A Phase Registration Method for Lesion Feature Motion Curve and Respiratory Fluctuation Curve

Abstract

The present invention discloses a phase registration method for a lesion feature motion curve and a respiratory undulation curve, comprising the steps of: (1) obtaining a lesion feature motion curve and a respiratory undulation curve; (2) performing offset traversal on the external respiratory undulation curve by using the lesion feature motion curve on the time axis to obtain a lesion feature motion mapping curve; (3) obtaining the respiratory undulation curve segment corresponding to the sampling time of the lesion feature motion mapping curve during each movement, and calculating the matching value between the two; (4) selecting the registration relationship corresponding to the minimum matching value calculated in step (3), obtaining the respiratory undulation curve segment with the best mapping registration degree with the lesion feature motion mapping curve, and thereby obtaining the mapping relationship between the lesion feature motion curve and the respiratory undulation curve. The present invention obtains an accurate mapping relationship between the lesion feature motion curve and the external respiratory undulation curve through the above method, and doctors can achieve precise puncture under the guidance of this mapping relationship.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a phase registration method for a lesion feature motion curve and a respiratory fluctuation curve. Background Art

[0002] Respiratory monitoring technology is an important and necessary technology for puncture of pulmonary nodules, precise radiotherapy of tumors, etc., which can reduce the pain of clinical patients and greatly improve the treatment effect. The key and difficult part of respiratory monitoring technology is to establish the mapping relationship between the diaphragm movement caused by respiration and the lesion movement.

[0003] The lesion feature motion data is derived from 4D-CT images. Features that can represent the lesion motion curve are extracted from the images, such as the lesion itself, the bronchi near the lesion, the vascular bifurcation points, the subtracted motion part, etc. The centroid of the above features is extracted through medical image processing algorithms to form the lesion feature motion data that fluctuates with time. Respiratory motion collectors such as optical trackers, electromagnetic navigation sensors, and respiratory flow sensors placed on the abdomen can directly obtain the data of the diaphragm movement caused by respiration, so as to obtain the respiratory fluctuation data that fluctuates with time. This method is relatively accurate and has a high sampling rate. At present, the data reflecting the lesion motion characteristics is collected based on CT, DR, etc. Due to the time resolution limitation of CT, the sampling rate is low (not higher than 4Hz). In addition, due to the uncertainty of the lesion feature algorithm and the artifacts generated during inhalation and exhalation, the centroid extraction has a high degree of unbelievability, resulting in a complex and inaccurate solution for the mapping relationship between the respiratory fluctuation curve and the lesion motion curve. Summary of the Invention

[0004] Object of the Invention: Aiming at the above deficiencies, the present invention proposes a phase registration method for a lesion feature motion curve and a respiratory fluctuation curve, and establishes the phase matching between the two based on the respiratory motion data and the lesion feature motion data to solve the mapping relationship between the two.

[0005] Technical Solution:

[0006] A phase registration method for a lesion feature motion curve and a respiratory fluctuation curve includes the steps of:

[0007] (1) Obtain the lesion feature motion curve and the respiratory fluctuation curve;

[0008] (2) On the time axis, use the lesion feature motion curve to perform offset traversal on the external respiratory fluctuation curve to obtain the corresponding lesion feature motion mapping curve;

[0009] (3) Align the phases on the respiratory fluctuation curve and the lesion feature motion mapping curve to obtain the respiratory fluctuation curve segments corresponding to the sampling times of the lesion feature motion mapping curve during each movement, and calculate the matching value between the lesion feature motion mapping curve and the corresponding respiratory fluctuation curve segments;

[0010] (4) Select the registration relationship corresponding to the minimum matching value among the matching values calculated in step (3) to obtain the respiratory fluctuation curve segment with the best mapping registration degree with the lesion feature motion mapping curve, and thereby obtain the mapping relationship between the lesion feature motion curve and the respiratory fluctuation curve.

[0011] Between step (1) and step (2), there is also a step of respectively normalizing the lesion feature motion curve and the respiratory fluctuation curve using the min-max normalization method. The normalization formula is as follows:

[0012]

[0013] where, x scale represents the amplitude of the corresponding data set after normalization, x represents the amplitude of the corresponding data set before normalization, x min represents the minimum amplitude of the corresponding data set before normalization, x max represents the maximum amplitude of the corresponding data set before normalization.

