4 As a model of a neural networkRespectively corresponding to the grinding speed, the grinding force characteristic value (i.e. the target grinding force signal), the grinding depth and the ultrasonic knife power, H 1 ～H 3 For the median value of the hidden layer, Y is the output value, which can be set to 0 and 1 for the purpose of determining whether the current grinding layer is lamino-cortical bone, based on model identification. v. of 1h ～v 4h As weights of the input layer to the hidden layer, w 11 ～w 31 The weights from the hidden layer to the output layer. In addition, considering that the output value Y of the output layer is not necessarily 0 or 1 of an integer during training, a section function is set to divide the interval [0,0.5 ]]And interval (0.5, 1)]The output value of (a) is made 0 or 1, the piecewise function is:
wherein f (Y) is the output value of the piecewise function, and Y is the input value of the piecewise function, wherein the input value of the piecewise function is the output value of the output layer of the neural network.
Based on the constructed model, the BP neural network model is trained by using a data set, wherein the data set comprises a target grinding force signal and a related signal which are collected in advance, and a corresponding label (namely corresponding bone layer information), and the data set can be obtained through experiments. In consideration of the fact that the target grinding force signal and the related signal in the data set are independent from each other and the difference of the magnitude of each signal may occur, the target grinding force signal and the related signal are normalized before being input into the BP neural network model. In this embodiment, the normalization process adopts a maximum-minimum method, and the maximum-minimum method is:
in the formula, x k For the target grinding force signal or associated signal, X, in the data set k For the corresponding normalized signal, x min For data sequences corresponding to signals of the same classMinimum value of (1), x max Is the maximum value of the data sequence corresponding to the class signal.
In the training process, considering that a large number of grinding force signals and associated signals such as grinding speed, grinding depth and ultrasonic knife power can be acquired by single-layer grinding operation in the process of grinding the vertebral plate by the ultrasonic knife, if all the grinding force signals and the associated signals are input into the neural network for training, the calculated amount can be greatly increased, and the bone recognition difficulty is improved.
In the training process, the one-time forward process of the BP neural network is as follows:
in the formula, a j Is a hidden layer threshold, b is an output layer threshold, σ is a neuron activation function, v ij Is the weight, x, of the ith input element to the jth intermediate value i Is the ith input element, m is the number of input elements, H is the number of intermediate values, H j Is the jth intermediate value, ω j The weight from the jth intermediate value to the output layer. The introduction of the activation function enables the model to adapt to nonlinear mapping, and considering that the number of layers of the neural network is small, the sigmoid function has good derivative property, and infinite signals can be mapped to (0,1), so that the activation function in the embodiment selects the sigmoid function, and the sigmoid function is as follows:
whereinX in sigmoid function
After a forward process is performed once, an error generated in a single process is calculated, in this embodiment, a mean square error function is selected as a loss function, where the mean square error function is:
where err represents the convergence error, y represents the output value of the neural network output layer,
representing the corresponding label.
After the convergence error is obtained, feeding the error value back to the neural network, and performing weight correction, namely a back propagation process of the BP neural network, wherein the weight is updated as follows:
v ij ＝v ij +ηH j (1-H j )x i ω j err
ω j ＝ω j +ηH j err
the threshold is updated as:
a j ＝a j +ηH j (1-H j )ω j err
b＝b+err
where η is the learning rate of the neural network. Setting the convergence error err of the neural network to 10 -4 The learning rate eta is 0.001, and v is repeatedly calculated in an iterative way along with the forward process and the backward propagation of the BP neural network ij And ω j And continuously changing until the optimal condition is met, thereby obtaining a trained BP neural network model.
In this embodiment, after obtaining the trained BP neural network model, in step S12, the target grinding force signal to be identified (i.e., the target grinding force signal processed in step S11) and the associated signal to be identified (i.e., the associated signal obtained in step S10 in real time) are input into the trained BP neural network model for bone identification to determine whether to enter the inner cortical bone.
