Cardiac Region Localization Method Based on Electrical Impedance Tomography Technology

Abstract

The present invention discloses a cardiac region localization method based on electrical impedance tomography technology, belonging to the field of electrical impedance tomography technology, which comprises the following steps: collecting EIT data, and dividing into a localization group and a control group; reconstructing the EIT data to obtain an EIT time series signal map; processing each pixel in the EIT time series signal map to obtain a time series signal intensity map of each pixel; performing standard deviation normalization calculation on the time series signal intensity of each pixel to obtain a time series signal intensity distribution map, selecting the pixel point where the highest intensity value is located as the central position of the cardiac region of interest, and calculating the phase difference between each pixel position and the central position; obtaining the cardiac region localization results of the localization group and the control group according to the phase difference and the time series signal intensity map; comparing the results of the localization group and the control group, and determining the final cardiac localization region after verification, which no longer relies on simple signal intensity threshold division, can expand the clinical application scenarios of EIT, and further enhances the non-invasive superiority of EIT technology.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical impedance tomography, and particularly to a method for localizing the cardiac region based on electrical impedance tomography technology. Background Art

[0002] Electrical Impedance Tomography (EIT) is a technique that uses the internal impedance distribution of an object to form an image. It is currently widely used in the real-time monitoring of the pulmonary ventilation and cardiac function of patients, and can provide dynamic information on pulmonary ventilation and perfusion as well as non-invasive measurement of cardiac output. However, the EIT technique itself has problems with resolution and image quality, and there are also errors in the human impedance distribution model. Therefore, accurate information cannot be obtained when collecting real-time data from clinical patients. In particular, the measurement accuracy of cardiac output and various pulmonary indicators depends heavily on the accuracy of the division of the cardiac region of interest and the pulmonary region of interest, so it will affect clinical diagnosis.

[0003] Currently, the localization of the cardiac region is achieved by defining empirical dual thresholds of global normalized signal intensity to divide the cardiac-lung junction line and the cardiac-lung combined region, and indirectly segmenting the cardiac region of interest (ROI heart ). However, this method has two problems: First, the empirical thresholds do not have universal applicability, and there may be structural differences in the cardiopulmonary signal distributions of different patients; Second, the signal intensity cannot fully reflect the differences in the cardiopulmonary regions, and more information hidden in the EIT signals is ignored. Currently, there is no accurate and standardized way to localize the cardiac region of interest (ROI heart ) in the industry. Summary of the Invention

[0004] In view of the above problems, the present invention provides a method for localizing the cardiac region based on electrical impedance tomography technology, which can improve the accuracy of dividing the cardiac region of interest and can effectively assist clinical analysis and treatment.

[0005] The technical solution of the present invention is as follows:

[0006] A method for localizing the cardiac region based on electrical impedance tomography technology includes the following steps:

[0007] S1. Collect EIT data and divide into a localization group and a control group;

[0008] S2. Reconstruct the EIT data to obtain an EIT time-series signal map;

[0009] S3. Use a high-pass filter to remove the linear drift signal for each pixel in the EIT time-series signal map, and then use a band-pass filter to extract the signal near the perfusion frequency to obtain the time-series signal intensity map of each pixel;

[0010] S4. Perform standard deviation normalization calculation on the temporal signal intensity of each pixel to obtain a temporal signal intensity distribution map. Select the pixel point with the highest intensity value as the center position of the heart region of interest, and calculate the phase difference between each pixel position and the center position.

[0011] S5. Obtain the heart region localization results of the positioning group and the control group according to the phase difference and the temporal signal intensity map.

[0012] S6. Compare the results of the positioning group and the control group, and determine the final heart localization region after verification.

[0013] The positioning group is the spontaneous breathing group, and the control group is the apnea group.

[0014] In step S3, the 3dB cut-off frequency of the high-pass filter is 0.1 - 0.3Hz, and the 3dB cut-off frequency of the band-pass filter is the average heart rate during sampling ±0.1Hz.

[0015] In step S4, before performing the normalization calculation, it is necessary to determine the center position of the heart region of interest in combination with prior information.

[0016] In step S4, the standard deviation normalization calculation uses Fourier transform.

[0017] In step S5, set a threshold for the positioning group, determine the optimal threshold condition, calculate the positioning result of the positioning group under the optimal threshold, calculate the positioning result of the control group using the optimal threshold, and perform manual review to obtain the final control group positioning result.

[0018] Before manual review, the control group needs to fill the holes inside the heart region of interest.

[0019] The beneficial effects of the present invention are:

[0020] This method effectively improves the accuracy of ROI heart localization in traditional EIT chest imaging based on experience. ROI heart localization no longer depends on simple signal intensity threshold division. By adding the utilization of EIT phase information, it overcomes the defect of "varying from person to person" in heart region localization, and can effectively improve the effectiveness and integrity of scene use on the basis of ensuring the robustness of the ROI heart localization technology. For patients receiving diagnosis and treatment, this technology does not require apnea, is applicable to patients with spontaneous breathing, can expand the clinical application scenarios of EIT, and further enhances the non-invasive superiority of EIT technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flowchart of the method for heart region localization based on electrical impedance tomography technology according to an embodiment of the present invention;

[0022] Figure 2 is the logic flow chart of the heart region localization method based on electrical impedance tomography technology according to the embodiments of the present invention. Detailed implementation manners

[0023] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0024] Embodiment:

[0025] As Figure 1 、 Figure 2 shown, the heart region localization method based on electrical impedance tomography technology includes the following steps:

[0026] S1. EIT data acquisition, dividing the positioning group and the control group;

[0027] The patient lies flat, and the electrode suitable for the patient's chest size is properly fixed at the T4 / 5 intercostal space. For each patient, 120 seconds of EIT data (EIT spontaneous ) is collected under spontaneous breathing; after general anesthesia is implemented and mechanical ventilation is stable, the mechanical ventilation is paused for 30 seconds using the breath-holding technique, and EIT data (EIT apnea ) is collected.

[0028] The positioning group is the spontaneous breathing group, and the control group is the apnea group.

[0029] S2. Reconstructing the EIT data to obtain the EIT time series signal diagram;

[0030] Using the reconstruction algorithm of the human chest model provided by Draeger Company, an EIT time series signal with a resolution of 32*32*N is generated. (32*32 represents the spatial resolution of the reconstructed image, and N represents the number of signal frames collected. N = t*Fs, where t is the acquisition time in seconds, and Fs represents the sampling frequency of the EIT device, generally 50Hz).

[0031] S3. Using a high-pass filter to remove the linear drift signal for each pixel in the EIT time series signal diagram, and then using a band-pass filter to extract the signal near the perfusion frequency to obtain the time series signal intensity diagram of each pixel:

[0032] Specifically as follows: The collected time series signal is regarded as a digital signal with a length of N and a resolution of 32*32 pixels. The linear drift of the signal for each pixel is removed using a high-pass filter, and then the signal near the perfusion frequency is extracted using a band-pass filter to obtain the time series signal intensity diagram of each pixel. Using high-pass filtering (the 3dB cut-off frequency is set to 0.2Hz, the filter type is IIR Butterworth, and the order is 50) for the EIT spontaneous and EITapnea The timing signal is filtered to remove the linear signal drift, and for the EIT-derived spontaneous An additional low-pass filter is set (3dB cut-off frequency is the average heart rate during sampling ±0.1Hz, filter type is FIR EQUIRIPPLE, order is 1500). Among them, if there is heart rate data, it is directly read from the standard input signal; if the heart rate is unknown, the timing signal corresponding to the pixel point with the maximum signal intensity after high-pass filtering is selected for Fourier transform, and the frequency corresponding to the maximum value with a frequency greater than 0.8Hz is empirically regarded as the average heart rate. Both filtrations need to remove the signal part related to the filter delay. Calculate the standard deviation of the timing signal calculated pixel by pixel for the filtered timing signal (represented by a 32*32*N array), and normalize the standard deviation values of each pixel of the entire image to generate a 32*32 intensity distribution map of the timing signal, denoted as Imap (Intensity map).

[0033] S4. Perform standard deviation normalization calculation on the intensity of the timing signal of each pixel to obtain the intensity distribution map of the timing signal. Select the pixel point where the highest intensity value is located as the central position of the heart region of interest, and calculate the phase difference between each pixel position and the central position:

[0034] The phase of the timing signal at any pixel at the perfusion frequency is obtained in the following way: Use Fourier transform to convert the timing signal from the time domain to the frequency domain. On the phase spectrum of the frequency domain signal, find the phase value corresponding to the perfusion frequency (heart rate), which is the obtained phase. The position of the maximum pixel value in Imap is regarded as the "central position" of the heart region. Subtract the phase of the timing signal corresponding to each pixel from the phase of the pixel at the "central position" of the heart region, calculate the "phase difference" between the two at each pixel corresponding position (theoretically in the range of [-π, π)), and take its absolute value (in the range of [0, π)). Denote the 32*32 phase distribution map represented by this absolute value of the phase difference as Pmap (Phase map).

[0035] S5. Obtain the heart region localization results of the localization group and the control group according to the phase error and the timing signal intensity map;

[0036] The specific steps are as follows:

[0037] Step 1: For the EIT-derived apneaSet a set of initial intensity thresholds (denoted as I_t, within [0, 1]) and phase delay thresholds (denoted as P_t, within [0, π]), marked as I_t / P_t, for the intensity distribution map (Imap) and phase distribution map (Pmap). In Imap, select all pixels with values greater than I_t to form a pixel set, denoted as RI (Region of Intensity-valid); in Pmap, select all pixels with values less than P_t to form a pixel set, denoted as RP (Region of Phase-valid). Take the intersection of the pixel sets RI and RP, denoted as RIP (Region of Intensity-and Phase-valid). By generating a binary mask (32*32), locate the RIP and non-RIP regions (represented by 1 and 0 respectively); to ensure connectivity, remove all pixels in RIP that cannot have an adjacent relationship with the "central position" pixel of the heart region through any other RIP pixel, and further fill all holes within the RIP region. The finally generated binary mask is denoted as RIP mask:i (i = 1), which is used to represent the ROI heart . In Imap, set the regions with a value of 1 in RIP mask:i to 0, that is, the output contains the ROI heart Perfusion signal intensity distribution map with localization information.

[0038] Step 2: Repeat Step 1 cyclically and adjust the I_t / P_t values successively. Manually measure the accuracy of the detection results (to determine whether to increase or decrease the I_t / P_t values in the next repetition) to obtain a new RIP mask:i (i = number of detections). Repeat the adjustment until it is subjectively considered that the localization accuracy of the heart region cannot be significantly improved by changing the I_t / P_t, and then terminate. Select the I_t / P_t value selected at this time as the optimal value.

[0039] Step 3: Use the optimal I_t / P_t value obtained in Step 2 for the Imap and Pmap derived from EIT spontaneous , and then execute Step 1 to finally obtain the perfusion signal intensity distribution map containing the ROI heart Perfusion signal intensity distribution map with localization information.

[0040] Perform a manual review process synchronously. Compare the Imap obtained through digital signal processing in S3 with the RIP mask result obtained in Step 3 to verify whether the positioning result of the heart area is reliable. In existing cases, usually, the RIP mask is reliable (95%). In case of obvious errors (mostly caused by incorrect "central position" of the automatically selected heart area), the backup plan can be run, that is, manually select the position of the pixel of the "central position" of the heart area through the visualization interface for correction, and the rest of the technical route is the same as the previous plan.

[0041] S6. Determine the final heart positioning area after verification.

[0042] By comparing the RIP masks of data from EIT apnea and EIT spontaneous under the same I_t / P_t settings, evaluate the effectiveness of the patent ROI heart positioning algorithm. Mark the RIP mask result of EIT spontaneous as R1, and the RIP mask result of EIT apnea as R2. Use the IoU (Intersection of Union) metric to quantify the consistency.

[0043]

[0044] The closer this value is to 1, the better the consistency between the two. Use IoU_mean to represent the average value of IoU calculated in all EIT apnea to measure whether the algorithm is effective enough.

[0045] The above-described embodiments only represent the specific implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention.

Claims

1. A method for cardiac region localization based on electrical impedance tomography technology, characterized in that It includes the following steps: S1. Collect EIT data and divide it into a positioning group and a control group; S2. Reconstruct the EIT data to obtain an EIT time-series signal diagram; S3. Use a high-pass filter to remove the linear drift signal for each pixel in the EIT time-series signal diagram, and then use a band-pass filter to extract the signal near the perfusion frequency to obtain the time-series signal intensity diagram of each pixel; S4. Perform standard deviation normalization calculation on the time-series signal intensity of each pixel to obtain a time-series signal intensity distribution diagram, select the pixel point with the highest intensity value as the central position of the heart region of interest, and calculate the phase difference between each pixel position and the central position; S5. Obtain the heart region positioning results of the positioning group and the control group according to the phase difference and the time-series signal intensity diagram; S6. Compare the results of the positioning group and the control group, and determine the final heart positioning region after verification; Among them, S5 is specifically as follows: According to the phase difference of the EIT data of the positioning group and the control group and the time-series signal intensity distribution diagram, set the initial intensity threshold and the phase delay threshold, construct a pixel set based on the set thresholds, select the intersection of the pixel sets to generate a binary mask, locate the heart region through the binary mask, and after manual review, perform cyclic adjustment to determine the best heart region positioning result, and finally complete the heart region positioning of the positioning group and the control group.

2. The method for cardiac region localization based on electrical impedance tomography technology according to claim 1, wherein, The positioning group is the spontaneous breathing group, and the control group is the apnea group.

3. The method for localizing the cardiac region based on electrical impedance tomography technology according to claim 1, wherein In step S3, the 3dB cut-off frequency of the high-pass filter is 0.1 - 0.3Hz, and the 3dB cut-off frequency of the band-pass filter is the average heart rate during sampling ± 0.1Hz.

4. The method for localizing the cardiac region based on electrical impedance tomography technology according to claim 1, wherein In step S4, before performing the normalization calculation, it is necessary to determine the central position of the heart region of interest in combination with prior information.

5. The method for localizing the cardiac region based on electrical impedance tomography technology according to claim 1, wherein In step S4, the standard deviation normalization calculation uses Fourier transform.

6. The method for cardiac region localization based on electrical impedance tomography technology according to claim 1, wherein In step S5, set a threshold for the positioning group, determine the optimal threshold condition, calculate the positioning result of the positioning group under the optimal threshold, calculate the positioning result of the control group using the optimal threshold, and perform manual review to obtain the final control group positioning result.

7. The method for localizing the cardiac region based on electrical impedance tomography according to claim 6, wherein The control group needs to fill the inside of the heart region of interest before manual review.

Images