[0014] The formula for calculating the matching value in step (3) is as follows:

[0015]

[0016] where, y ei ′ and y qi ′ are respectively the amplitudes of the corresponding phase points on the respiratory fluctuation curve segment and the lesion feature motion mapping curve after normalization, f w is the weight value at the corresponding amplitude point, and K represents the sampling time length of the lesion feature motion curve.

[0017] The formula for f w is:

[0018]

[0019] where, x inner is the real-time amplitude of the lesion feature motion curve after normalization, x inner_max is the maximum amplitude of the lesion feature motion curve after normalization, x inner_min is the minimum amplitude of the lesion feature motion curve after normalization, a represents the maximum assignment weight coefficient, and b represents the minimum assignment weight coefficient.

[0020] The formula for calculating the matching value in step (3) is as follows:

[0021]

[0022] Where y ei and y qi are the amplitudes of the corresponding phase points on the respiratory fluctuation curve segment and the lesion characteristic motion mapping curve respectively, f w is the weight value at the corresponding amplitude point, and K represents the sampling time length of the lesion characteristic motion curve.

[0023] The formula for f w is:

[0024]

[0025] Where x inner is the real-time amplitude of the lesion characteristic motion curve, x inner_max is the maximum amplitude of the lesion characteristic motion curve, x inner_min is the minimum amplitude of the lesion characteristic motion curve, a represents the maximum assignment weight coefficient, and b represents the minimum assignment weight coefficient.

[0026] For the weight assignment of the amplitudes of the lesion characteristic motion curve, it gradually decreases from the end of inspiration and the end of expiration to the intermediate weight value.

[0027] The formula for calculating the matching value in step (3) is as follows:

[0028]

[0029] Where y ei ′ and y qi ′ are the amplitudes of the corresponding phase points on the respiratory fluctuation curve segment and the lesion characteristic motion mapping curve after normalization respectively, and K represents the sampling time length of the lesion characteristic motion curve.

[0030] The formula for calculating the matching value in step (3) is as follows:

[0031]

[0032] Where y ei and y qi are the amplitudes of the corresponding phase points on the respiratory fluctuation curve segment and the lesion characteristic motion mapping curve respectively, and K represents the sampling time length of the lesion characteristic motion curve.

[0033] Beneficial effects: By extracting the data corresponding to the time components of the lesion feature motion curve and the external respiratory fluctuation curve, calculating the corresponding matching values based on the data after moving traversal, and selecting the matching result corresponding to the minimum matching value as the best matching result, the registration relationship between the motion amplitude of the lesion feature and the respiratory fluctuation amplitude after registration, that is, the mapping relationship, can be obtained. Guided by the accurate mapping relationship between the lesion feature motion curve and the external respiratory fluctuation curve obtained by the above method, doctors can achieve precise puncture. Description of the Drawings

[0034] Figure 1 is the flowchart of the phase registration method of the present invention;

[0035] Figure 2 is the scatter plot of the original lesion feature motion curve and the external respiratory fluctuation curve; among them, the curve with a long sampling time is the lesion feature motion curve, and the curve with a short sampling time is the respiratory fluctuation curve;

[0036] Figure 3 is the weight assignment function graph;

[0037] Figure 4 is the data registration result graph. Detailed Embodiment

[0038] The present invention will be further illustrated below in conjunction with the drawings and specific embodiments.

[0039] There is a certain periodic law between the lesion feature motion curve derived from 4D-CT and the external respiratory fluctuation curve derived from a respiratory sampler. The respiratory gating technology predicts the lesion motion condition based on the phase of the external respiratory fluctuation curve to achieve precise puncture or tumor radiotherapy and other surgeries. Due to the complexity and uncertainty of the extraction of the lesion feature motion curve, there is a non-linear mapping relationship between the lesion feature motion curve and the external respiratory fluctuation curve. The present invention obtains the mapping relationship between the two after registering the corresponding data of the two on the time axis.

[0040] Figure 1 is the flowchart of the phase registration method of the present invention. As Figure 1 shown, the phase registration method of the lesion feature motion curve and the respiratory fluctuation curve of the present invention includes the following steps:

[0041] (1) Obtain the lesion feature motion curve and the respiratory fluctuation curve:

[0042] (11) Collect the 4D-CT images of the patient's lesion and extract the lesion motion characterization data D inner ={(t qi , y qi )|i = 0, 1…m}, where y qi is the moment tqi The corresponding amplitude of lesion movement

[0043] (12) Obtain the external respiratory fluctuation data D corresponding to the patient's respiration ext ={(t ei , y ei ) | i = 0, 1…n}, where y ei is the respiratory fluctuation amplitude corresponding to time t ei ;

[0044] The reference data is as Figure 2 shown, t ei and t qi represent the elapsed time, starting from zero;

[0045] (2) Normalize the amplitudes of the two groups of curves to eliminate the dimension difference:

[0046] Respectively process the lesion characteristic motion curve and the respiratory fluctuation curve obtained in step (1) using the min-max normalization method, and normalize their amplitudes to the range of 0 to 1 to eliminate the difference introduced by different amplitudes; the specific min-max normalization formula is as follows:

[0047]

[0048] where x scale represents the amplitude of the corresponding data set after normalization, x represents the amplitude of the corresponding data set before normalization, x min represents the minimum amplitude of the corresponding data set before normalization, and x max represents the maximum amplitude of the corresponding data set before normalization;

[0049] After performing the normalization in this step, the normalized lesion motion characterization data D inner ′ = {(t qi , y qi ′) | i = 0, 1…m} and the external respiratory fluctuation data D ext ′ = {(t ei , y ei ′) | i = 0, 1…n} are obtained, where y qi ′ is the lesion motion amplitude corresponding to time t qi after normalization, and y ei ′ is the respiratory fluctuation amplitude corresponding to time t ei ;

[0050] (3) On the time axis, use the lesion characteristic motion curve to perform offset traversal on the external respiratory fluctuation curve to obtain the corresponding lesion characteristic motion mapping data set D inner-off ′:

[0051] Since the lesion characteristic motion curve Dinner has a low sampling rate, a short sampling time range, and few sampling points. Sequentially, using the external respiratory fluctuation curve D ext ′s sampling interval m step as the step size, starting from the t ext of the respiratory fluctuation curve D qi ′, gradually move the lesion feature motion curve D ext ′ on the time axis of the respiratory fluctuation curve D inner ′, and when the k-th movement is obtained, that is, the sampling time is t qi +k*m step , obtain the lesion feature motion mapping amplitude y inner ′ of the lesion feature motion curve D ext ′ and the respiratory fluctuation curve D qi ′ at the corresponding phase, so as to obtain the corresponding lesion feature motion mapping curve D inner-off ′={(t qi +k*m step ,y qi ′)|i=0,1…K;}. When moving to the end of the time axis of the respiratory fluctuation curve D ext ′, the movement is terminated. k is the number of movements, and K represents the sampling time length of the lesion feature motion curve;

[0052] (4) Matching value calculation:

[0053] (41) According to the amplitudes on the lesion feature motion mapping curve D inner-off ′ obtained in step (4), align the phases on the respiratory fluctuation curve D ext ′ and the lesion feature motion mapping curve D inner-off ′, and obtain the respiratory amplitude corresponding to the same phase on the external respiratory fluctuation curve D ext ′ as the lesion feature motion mapping curve D inner-off ′, so as to obtain the respiratory fluctuation curve segment D inner-off ′ corresponding to the sampling time of the lesion feature motion mapping curve D ext-off ′ at the k-th movement; if there is no corresponding respiratory amplitude value for the corresponding phase on the respiratory fluctuation curve D ext ′ and the lesion feature motion mapping curve D inner-off ′, obtain the corresponding amplitude by linear interpolation according to the characteristics of the respiratory fluctuation curve D ext ′; thus obtain k segments of respiratory fluctuation curve segments D inner-off corresponding to the lesion feature motion mapping curve D ext-off ′;

[0054] (42) For the lesion feature motion mapping curve D inner-off ′ and the corresponding k-th segment of the respiratory fluctuation curve segment Dext-off 'Calculate the matching value between the two, and the calculation formula is as follows:

[0055]

[0056] Among them, y ei ' and y qi ' are the amplitudes corresponding to the corresponding phases on the corresponding respiratory fluctuation curve segment D ext-off ' and the lesion feature motion mapping curve D inner-off '. K represents the sampling time length of the lesion feature motion curve; f w is the weight at the corresponding amplitude point, and the specific weight assignment is as follows:

[0057] Because the amplitude accuracies of the data sampled at different phases are different, by assigning small weights to the data amplitudes with low accuracy and large weights to the data amplitudes with high accuracy, the confidence of the data amplitude results is improved. As Figure 3 shown, the calculation formula of f w is:

[0058]

[0059] Among them, x inner is the normalized real-time amplitude of the lesion feature motion curve, x inner_max is the normalized maximum amplitude of the lesion feature motion curve, x inner_min is the normalized minimum amplitude of the lesion feature motion curve, a represents the maximum assignment weight coefficient, and b represents the minimum assignment weight coefficient;

[0060] It is also possible not to calculate the weight value using the above formula, but for the doctor to assign weights according to experience. Specifically, considering that the periodic curve of respiratory fluctuation is similar to a sine wave, the first derivative is zero near the peak, and the first derivative first increases and then decreases with time. Therefore, the residence time is longer at the end of exhalation and inhalation. As a result, the target extracted from the lesion features corresponding to this phase is stable without artifacts and has high credibility. Then, a larger weight value is assigned to the amplitude of this part of the lesion feature motion curve, and a lower weight value is assigned to the amplitude tending to the middle part of the lesion feature motion curve;

[0061] (5) Optimal matching output:

[0062] According to the matching value calculated in step (4), select the registration relationship corresponding to the minimum matching value. The one corresponding to this minimum matching value is the state with the best registration degree between the lesion feature motion mapping curve D inner-off ' and the respiratory fluctuation curve segment D ext-off '. That is, obtain the respiratory fluctuation curve segment D inner with the best registration degree with the lesion feature motion curve D ext-off The registration result is asFigure 4 as shown;

[0063] (6) Phase mapping:

[0064] Based on the lesion feature motion curve D obtained in step (5) inner and the respiratory undulation curve segment D with the best registration accuracy ext-off , the mapping relationship between the lesion feature motion curve and the respiratory undulation curve is obtained to predict the internal lesion motion.

[0065] In other embodiments, the step of normalizing the amplitudes of the two sets of curves to eliminate the dimensional difference can be omitted. In this case, the registration accuracy will decrease, but it can still meet the usage requirements. For this situation, the calculation formula for the matching value of the two is as follows:

[0066]

[0067] where y ei and y qi are respectively the amplitudes of the corresponding phase points of the respiratory undulation curve segment and the lesion feature motion mapping curve that have not been normalized, f w is the weight value at the corresponding amplitude point, and K represents the sampling time length of the lesion feature motion curve.

[0068] Correspondingly, the weighted amplitude at the amplitude point of the lesion feature motion curve is similar to that in step (2), except that the lesion feature motion curve used therein is not normalized.

[0069] Furthermore, in other embodiments, when calculating the matching value for the two sets of normalized data, the amplitude of the lesion feature motion curve may not be weighted, and the calculation formula for the matching value is as follows:

[0070]

[0071] where y ei ′ and y qi ′ are respectively the amplitudes of the corresponding phase points on the respiratory undulation curve segment and the lesion feature motion mapping curve after normalization, and K represents the sampling time length of the lesion feature motion curve.

[0072] Even further, in other embodiments, when calculating the matching value for the two sets of unnormalized data, the amplitude may not be weighted, and the calculation formula for the matching value is as follows:

[0073]

[0074] where y ei and y qiThey are the amplitudes of the corresponding phase points of the respiratory fluctuation curve segment and the lesion characteristic motion mapping curve respectively, and K represents the sampling time length of the lesion characteristic motion curve.

[0075] The present invention extracts the data corresponding to the time components of the lesion characteristic motion curve and the external respiratory fluctuation curve, calculates the corresponding matching values based on the data after moving traversal, and selects the matching result corresponding to the minimum matching value as the best matching result, so as to obtain the registration relationship, that is, the mapping relationship, between the amplitude of the lesion characteristic motion and the amplitude of the respiratory fluctuation after registration. Under the guidance of the accurate mapping relationship between the lesion characteristic motion curve and the external respiratory fluctuation curve obtained by the above method, doctors can achieve precise puncture.

[0076] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations (such as quantity, shape, position, etc.) can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A phase registration method for a lesion feature motion curve and a respiratory fluctuation curve, characterized in that: Including the steps: (1) Obtain the lesion characteristic motion curve and the respiratory fluctuation curve; (2) Taking the sampling interval of the external respiratory fluctuation curve as the step size, starting from a certain moment of the external respiratory fluctuation curve, gradually move the lesion characteristic motion curve on its time axis to perform offset traversal on the external respiratory fluctuation curve, and obtain the lesion characteristic motion mapping amplitudes at the corresponding phases of the lesion characteristic motion curve and the external respiratory fluctuation curve at different moments, so as to obtain the corresponding lesion characteristic motion mapping curve; (3) Align the phases on the respiratory fluctuation curve and the lesion characteristic motion mapping curve, obtain the respiratory fluctuation curve segments corresponding to the sampling times of the lesion characteristic motion mapping curve during each movement, and calculate the matching values between the lesion characteristic motion mapping curve and the corresponding respiratory fluctuation curve segments; (4) Select the registration relationship corresponding to the minimum matching value calculated in step (3), obtain the respiratory fluctuation curve segment with the best mapping registration degree with the lesion characteristic motion mapping curve, and thereby obtain the mapping relationship between the lesion characteristic motion curve and the respiratory fluctuation curve.

2. The phase registration method for the lesion feature motion curve and the respiratory fluctuation curve according to claim 1, characterized in that: Between the step (1) and the step (2), there is also a step of respectively performing normalization processing on the lesion characteristic motion curve and the respiratory fluctuation curve by using the maximum-minimum normalization method, and the normalization formula is as follows: Among them, x scale represents the amplitude of the corresponding data set after normalization, x represents the amplitude of the corresponding data set before normalization, x min represents the minimum amplitude of the corresponding data set before normalization, x max represents the maximum amplitude of the corresponding data set before normalization.

3. The phase registration method for the lesion feature motion curve and the respiratory fluctuation curve according to claim 2, wherein: The formula for calculating the matching value in the step (3) is as follows: where y ei ′ and y qi ′ are the amplitudes of the corresponding phase points on the respiratory fluctuation curve segment and the lesion characteristic motion mapping curve respectively after normalization, f w is the weight value at the corresponding amplitude point, and K represents the sampling time length of the lesion characteristic motion curve.

4. The phase registration method for the lesion feature motion curve and the respiratory fluctuation curve according to claim 3, characterized in that: The said f w The calculation formula is as follows: Among them, x inner is the real-time amplitude after normalization of the lesion feature motion curve, and x inner_max is the maximum amplitude after normalization of the lesion feature motion curve, and x inner_min is the minimum amplitude after normalization of the lesion feature motion curve. a represents the maximum assignment weight coefficient, and b represents the minimum assignment weight coefficient.

5. The phase registration method for the lesion feature motion curve and the respiratory fluctuation curve according to claim 1, wherein: The formula for calculating the matching value in the step (3) is as follows: where y ei and y qi are the amplitudes of the corresponding phase points of the respiratory fluctuation curve segment and the lesion feature motion mapping curve respectively, f w is the weight value at the corresponding amplitude point, and K represents the sampling time length of the lesion feature motion curve.

6. The phase registration method for the lesion feature motion curve and the respiratory fluctuation curve according to claim 5, characterized in that: The said f w The calculation formula is as follows: Among them, x inner is the real-time amplitude of the lesion feature motion curve, and x inner_max is the maximum amplitude of the lesion feature motion curve, and x inner_min is the minimum amplitude of the lesion feature motion curve. a represents the maximum assignment weight coefficient, and b represents the minimum assignment weight coefficient.

7. The phase registration method for the lesion feature motion curve and the respiratory fluctuation curve according to claim 3 or 5, characterized in that: The weight assignment for the amplitude of the lesion characteristic motion curve gradually decreases from the end of inspiration and the end of expiration to the intermediate weight value.

8. The phase registration method for the lesion feature motion curve and the respiratory fluctuation curve according to claim 2, characterized in that: The formula for calculating the matching value in the step (3) is as follows: where y ei ′ and y qi ′ are the amplitudes of the corresponding phase points on the respiration fluctuation curve segment and the lesion feature motion mapping curve respectively after normalization processing, and K represents the sampling time length of the lesion feature motion curve.

9. The phase registration method for the lesion feature motion curve and the respiratory fluctuation curve according to claim 1, wherein: The formula for calculating the matching value in the step (3) is as follows: where y ei and y qi are the amplitudes of the corresponding phase points of the respiratory fluctuation curve segment and the lesion characteristic motion mapping curve respectively, and K represents the sampling time length of the lesion characteristic motion curve.

Images