In step S12, before the target grinding force signal to be recognized and the related signal are input to the trained BP neural network model, normalization processing is performed on the target grinding force signal and the related signal, respectively. Thus, the order of magnitude difference between various data can be avoided.
In the present embodiment, it is determined whether the grinding end point is reached (i.e., whether the lamino cortex bone is reached) based on the recognition result of step S12, and if so, the grinding is completed, and if not, the grinding is continued.
In some embodiments, considering that the vertebral plate of the spine is constructed in a "Pigsonia skin" hamburger structure, the inner cortical bone has a certain thickness and hardness, and according to the experience of the clinician, the thickness of the inner cortical bone of the vertebral plate is 1-2mm, and each grinding depth is 0.5 mm. Therefore, after the trained BP neural network model identifies the inner cortical bone, the vertebral plate is continuously ground for 2 layers, the vertebral plate can be conveniently removed after grinding is completed by a doctor while the vertebral plate is not worn, and therefore the grinding strategy can be formulated. FIG. 8 is a schematic diagram of the results of an ex vivo bone laminectomy trial provided by embodiments of the present invention. In order to verify the identification method, an in-vitro spinal bone grinding experiment is carried out, and after the in-vitro spinal bone grinding experiment is completed, the method is verified by visual observation and CT images (see figure 8) to achieve that a thin layer of cortical bone of the inner layer of the vertebral plate can be left, so that the requirement of grinding the vertebral plate is met.
In the method of the embodiment, considering that the vertical force is less affected by the grinding depth during grinding, and the peak value change trend can obviously reflect the force characteristics (namely grinding force value information) of different bones, the grinding force in the vertical direction is obtained, the grinding force in the vertical direction is used as the input data of the model, and the BP neural network model is used for identifying the bones so as to more accurately identify the bones, thereby solving the problem of low bone identification accuracy during grinding in the prior art, in addition, the grinding force signal in the vertical direction, the four-dimensional data of the grinding speed, the grinding depth and the ultrasonic knife power are comprehensively considered, avoiding the misjudgment caused by comparing the collected grinding force signal with the preset value, ensuring that the end point of the grinding operation can be accurately judged during grinding, improving the bone identification accuracy, and the method of the embodiment is suitable for bone identification during ultrasonic knife grinding vertebral plates, can realize the real-time, accurate and safe vertebral plate grinding task.
In order to realize the above embodiment, the present invention further provides a bone identification device for bone tissue grinding.
Fig. 9 is a schematic structural diagram of an apparatus for bone identification for bone tissue grinding according to an embodiment of the present invention.
As shown in fig. 9, the apparatus for bone recognition for bone tissue grinding (which may be simply referred to as a bone recognition apparatus) 1 includes an acquisition module 10, a denoising module 20, and a recognition module 30.
In the present embodiment, the acquisition module 10 is used for acquiring a grinding force signal and a related signal in real time when the bone tissue is ground, wherein the related signal comprises at least one of grinding speed, grinding depth and ultrasonic blade power. The denoising module 20 is configured to receive the grinding force signal output by the acquisition module 10, perform wavelet transformation denoising on the grinding force signal, and then extract the denoised grinding force signal in the vertical direction as a target grinding force signal. The recognition module 30 is configured to receive the target grinding force signal output by the denoising module 20 and the associated signal output by the acquisition module 10, and input the target grinding force signal and the associated signal into a trained BP neural network model for bone recognition to determine whether to enter the inner cortical bone.
The wavelet transform denoising performed by the denoising module 20 in this embodiment includes: decomposing the grinding force signal by using Daubechies series wavelet basis functions to obtain an approximate component and a detail component; selecting a threshold value, and then filtering the detail component by using a soft threshold value function; the approximation component and the filtered detail component are reconstructed to obtain a noise reduced grinding force signal. Thereby, signal spikes and abrupt signals can be better protected and subsequently reconstructed signals can be made smoother.
In this embodiment, the denoising module 20 needs to pre-process the denoising grinding force signals in the vertical direction, where the pre-processing includes sorting all the denoising grinding force signals of the single-layer bone tissue; respectively filtering data in a preset proportion before and after the sequence; and calculating the average value of the residual data as a target grinding force signal of the layer of bone tissue. Thus, the fluctuation of the force signal generated by the disturbance outside the system during the grinding process can be removed.
It should be noted that the foregoing explanation of the embodiment of the method for identifying bone substance for bone tissue grinding is also applicable to the device for identifying bone substance for bone tissue grinding of this embodiment, and will not be described again here.
In order to implement the above embodiments, the present invention further provides an electronic device, including: at least one processor; and a memory communicatively coupled to the at least one processor, the memory storing instructions executable by the at least one processor to cause the at least one processor to perform the bone identification method of the foregoing method embodiment of bone identification for bone tissue milling of the present invention.
In order to achieve the above embodiments, the present invention also proposes a non-transitory computer readable storage medium storing computer instructions for causing a computer to execute the bone identification method in the aforementioned method embodiment for bone identification for bone tissue grinding of the present invention.
For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory. The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc.
Although embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are exemplary and not to be construed as limiting the present invention, and that changes, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention. It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments. In addition, functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Claims (10)
1. A method of bone identification for bone tissue grinding, comprising:
acquiring a grinding force signal and a related signal in real time when the bone tissue is ground, wherein the related signal comprises at least one of grinding speed, grinding depth and ultrasonic knife power;
performing wavelet transformation noise reduction on the grinding force signal, and then extracting the noise-reduced grinding force signal in the vertical direction as a target grinding force signal;
and inputting the target grinding force signal and the correlation signal into a trained BP neural network model for bone identification to determine whether to enter inner cortical bone.
2. The method of claim 1, wherein the wavelet transform denoising comprises:
decomposing the grinding force signal by using a wavelet basis function to obtain an approximate component and a detail component;
selecting a threshold value, and then filtering the detail component by using a soft threshold value function;
the approximation component and the filtered detail component are reconstructed to obtain a noise reduced grinding force signal.
3. The method of claim 2, further comprising:
wavelet basis functions adopt Daubechies series.
4. The method of claim 1, further comprising:
preprocessing a grinding force signal subjected to noise reduction in the vertical direction, wherein the preprocessing comprises the following steps: sequencing all the grinding force signals after noise reduction of the single-layer bone tissue; respectively filtering data in a preset proportion before and after the sequence; and calculating the average value of the residual data as a target grinding force signal of the layer of bone tissue.
5. The method of claim 1 or 4, further comprising:
and respectively carrying out normalization processing on the target grinding force signal and the correlation signal before inputting the target grinding force signal and the correlation signal into a trained BP neural network model.
6. An apparatus for bone identification for bone tissue grinding, comprising:
the acquisition module is used for acquiring a grinding force signal and a related signal in the process of grinding bone tissues in real time, wherein the related signal comprises at least one of grinding speed, grinding depth and ultrasonic knife power;
the de-noising module is used for receiving the grinding force signal output by the acquisition module, performing wavelet transformation de-noising on the grinding force signal, and extracting the de-noised grinding force signal in the vertical direction as a target grinding force signal;
and the identification module is used for receiving the target grinding force signal output by the denoising module, receiving the correlation signal output by the acquisition module, and inputting the target grinding force signal and the correlation signal into a trained BP neural network model for bone identification to determine whether the bone enters the inner cortical bone.
7. The apparatus of claim 6, wherein in the denoising module, the wavelet transform denoising comprises: decomposing the grinding force signal by using Daubechies series wavelet basis functions to obtain an approximate component and a detail component; selecting a threshold value, and then filtering the detail component by using a soft threshold value function; the approximation component and the filtered detail component are reconstructed to obtain a noise reduced grinding force signal.
8. The apparatus of claim 6, further comprising:
in the denoising module, preprocessing is required to be performed on a grinding force signal after denoising in the vertical direction, and the preprocessing includes: sequencing all the grinding force signals after noise reduction of the single-layer bone tissue; respectively filtering data in a preset proportion before and after the sequence; and calculating the average value of the residual data as a target grinding force signal of the layer of bone tissue.
9. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.
10. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210668751.5A CN115081476A (en) | 2022-06-14 | 2022-06-14 | Method and device for bone recognition for bone tissue grinding |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210668751.5A CN115081476A (en) | 2022-06-14 | 2022-06-14 | Method and device for bone recognition for bone tissue grinding |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115081476A true CN115081476A (en) | 2022-09-20 |
Family
ID=83250530
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210668751.5A Pending CN115081476A (en) | 2022-06-14 | 2022-06-14 | Method and device for bone recognition for bone tissue grinding |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115081476A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116570367A (en) * | 2023-05-12 | 2023-08-11 | 北京长木谷医疗科技股份有限公司 | Intelligent sensing prediction method device and equipment for bone grinding and bone quality of robot operation |
-
2022
- 2022-06-14 CN CN202210668751.5A patent/CN115081476A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116570367A (en) * | 2023-05-12 | 2023-08-11 | 北京长木谷医疗科技股份有限公司 | Intelligent sensing prediction method device and equipment for bone grinding and bone quality of robot operation |
| CN116570367B (en) * | 2023-05-12 | 2024-08-16 | 北京长木谷医疗科技股份有限公司 | Intelligent sensing prediction method device and equipment for bone grinding and bone quality of robot operation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11450435B2 (en) | Spinal stenosis detection and generation of spinal decompression plan | |
| EP3694419B1 (en) | Advanced ultrasonic detection of different tissue types | |
| CN103690160B (en) | Electroencephalogram feature extraction method based on non-Gaussian time sequence model | |
| US20230371876A1 (en) | Systems, methods, and devices for detecting the threshold of nerve-muscle response using variable frequency of stimulation | |
| CN110418604B (en) | Medical system and method for detecting electrophysiological evoked potential changes | |
| Jahromi et al. | Cross comparison of motor unit potential features used in EMG signal decomposition | |
| CN115081476A (en) | Method and device for bone recognition for bone tissue grinding | |
| CN110693488A (en) | Electrocardiosignal processing system, electrocardiosignal processing method, electronic equipment and readable storage medium | |
| CN116830103A (en) | Method and apparatus for variable speed osteotomy | |
| CN114470516A (en) | Control method, control system, electronic device, and medium for stimulation mode | |
| CN106137185A (en) | A kind of epileptic chracter wave detecting method based on structure of transvers plate small echo | |
| Al-Salman et al. | Detection of k-complexes in EEG signals using a multi-domain feature extraction coupled with a least square support vector machine classifier | |
| CN114176602B (en) | Method for simultaneously positioning electrocardiograph P wave, QRS wave and T wave based on deep learning multi-target detection | |
| Pauk | 419. Different techniques for EMG signal processing. | |
| CN113951988B (en) | Grinding method, device and system for ultrasonic bone knife | |
| CN117501307A (en) | Device and method for diagnosing pneumonia by frequency analysis of ultrasonic signals | |
| CN113974647B (en) | System and method for reconstructing sudden cardiac death risk factor | |
| CN117297769A (en) | Bone layer identification method in hard bone tissue operation | |
| CN114931384A (en) | MRCP decoding method | |
| CN118522439A (en) | Digital traditional Chinese medicine physiotherapy method and system | |
| Sathesh et al. | Clearer HeartBeats: Enhancement of cardiac sounds using Adaptive Filtering and Wavelet Decomposition | |
| Gheni et al. | Model for classifying EEG signals using deep learning to identify epileptic seizures | |
| CN112402010A (en) | Control method, device and system of ablation pulse, electronic equipment and storage medium | |
| CN116849609A (en) | Automatic positioning method, system, equipment and storage medium for epilepsy inducing area | |
| Walbran et al. | A technique for optimizing electrode placement for electromyographic control of prostheses |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |