WO2022199619A1 - Electrical impedance tomography based lung assessment - Google Patents

Electrical impedance tomography based lung assessment Download PDF

Info

Publication number
WO2022199619A1
WO2022199619A1 PCT/CN2022/082511 CN2022082511W WO2022199619A1 WO 2022199619 A1 WO2022199619 A1 WO 2022199619A1 CN 2022082511 W CN2022082511 W CN 2022082511W WO 2022199619 A1 WO2022199619 A1 WO 2022199619A1
Authority
WO
WIPO (PCT)
Prior art keywords
lung
subject
eit
computer
implemented method
Prior art date
Application number
PCT/CN2022/082511
Other languages
French (fr)
Inventor
Russell Wade CHAN
Fedi ZOUARI
Luca MINCIULLO
Dipyaman MODAK
Pak Heng Justin CHAN
Original Assignee
Gense Technologies Limited
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Gense Technologies Limited filed Critical Gense Technologies Limited
Publication of WO2022199619A1 publication Critical patent/WO2022199619A1/en

Links

Images

Classifications

    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/05Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves 
    • A61B5/053Measuring electrical impedance or conductance of a portion of the body
    • A61B5/0536Impedance imaging, e.g. by tomography
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/08Detecting, measuring or recording devices for evaluating the respiratory organs
    • A61B5/0809Detecting, measuring or recording devices for evaluating the respiratory organs by impedance pneumography
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/08Detecting, measuring or recording devices for evaluating the respiratory organs
    • A61B5/087Measuring breath flow
    • A61B5/0871Peak expiratory flowmeters
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/08Detecting, measuring or recording devices for evaluating the respiratory organs
    • A61B5/091Measuring volume of inspired or expired gases, e.g. to determine lung capacity

Definitions

  • the invention relates to systems and methods for analyzing electrical impedance tomography (EIT) data for lung health assessment.
  • EIT electrical impedance tomography
  • the invention can be implemented as a diagnostic tool.
  • Lung diseases can be irreversible, and early diagnosis of these diseases may not be readily available.
  • COPD chronic obstructive pulmonary disease
  • Conventional clinical assessment and pulmonary function tests typically utilizes spirometry and CT scans. These methods lack regional lung functional assessment and may not be sensitive to early-stages of lung diseases.
  • these existing diagnosing techniques are mostly non-portable, require skilled operators, expensive, and/or harmful.
  • a method /technique for calculating or determining voxel-wise respiratory-related parameters to assess the lungs health of the users comprises: analyzing EIT signals measured at, at least, one or more or all of the following breathing modes: (1) guided breathing, (2) spontaneous breathing, and (3) forced expiration and inspiration.
  • the guided breathing mode is used for calibrating the measurement system and extracting lung clusters.
  • the guided breathing and spontaneous breathing are used to infer respiratory parameters such as the breathing amplitude, relative phase delay of different lung segments (voxels) , spatial lung homogeneity, etc.
  • the forced expiration and inspiration breathing mode is used to infer respiratory parameters related to the maximum lung capacity and speed to inhale and exhale air.
  • the forced expiration and inspiration follows the standard spirometry test protocol set by the American Thoracic Society.
  • voxel-wise standard spirometry indicators can be inferred.
  • Additional breathing mode (s) may be used.
  • a system for determining or calculating respiratory-related parameters to assess the lungs health of the users comprising: means for measuring EIT signals at, at least, three breathing modes: (1) guided breathing mode, (2) spontaneous breathing mode, and (3) forced expiration and inspiration mode.
  • the system further comprises means configured to implement techniques for inferring activated lungs regions, quantifying respiratory-related indicators and their spatial distribution across the lungs.
  • a computer-implemented method for lung health assessment comprising: receiving EIT data associated with a lung of a subject; and processing the received EIT data to determine a health condition of the lung of the subject.
  • the EIT data is raw EIT data.
  • the EIT data is obtained from the subject or the subject’s lung during a guided breathing operation.
  • the subject periodically or regularly inhales and exhales, optionally with a predetermined number of breaths per minute.
  • the EIT data is obtained from the subject or the subject’s lung during a forced breathing operation.
  • the subject first takes a predetermined number of shallow breaths, then inhales to maximum or medium lung capacity, then exhales abruptly or slowly, then inhales to maximum or medium lung capacity, then exhales to normal breathing exhalation volume, and finally takes a predetermined number of shallow breaths.
  • the EIT data is obtained from the subject or the subject’s lung during a spontaneous breathing operation.
  • the spontaneous breathing operation is an unguided operation in which the subject can breathe unguided.
  • the processing comprises: denoising the received EIT data; and reconstructing time-difference EIT images based on the denoised EIT data.
  • the denoising comprises: filtering the EIT data; and/or adjusting outlier measurements in the EIT data based on reference voltage data.
  • the reconstruction is performed using a Gauss–Newton method, such as a one-step linear Gauss-Newton solver optionally with a regularization matrix based on the Newton's one-step error re-constructor prior.
  • a Gauss–Newton method such as a one-step linear Gauss-Newton solver optionally with a regularization matrix based on the Newton's one-step error re-constructor prior.
  • the processing further comprises, for each time series of the time-difference EIT images: determining a global conductivity measure or curve that represents a dominant conductivity signal in the time series due to breathing of the subject; determining an amplitude measure or map that represents a local conductivity variation at different voxels of the images, and/or determining a correlation measure or map that represents voxel-wise correlation between the global conductivity-time curve and the respective local conductivity-time curve.
  • the global conductivity measure or curve is determined based on feature selection through filtering and signal extraction.
  • the processing further comprises analyzing the global conductivity curve, the amplitude map, and/or the correlation map to determine the health condition of the lung of the subject.
  • the processing further comprises, for each time series of the time-difference EIT images: determining one or more parametric functional indicator measures or maps, each of which is related to a respective indicator measure.
  • the one or more parametric functional indicator measures or maps are determined based on the global conductivity curve and a voxel-wise conductivity curve.
  • the indicator measure is spirometry indicator measure
  • the one or more parametric functional indicator comprise: maximum volume engaged (MVE) corresponding to forced vital capacity (FVC) measure; exhaled volume in 1 second (EV1) corresponding to forced expiration volume in 1 second (FEV1) measure; exhaled volume in 1 second (EV1) to maximum volume engaged (MVE) ratio corresponding to forced expiration volume in 1 second (FEV1) to forced vital capacity (FVC) ratio measure; maximum expiration flow (MVE) corresponding to peak expiratory flow (PEF) measure; and/or expiratory flow at 25-75%of maximum volume engaged (EF25-75%) corresponding to forced expiratory flow at 25–75%of forced vital capacity (FEF25–75%) measure.
  • MVE maximum volume engaged
  • FVC forced vital capacity
  • the processing further comprises: processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine a spirometry indicator measure of the subject.
  • This spirometry indicator measure is determined for EIT data obtained using the forced breathing operation.
  • the processing further comprises: processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine an indicator of the subject indicative of lung condition, the indicator comprises amplitude, activated voxels, and /or coefficient of variation, which are non-spirometry indicator.
  • a model e.g., a regression model
  • the non-spirometry indicator is determined for EIT data obtained using the guided breathing operation or spontaneous breathing operation.
  • a system for lung health assessment comprising one or more processors arranged to: receive EIT data associated with a lung of a subject; and process the received EIT data to determine a health condition of the lung of the subject.
  • the one or more processors may be arranged to perform the method of the third aspect.
  • the system further comprises a display for displaying the processing results.
  • non-transitory computer-readable medium comprising instructions which, when executed by one or more processors, causes the one or more processors to perform the method of the third aspect.
  • Figure 1 is a schematic diagram of a data analysis pipeline for guided breathing mode in one embodiment.
  • Figure 2 is a schematic diagram of a data analysis pipeline for spontaneous breathing mode in one embodiment.
  • Figure 3 is a schematic diagram of a data analysis pipeline for forced expiration and inspiration mode in one embodiment.
  • Figure 4A is a schematic diagram illustrating a method for lung health assessment in one embodiment.
  • Figure 4B is a schematic diagram illustrating four guided breathing paradigms, including a combination of full or mid capacity inhale, and fast or slow exhale, under the forced breathing mode in some embodiments.
  • Figure 4C contain graphs showing the correlating of EIT-derived parameters and spirometry indicators in some embodiments.
  • MVE Maximal volume engaged
  • EV1 exhaled volume in 1 second
  • MEF maximum expiratory flow
  • EF25-75% expiratory flow at 25-75%of maximum volume engaged
  • Figure 4D contain parametric maps of the EIT-derived spirometry indicators in some embodiments.
  • Figure 5A is a schematic diagram illustrating a method for lung health assessment in one embodiment.
  • Figure 5B contain graphs and plots showing the plots of shallow and deep breathing modes in one embodiment.
  • Figure 5C contain graphs and plots showing measurements taken from COVID-19 discharged subject and controls.
  • Figure 5D contain graphs and plots showing measurements taken from COVID-19 discharged subject and controls.
  • Figure 5E contain graphs and plots showing measurements taken from COVID-19 discharged subject and controls.
  • Figure 6 is a schematic block diagram of a portable EIT console in one embodiment.
  • Figure 7 is a schematic block diagram of a lung health assessment system in one embodiment.
  • Figure 8 is a flow diagram illustrating a general processing method for processing EIT lung data for lung health assessment in one embodiment.
  • Figure 9A contain plots illustrating the definition of lung ROI in one embodiment.
  • Figure 9B contain graphs illustrating regional EIT-derived indicators in one embodiment.
  • Figure 10 contain graphs illustrating measurements taken from shallow and deep breathing modes.
  • Figure 11 contain graphs of both training and testing sets of the EIT-derived spirometry indicators in some embodiments.
  • Some embodiments of the present invention are to measure EIT signals for three breathing modes: (1) guided breathing, (2) spontaneous breathing, and/or (3) forced expiration and inspiration.
  • FIG. 1 a data analysis pipeline for the guided breathing mode is illustrated.
  • the tomographic conductivity images (106) are computed by a standard EIT image reconstruction technique.
  • the series of conductivity images are spatially filtered to remove spatial noise and temporally filtered to reject signals that are not related to breathing (108) .
  • a voxel-wise breathing amplitude map is calculated from the filtered waveforms (110) . From the waveform at each voxel, a set of maximum conductivity changes are computed over a time-window that includes at least one breathing cycle.
  • the breathing amplitude at a specific voxel is defined as an average of the set of maximum conductivity changes.
  • a major breathing waveform (114) is extracted using feature selection (through filtering and signal extraction) over the breathing waveforms of the voxels with the largest breathing amplitude (112) .
  • the correlation between the major waveform and the voxel’s waveform are computed, resulting in a correlation map (116) .
  • the lung clusters are obtained by segmentation of the breathing amplitude map weighed with the correlation map (118) .
  • the breathing waveforms at the lung clusters are then analyzed to extract indicators.
  • the preferred data analysis pipeline for the spontaneous breathing mode is illustrated.
  • the filtered series of the conductivity images (208) is obtained as in the guided breathing mode following the steps (202-206) .
  • the lung clusters (118/210) obtained from the guided breathing test can be employed to analyze the waveforms within the lung regions.
  • the analysis of the waveforms within the lung regions is used to infer respiratory related parameters including but not limited to breathing amplitude, relative phase delay of different lung segments (voxels) , and spatial lung homogeneity (212) .
  • Quantitative measures of lung health are assessed by multiple order statistics of the aforementioned parameters at specific regions of interest (ROIs) (214) .
  • ROIs regions of interest
  • FIG 3 the preferred data analysis pipeline for the forced expiration and inspiration mode is illustrated.
  • the filtered series of the conductivity images (308) is obtained as in the guided breathing mode following the steps (302-306) .
  • the lung clusters (118/310) obtained from the guided breathing test can be employed to analyze the waveforms within the lung regions.
  • the conductivity waveforms at different voxels are regularized by fitting to a mathematical model, including but not limited to a generalized high order piecewise polynomial and a double logistic curve (312) .
  • the regularized waveforms induce a better estimation of the respiratory parameters.
  • the analysis of the conductivity waveforms and conductivity loops within the lung regions is used to infer respiratory parameters related to the maximum lung capacity and speed to inhale and exhale air (316) .
  • the forced expiration and inspiration follows the standard spirometry test protocol set by the American Thoracic Society. With appropriate calibration of the EIT measurement, voxel-wise standard spirometry indicators can be inferred.
  • Quantitative measures of lung health are assessed by multiple order statistics of the aforementioned parameters at specific regions of interest (ROIs) (318) .
  • the inventors of the present invention have designed and devised, through research, experiments and trials, a portable EIT system for use in lung function assessment.
  • FIG. 6 shows an embodiment of a portable EIT console in one embodiment of the invention.
  • the EIT console generally includes a power management module for constant power supply, a current generation module for alternating current generation, a signal distribution and readout module for current injection and voltage readout, a data acquisition module for potential difference measurement, amplification and acquisition, and a control and output module for module coordination, data processing and cloud-server communication.
  • the EIT console includes an EIT console having 5 major modules:
  • Current generation module hat primarily includes a digitally programmable analog sine wave generator and a constant current generator successively to generate an alternating current of 1 mApp and a voltage amplitude of 1 Vpp.
  • a low-pass filter is used to suppress total harmonic distortion and ambient electromagnetic interference (e.g., power line noise) .
  • C Signal distribution and readout module that introduces the generated current to the subject via the 16-electrode belt using a set of CMOS multiplexers (MUXs) .
  • MUXs CMOS multiplexers
  • Four MUXs are used, in which two MUXs are employed for current injection and the other two for voltage readout.
  • the MUXs are configured into the adjacent-scan pattern through the microcontroller unit (MCU) .
  • D Data acquisition module that is the analog front-end (AFE) that acquires, measures and amplifies the potential differences from the electrodes.
  • the AFE comprises a four-stage wide input differential amplifier with high common-mode rejection ratio (CMRR) , and a bandpass filter.
  • CMRR common-mode rejection ratio
  • Control and output module that includes an analog-to-digital converter (ADC) , a MCU and a wireless communication chip.
  • ADC analog-to-digital converter
  • MCU MCU
  • wireless communication chip The potential differences obtained from the data acquisition module are digitized by a 12-bit ADC, processed in the MCU unit, and transferred to the cloud server for image reconstruction and processing.
  • FIG. 7 shows a lung health assessment system in one embodiment.
  • the lung health assessment system includes, generally, a mobile app, a cloud server, and an EIT console, which may be the console of Figure 6.
  • the mobile app, cloud server, and EIT console cooperate to facilitate lung health assessment.
  • the EIT console is first connected to Wi-Fi, either automatically or manually. Users will then be instructed to wear the electrode belt, and electrode quality will be analyzed. Once a high signal-to-noise ratio (SNR) is achieved or a force-start is triggered, the lung function assessment will begin. Finally, the acquired raw data will be transferred to the image reconstruction and processing pipeline, and results will be displayed.
  • SNR signal-to-noise ratio
  • the mobile app guides users to connect the console and wear the electrode belt, instructs users to perform breathing paradigms, commands the console to acquire and transfer EIT-data for processing, and displays lung functional assessment results.
  • the acquired data is then denoised, and time-difference EIT images are reconstructed and processed to generate functional maps, and EIT-derived indicators are extracted.
  • FIG 8 An exemplary operation pipeline (that makes use of the EIT system) is shown in Figure 4A and a general processing flow (that makes use of the EIT system) in shown in Figure 8.
  • Figure 8 illustrates that global and regional indicators and functional maps are obtained through multiple processing stages.
  • a reference is first chosen from the measured voltage data then the raw data is denoised.
  • Time difference conductivity images are then reconstructed from all measured frames and from the reference data.
  • spatial and temporal filtering is applied to the series of reconstructed images from which global functional indicators and waveform are extracted.
  • the series of EIT images and the global waveform are used to define the regions of interest which are then used to extract regional indicators and functional maps.
  • simultaneous EIT and spirometry are applied with four different breathing paradigms to simulate a wide dynamic range, including a combination of full or mid capacity inhale, and fast or slow exhale (Figure 4B) .
  • Maximal volume engaged (MVE) exhaled volume in 1 second (EV1) , EV1/MVE ratio, maximum expiratory flow (MEF) , and expiratory flow at 25–75%of maximum volume engaged (EF25-75%) are calculated (further described below) .
  • EIT-derived indicators are significantly correlated with spirometry indicators across a wide dynamic range, demonstrating the portable EIT system has standard spirometry capabilities (Figure 4C) .
  • EIT parametric maps (Figure 4D) and regional EIT-derived indicators are consistent with the four corresponding breathing paradigms, signifying the portable EIT system allows regional lung function assessment.
  • Some embodiments of the invention also provide a guided breathing paradigm which includes a periodic inhalation and exhalation at a predetermined number of (e.g., 12) breaths per minute (Figure 5A) , and its corresponding processing pipeline (further described below) .
  • the amplitude maps, total amplitude, conductivity-time curve, and frequency spectra exhibited higher amplitude during deep breathing compared to shallow breathing, demonstrating this close-to-effortless method can reflect global and regional changes in lung function (Figures 5B and 10) . It is observed that the right lung has significantly more activated voxels and significantly higher total amplitude compared to the left (Figure 5B) , likely due to the positioning of the heart within the left thorax.
  • COVID-19 discharged subject is longitudinally monitored with two age-and gender-matched controls.
  • the COVID-19 discharged subject have higher coefficient of variation (C.V.; Figure 5C) , suggesting lung function deterioration.
  • C.V. significantly decreased across time in the left lung of the COVID-19 discharged subject ( Figure 5D) , suggesting a functional deterioration at the beginning followed by a recovery.
  • Regional analysis further pin-pointed the potential deterioration and recovery is in the anterior left lung ( Figure 5E) .
  • Figure 10 illustrates EIT with guided breathing paradigm in some embodiments can quantify lung functional changes.
  • Shallow and deep breathing modes were applied to test the sensitivity of this close-to-effortless breathing paradigm.
  • the right lung has significantly more activated voxels and significantly higher total amplitude compared to the left in both shallow and deep breathing modes, likely due to the positioning of the heart within the left thorax.
  • the total amplitude was higher during deep breathing compared to shallow breathing, while activated voxels remained similar.
  • ***p ⁇ 0.001. Error bars indicate ⁇ standard error of mean.
  • Figures 4A-4D generally illustrate that EIT or EIT system enables global and regional lung function assessment.
  • Figure 4A shows an EIT system in one embodiment including a portable console, electrode belt, mobile app interface, and cloud-based processing pipeline.
  • Raw EIT data is first acquired followed by denoising, time-difference EIT image reconstruction, and lung function mapping. Further analysis is applied to extract lung functional indicators.
  • Figure 4A shows four different breathing paradigms, including a combination of full or mid capacity inhale, and fast or slow exhale. Simultaneous EIT and spirometry were applied with these paradigms.
  • Figure 4C shows that EIT-derived indicators are significantly correlated with spirometry indicators across a wide dynamic range, demonstrating the EIT system has standard spirometry capabilities.
  • Figure 4D shows calculated parametric maps that allow regional lung function assessment.
  • MVE Maximal volume engaged
  • EV1 exhaled volume in 1 second
  • MEF maximum expiratory flow
  • EF25-75% expiratory flow at 25-75%of maximum volume engaged
  • Figures 5A-5E generally illustrate a guided breathing paradigm that can quantify global and regional lung functional changes in some embodiments.
  • Figure 5A shows that the paradigm is a periodic inhale and exhale pattern at 12 breaths per minute (bpm) .
  • Figure 5B shows shallow and deep breathing modes were applied to test the sensitivity of this method.
  • Activated voxels and total amplitude were extracted from the left and right lungs.
  • the right lung has significantly more activated voxels and significantly higher total amplitude compared to the left.
  • the amplitude maps, total amplitude, conductivity time curve, and frequency spectra exhibited higher amplitude during deep breathing compared to shallow breathing, while activated voxels remained similar.
  • Figures 5C-E show measurements for a COVID-19 discharged subject (longitudinally monitored) along with two age-and gender-matched controls.
  • Figure 5C shows the COVID-19 discharged subject had higher C.V., suggesting lung function deterioration. Note the right lung has significantly more activated voxels and significantly higher total amplitude.
  • Figure 5D shows C.V. significantly decreased across time in the left lung of the COVID-19 discharged subject, suggesting a functional deterioration at the beginning followed by a recovery.
  • Figure 5E shows regional analysis further pin-pointed deterioration and recovery in the anterior left lung. *p ⁇ 0.05, **p ⁇ 0.01, and ***p ⁇ 0.001. Error bars indicate ⁇ standard error of mean. Abbreviations: arbitrary unit (a.u. ) ; coefficient of variation (C.V. ) .
  • the EIT system includes an electrode lung belt for emitting and receiving electrical signals and a portable EIT console ( Figure 4A) .
  • the electrode belt includes an elastic band with 16 equally spaced carbon gel electrodes (BJD-A, Bestpad, Shenzhen, China) .
  • the elastic electrode belts come in sizes ranging from 65 cm –120 cm in length with extendable range of 10%of its original length. Sixteen 4 x 4 cm 2 electrodes are used for belts longer than or equal to 75 cm, while the same electrodes are trimmed to 3 x 4 cm 2 for belts shorter than 75 cm to avoid contact between adjacent electrodes.
  • the portable EIT console includes the following functional modules.
  • a power management module provides constant power supply to all other electronic modules through the power socket or the Li-ion battery.
  • a current generation module primarily includes a digitally programmable analog sine wave generator and a current generator successively to generate an alternating current (a.c) of 1 mApp and a voltage amplitude of 1 Vpp.
  • a lowpass filter is used to suppress total harmonic distortion and ambient electromagnetic interference (e.g., power line noise) .
  • a signal distribution and readout module introduces the generated current to the subject via the 16-electrode belt using a set of CMOS multiplexers (MUXs) .
  • MUXs CMOS multiplexers
  • a data acquisition module is the analog front-end (AFE) that acquires, measures, and amplifies the differential voltage from the electrodes.
  • the AFE comprises a four-stage wide input differential amplifier with high common-mode rejection ratio (CMRR) , and a bandpass filter.
  • a control and output module includes an analog-to-digital converter (ADC) , an MCU and a wireless communication chip.
  • ADC analog-to-digital converter
  • the differential voltage obtained from the data acquisition module are digitized by three 12-bit ADCs using triple interleaved ADC method to achieve higher sampling rate, processed in the MCU unit, and transferred to the cloud server for image reconstruction and processing.
  • an alternating current (AC) of frequency 35kHz is injected sequentially between all adjacent electrode pairs and the potential differences are measured across other 13 adjacent electrode pairs.
  • a data frame consisting of 208 (16 ⁇ 13) differential voltage measurements is acquired at a rate of 33 frames per second.
  • the observed signal-to-noise ratio (SNR) ranged from 45-55 dB.
  • two breathing paradigms are defined: a forced breathing paradigm and a guided breathing paradigm.
  • the subjects performed all breathing modes in the upright position.
  • the subjects are guided by the curves shown in Figure 4B to perform three shallow breaths at a rate of 18 cycles per minute, followed by an inhalation to their maximum (or mid) lung capacity. Subsequently, they are instructed to exhale fast and abruptly (or slowly) during an exhalation segment lasting for 6 seconds. At the end of the exhalation phase, the subjects are guided to inhale again to their maximum (or mid) lung capacity, followed by an exhalation to their normal breathing exhalation volume, and lastly another three guided shallow breaths at a rate of 18 cycles per minute. The full paradigm lasted about 26 seconds.
  • the middle section of the inhale and exhale have four variants resulting in four types of breathing efforts which we denoted as full inhale and fast exhale, full inhale and slow exhale, mid inhale and fast exhale, and mid inhale and slow exhale.
  • the subjects are instructed to breathe into and out of the spirometry device (Spirobank Smart, Medical International Research, Italy) via a mouthpiece to acquire the spirometry indicators and the volume-time curves during the forced exhalation phase.
  • the subjects are instructed to follow the breathing instructions on a screen for 60 seconds as shown in Figure 5A, which includes regular inhale-and-exhale cycles at a constant rate of 12 breaths per minute.
  • forced breathing paradigm and the guided breathing paradigm can be modified in other embodiments to have other pattern and/or duration.
  • a total of 26 subjects are recruited for testing the system/method embodiments of the invention. 14 subjects performed around 20 repetitions of forced breathing paradigm (Figure 4B-4D) , 9 subjects performed around 4 repetitions of guided breathing with deep and shallow breathing modes (Figure 5B) , and a COVID-19 discharged subject and two age-and gender-matched healthy controls performed guided breathing across a span of 10 days (Figure 5C-5E) .
  • the COVID-19 discharged subject conducted the tests by himself with a similar but separate system.
  • the current generation module of this system included a 12-bit digital-to-analog converter (DAC) and a current generator successively to generate an alternating current (ac) of amplitude 1-2 mApp.
  • the AFE comprised two non-inverting amplifiers and two 12-bit ADC for signal acquisition.
  • FVC forced vital capacity
  • FEV1 forced expiration volume in 1 second
  • FEV1/FVC ratio peak expiratory flow
  • PEF peak expiratory flow
  • FEF25–75% forced expiratory flow at 25–75%of forced vital capacity
  • the first type is the spirometry-like forced breathing paradigm, i.e., inhale to full capacity, and exhale as fast as possible. This is to obtain the largest possible values of MVE, EV1, EV1/MVE, MEF and EF25-75%.
  • the second type is to inhale to the full capacity but exhale slowly. This is to simulate a low EV1, EV1/MVE, MEF and EF25-75%, while MVE remains high.
  • the third type is to inhale to mid capacity, and exhale as fast as possible. This is to simulate a low MVE, and a high EV1/MVE.
  • the fourth type is to inhale to mid capacity and exhale slowly. This is to simulate a low MVE, EV1, EV1/MVE, MEF and EF25-75%.
  • Image reconstruction and processing steps in some embodiments are as follows.
  • the reference voltage data frame for time-difference EIT image reconstruction is set to the mean data frame across the full time-series after gain normalization.
  • Individual data frames are denoised by setting outlier voltage values above and below system thresholds to the corresponding value in the reference data frame.
  • NOSER Newton's one-step error reconstructor
  • the initially reconstructed time-series EIT images are interpolated into a new time-grid of regular sampling rate, resulting in a new series of timestamp-corrected EIT images. Further spatiotemporal filtering is applied on every 3D spatiotemporal image series to reduce temporal noise and spatial artifacts.
  • a 0.8s-wide moving average is used to filter the temporal waveforms at each voxel.
  • a 3rd order Butterworth filter of 0.083 –0.5 Hz passband is used to filter out the irrelevant signals embedded in temporal waveforms at each voxel, such as the cardiac related signals (i.e., 60 –80 beats per minute) , but to keep breathing related signals (i.e., 5 –30 breaths per minute) .
  • the filtered image series is then transformed from corresponding initial triangular simplices into a 64 ⁇ 64-voxel rectangular grid by using interpolation weights defined by a sigmoid function.
  • a global conductivity curve For each time-series of images, a global conductivity curve, an amplitude map and a correlation map are calculated.
  • the global conductivity curve represents the dominant conductivity signal due to breathing and is obtained by a weighted sum of the conductivity curves at all voxels.
  • the amplitude map represents the conductivity variation at different voxels.
  • the amplitude of the corresponding voxel is evaluated as the 50 th and 100 th percentile of the maximum change of conductivity at all time-segments for guided breathing and forced breathing, respectively.
  • the correlation map represents the voxel-wise correlation between the global conductivity-time curve and the respective conductivity-time curve.
  • additional global EIT indicators and functional maps corresponding to MVE, EV1, MVE/EV1, MEF and EF25-75% are calculated from the global conductivity curve and the voxel-wise conductivity curve.
  • the conductivity curve corresponding to the forced exhale and inhale is first extracted by excluding the first and last segments corresponding to the guided shallow breathing.
  • the starting time of the forced exhale is obtained using the back-extrapolation method as in standard spirometry data analysis 1 .
  • EIT indicators corresponding to MVE, EV1, MVE/EV1, MEF and EF25-75% are calculated as follows, (1) MVE is obtained by the difference between the maximum and minimum conductivity changes; (2) EV1is obtained by the difference between the conductivity change at the starting time of the exhalation and one second after; (3) EV1/MVE is the ratio of EV1 and MVE; (4) MEF is the maximum value of the time derivative of the conductivity curve; and (5) EF25-75%is the average flow during the expiration from 25%of MVE to 75%of MVE.
  • regions of interest are defined using a threshold-based segmentation of the amplitude-correlation product map.
  • the largest cluster with the most selected voxels localized in the left-half and right-half of the product image are considered as the left and right lung clusters, respectively.
  • a group threshold is applied onto all repetitions performed by the same subject in a single trial.
  • the group threshold value is determined by finding the mean of threshold values which selected the top 35%voxels with the largest amplitude-correlation product across repetitions of the fourth (weakest) breathing effort. This thresholding strategy is used to compare the activated voxels for different breathing effort.
  • the lung clusters are further divided into four ROIs, namely the anterior left, posterior left, anterior right and posterior right lungs.
  • the anterior-posterior division is defined by the horizontal line passing through the midpoint between the topmost and bottommost voxels in both lung clusters.
  • the left and right conductivity waveforms are obtained by averaging the temporal signals of voxels within both lung clusters.
  • the frequency spectra of the left-right waveforms are obtained by computing their fast Fourier transforms.
  • Regional EIT indicators including the number of activated voxels, total amplitude, and the coefficient of variation (CV) shown in Figure 5B-5E are computed from amplitude image.
  • the number of activated voxels is the count of voxels in the corresponding ROI.
  • the total amplitude is the sum of the voxel amplitude within different ROIs.
  • the CV is the ratio of the standard deviation over mean in the top 75%voxels within each ROI. CV are compensated for potential system-dependent SNR.
  • CV serves to monitor homogeneity of amplitudes as a significant consistent loss in homogeneity could imply decreased lung function.
  • additional regional indicators (MVE, EV1, EV1/MVE ratio, MEF and EF25-75%) are computed for each lung ROI, by averaging the corresponding functional indicator maps within each ROI.
  • Figures 9A and 9B more specifically illustrate that regional EIT-derived indicators were consistent with the four corresponding breathing paradigms which is the combination of full or mid capacity inhale and fast or slow exhale.
  • Figure 9A shows that individual functional image was used to define the regions of interest (ROIs) , including anterior left, posterior left, anterior right and posterior right.
  • ROIs regions of interest
  • Figure 9B shows that all regional indicators followed a similar trend, i.e., the MVE is higher in paradigms involving full capacity inhale, the EV1 is highest in full capacity inhale with fast exhale and is lowest in half capacity inhale with slow exhale, the EV1/MVE ratio is higher in paradigms involving fast exhale, the MEF is highest in full capacity inhale with fast exhale and is lowest in half capacity inhale with slow exhale, and the EF25-75%is highest in highest in full capacity inhale with fast exhale and is lowest in slow exhale.
  • trials are excluded by inspecting the correlation map and the global conductivity curve.
  • the accepted trials showed two positively correlated clusters in the correlation map (corresponding to two lung regions) and a global conductivity curve which followed the instructed curve.
  • a functional mapping to predict spirometry indicators from the global EIT indicators and anthropometrics is learnt from a sub-set (training set) of the collected data, then evaluated with the remaining sub-set (test set) .
  • the test-set is obtained by randomly excluding all data from two participants and another 10%of the data from the remaining participants.
  • the training-set is the remaining data from the global set after excluding the test-set.
  • the proportion of the test-set is 24%. This splitting strategy is adopted to verify whether the trained model is capable to generalize unseen data and subjects with different anthropometrics.
  • Figures 11 illustrate that spirometry indicators are predicted from EIT over a wide dynamic range for both train and test sets. These results demonstrated that the EIT system has standard spirometry capabilities.
  • the blue dots are the samples from the training set and the red dots are the samples from the testing set.
  • test-set is obtained from the global set by randomly excluding all data from two subjects and another 10%of the data from the remaining subjects.
  • the train-set is the remaining data from the global set after excluding the test-set.
  • the performance of the regression model is evaluated with the correlation coefficient ( ⁇ ) and mean absolute error percentage (MAE%) . The results show that the performance on the train-set and test-set are very similar which suggest that the model has a good generalization performance.
  • Table 1 Evaluation of the regression model used to predict the spirometry indicators.
  • Table 2 Coefficients associated with the different variables of the multiple linear regression model used to predict the spirometry indicators.
  • the amplitude maps are masked with the lung clusters.
  • the amplitude maps are normalized by the maximum voxel amplitude across all repetitions performed.
  • the average amplitude maps at shallow and deep breathing are obtained by computing the mean normalized amplitude maps at each breathing depth ( Figure 5B) .
  • each amplitude map is normalized by its maximum voxel amplitude, and the average amplitude maps are obtained by averaging across different subject groups (patient and healthy controls) ( Figure 5C) .
  • Two-way ANOVA followed by Bonferroni multiple comparisons is applied to compare the number of activated voxels and total amplitude between the left and right lungs across breathing depth or patient/control group ( Figure 5B-5C) .
  • the regional waveforms and frequency spectra are normalized by their maximum value across all trials, and the mean normalized breathing waveforms and frequency spectra are presented with ⁇ SEM ( Figure 5B) .
  • the mean left and right CV in the patient and healthy control groups are presented ( Figure 2C).
  • Two-way ANOVA followed by Bonferroni multiple comparisons is applied to compare the left and right lung CV between subject groups.
  • computing system any appropriate computing system architecture may be utilized. This will include stand-alone computers, network computers, dedicated or non-dedicated hardware devices. Where the terms “computing system” and “computing device” are used, these terms are intended to include (but not limited to) any appropriate arrangement of computer or information processing hardware capable of implementing the function described.
  • the invention has provided a computer-implemented method for lung health assessment, comprising: receiving EIT data associated with a lung of a subject; and processing the received EIT data to determine a health condition of the lung of the subject.
  • the EIT data is raw EIT data.
  • the EIT data is obtained from the subject or the subject’s lung during a guided breathing operation.
  • the subject periodically or regularly inhales and exhales, optionally with a predetermined number of breaths per minute.
  • the EIT data is obtained from the subject or the subject’s lung during a forced breathing operation.
  • the subject in the forced breathing operation, the subject first takes a predetermined number of shallow breaths, then inhales to maximum or medium lung capacity, then exhales abruptly or slowly, then inhales to maximum or medium lung capacity, then exhales to normal breathing exhalation volume, and finally takes a predetermined number of shallow breaths.
  • the EIT data is obtained from the subject or the subject’s lung during a spontaneous breathing operation.
  • the spontaneous breathing operation is an unguided operation in which the subject can breathe unguided.
  • the processing comprises: denoising the received EIT data; and reconstructing time-difference EIT images based on the denoised EIT data.
  • the denoising comprises: filtering the EIT data; and/or adjusting outlier measurements in the EIT data based on reference voltage data.
  • the reconstruction is performed using a Gauss–Newton method, such as a one-step linear Gauss-Newton solver optionally with a regularization matrix based on the Newton's one-step error re-constructor prior.
  • the processing further comprises, for each time series of the time-difference EIT images: determining a global conductivity measure or curve that represents a dominant conductivity signal in the time series due to breathing of the subject; determining an amplitude measure or map that represents a local conductivity variation at different voxels of the images, and/or determining a correlation measure or map that represents voxel-wise correlation between the global conductivity-time curve and the respective local conductivity-time curve.
  • the global conductivity measure or curve is determined based on feature selection through filtering and signal extraction.
  • the processing further comprises analyzing the global conductivity curve, the amplitude map, and/or the correlation map to determine the health condition of the lung of the subject.
  • the processing further comprises, for each time series of the time-difference EIT images: determining one or more parametric functional indicator measures or maps, each of which is related to a respective indicator measure.
  • the one or more parametric functional indicator measures or maps are determined based on the global conductivity curve and a voxel-wise conductivity curve.
  • the indicator measure is spirometry indicator measure
  • the one or more parametric functional indicator comprise: maximum volume engaged (MVE) corresponding to forced vital capacity (FVC) measure; exhaled volume in 1 second (EV1) corresponding to forced expiration volume in 1 second (FEV1) measure; exhaled volume in 1 second (EV1) to maximum volume engaged (MVE) ratio corresponding to forced expiration volume in 1 second (FEV1) to forced vital capacity (FVC) ratio measure; maximum expiration flow (MVE) corresponding to peak expiratory flow (PEF) measure; and/or expiratory flow at 25-75%of maximum volume engaged (EF25-75%) corresponding to forced expiratory flow at 25–75%of forced vital capacity (FEF25–75%) measure.
  • MVE maximum volume engaged
  • FVC forced vital capacity
  • the processing further comprises: processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine a spirometry indicator measure of the subject.
  • This spirometry indicator measure is determined for EIT data obtained using the forced breathing operation.
  • the processing further comprises: processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine an indicator of the subject indicative of lung condition, the indicator comprises amplitude, activated voxels, and /or coefficient of variation, which are non-spirometry indicator.
  • the non-spirometry indicator is determined for EIT data obtained using the guided breathing operation and spontaneous breathing operation.
  • the invention has also provided a system and a non-transitory computer-readable medium for implementing the above method.
  • the method of the invention can be implemented using other EIT signal measurement system, which may or may not be portable, and may or may not be the same as the system specifically illustrated in the above embodiments.

Abstract

A computer-implemented method for lung health assessment comprises: receiving EIT data associated with a lung of a subject; and processing the received EIT data to determine a health condition of the lung of the subject. A system for lung health assessment comprises one or more processors arranged to: receiving EIT data associated with a lung of a subject; and processing the received EIT data to determine a health condition of the lung of the subject.

Description

ELECTRICAL IMPEDANCE TOMOGRAPHY BASED LUNG ASSESSMENT TECHNICAL FIELD
The invention relates to systems and methods for analyzing electrical impedance tomography (EIT) data for lung health assessment. The invention can be implemented as a diagnostic tool.
BACKGROUND
Lung diseases can be irreversible, and early diagnosis of these diseases may not be readily available. For example, chronic obstructive pulmonary disease (COPD) is a progressive disease that is often diagnosed at the severe stage. Conventional clinical assessment and pulmonary function tests typically utilizes spirometry and CT scans. These methods lack regional lung functional assessment and may not be sensitive to early-stages of lung diseases. Furthermore, these existing diagnosing techniques are mostly non-portable, require skilled operators, expensive, and/or harmful.
SUMMARY OF THE INVENTION
In a first aspect there is provided a method /technique for calculating or determining voxel-wise respiratory-related parameters to assess the lungs health of the users. The method comprises: analyzing EIT signals measured at, at least, one or more or all of the following breathing modes: (1) guided breathing, (2) spontaneous breathing, and (3) forced expiration and inspiration. The guided breathing mode is used for calibrating the measurement system and extracting lung clusters. The guided breathing and spontaneous breathing are used to infer respiratory parameters such as the breathing amplitude, relative phase delay of different lung segments (voxels) , spatial lung homogeneity, etc. The forced expiration and inspiration breathing mode is used to infer respiratory parameters related to the maximum lung capacity and speed to inhale and exhale air. Preferably, the forced expiration and inspiration follows the standard spirometry test protocol set by the American Thoracic Society. With appropriate calibration of the EIT measurement, voxel-wise standard spirometry indicators can be inferred. Additional breathing mode (s) may be used.
In a second aspect there is provided a system for determining or calculating respiratory-related parameters to assess the lungs health of the users. The system comprising: means for measuring EIT signals at, at least, three breathing modes: (1) guided breathing mode, (2) spontaneous breathing mode, and (3) forced expiration and inspiration mode. The system further comprises means configured to implement techniques for inferring activated lungs regions, quantifying respiratory-related indicators and their spatial distribution across the lungs.
In a third aspect there is provided a computer-implemented method for lung health assessment, comprising: receiving EIT data associated with a lung of a subject; and processing the received EIT data to determine a health condition of the lung of the subject.
Optionally, the EIT data is raw EIT data.
Optionally, the EIT data is obtained from the subject or the subject’s lung during a guided breathing operation.
Optionally, in the guided breathing operation, the subject periodically or regularly inhales and exhales, optionally with a predetermined number of breaths per minute.
Optionally, the EIT data is obtained from the subject or the subject’s lung during a forced breathing operation.
Optionally, in the forced breathing operation, the subject first takes a predetermined number of shallow breaths, then inhales to maximum or medium lung capacity, then exhales abruptly or slowly, then inhales to maximum or medium lung capacity, then exhales to normal breathing exhalation volume, and finally takes a predetermined number of shallow breaths.
Optionally, the EIT data is obtained from the subject or the subject’s lung during a spontaneous breathing operation. Unlike the guided breathing operation and the forced breathing operation in which the subject breaths according to predetermined patterns, the spontaneous breathing operation is an unguided operation in which the subject can breathe unguided.
Optionally, the processing comprises: denoising the received EIT data; and reconstructing time-difference EIT images based on the denoised EIT data.
Optionally, the denoising comprises: filtering the EIT data; and/or adjusting outlier measurements in the EIT data based on reference voltage data.
Optionally, the reconstruction is performed using a Gauss–Newton method, such as a one-step linear Gauss-Newton solver optionally with a regularization matrix based on the Newton's one-step error re-constructor prior.
Optionally, the processing further comprises, for each time series of the time-difference EIT images: determining a global conductivity measure or curve that represents a dominant conductivity signal in the time series due to breathing of the subject; determining an amplitude measure or map that represents a local conductivity variation at different voxels of the images, and/or determining a correlation measure or map that represents voxel-wise correlation between the global conductivity-time curve and the respective local conductivity-time curve.
Optionally, the global conductivity measure or curve is determined based on feature selection through filtering and signal extraction.
Optionally, the processing further comprises analyzing the global conductivity curve, the amplitude map, and/or the correlation map to determine the health condition of the lung of the subject.
Optionally, the processing further comprises, for each time series of the time-difference EIT images: determining one or more parametric functional indicator measures or maps, each of which is related to a respective indicator measure.
Optionally, the one or more parametric functional indicator measures or maps are determined based on the global conductivity curve and a voxel-wise conductivity curve.
Optionally, the indicator measure is spirometry indicator measure; and the one or more parametric functional indicator comprise: maximum volume engaged (MVE) corresponding to forced vital capacity (FVC) measure; exhaled volume in 1 second (EV1) corresponding to forced expiration volume in 1 second (FEV1) measure; exhaled volume in 1 second (EV1) to maximum volume engaged (MVE) ratio corresponding to forced expiration volume in 1 second (FEV1) to forced vital capacity (FVC) ratio measure; maximum expiration flow (MVE) corresponding to peak expiratory flow (PEF) measure; and/or expiratory flow at 25-75%of maximum volume engaged (EF25-75%) corresponding to forced expiratory flow at 25–75%of forced vital capacity (FEF25–75%) measure.
Optionally, the processing further comprises: processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine a spirometry indicator measure of the subject. This spirometry indicator measure is determined for EIT data obtained using the forced breathing operation.
Optionally, the processing further comprises: processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine an indicator of the subject indicative of lung condition, the indicator comprises amplitude, activated voxels, and /or coefficient of variation, which are non-spirometry indicator. The non-spirometry indicator is determined for EIT data obtained using the guided breathing operation or spontaneous breathing operation.
In a fourth aspect there is provided a system for lung health assessment, comprising one or more processors arranged to: receive EIT data associated with a lung of a subject; and process the received EIT data to determine a health condition of the lung of the subject. The one or more processors may be arranged to perform the method of the third aspect. Optionally the system further comprises a display for displaying the processing results.
In a fifth aspect there is provided non-transitory computer-readable medium comprising instructions which, when executed by one or more processors, causes the one or more processors to perform the method of the third aspect.
Other features and aspects of the invention will become apparent by consideration of the detailed description and accompanying drawings. Any feature (s) described herein in relation to one aspect or embodiment may be combined with any other feature (s) described herein in relation to any other aspect or embodiment as appropriate and applicable.
BRIEF DESCRIPTION OF THE DRAWINGS
Embodiments of the invention will now be described, by way of example, with reference to the accompanying drawings in which:
Figure 1 is a schematic diagram of a data analysis pipeline for guided breathing mode in one embodiment.
Figure 2 is a schematic diagram of a data analysis pipeline for spontaneous breathing mode in one embodiment.
Figure 3 is a schematic diagram of a data analysis pipeline for forced expiration and inspiration mode in one embodiment.
Figure 4A is a schematic diagram illustrating a method for lung health assessment in one embodiment.
Figure 4B is a schematic diagram illustrating four guided breathing paradigms, including a combination of full or mid capacity inhale, and fast or slow exhale, under the forced breathing mode in some embodiments.
Figure 4C contain graphs showing the correlating of EIT-derived parameters and spirometry indicators in some embodiments. (Abbreviations: Maximal volume engaged (MVE) , exhaled volume in 1 second (EV1) , maximum expiratory flow (MEF) , expiratory flow at 25-75%of maximum volume engaged (EF25-75%) .
Figure 4D contain parametric maps of the EIT-derived spirometry indicators in some embodiments. Figure 5A is a schematic diagram illustrating a method for lung health assessment in one embodiment.
Figure 5B contain graphs and plots showing the plots of shallow and deep breathing modes in one embodiment.
Figure 5C contain graphs and plots showing measurements taken from COVID-19 discharged subject and controls.
Figure 5D contain graphs and plots showing measurements taken from COVID-19 discharged subject and controls.
Figure 5E contain graphs and plots showing measurements taken from COVID-19 discharged subject and controls.
Figure 6 is a schematic block diagram of a portable EIT console in one embodiment.
Figure 7 is a schematic block diagram of a lung health assessment system in one embodiment.
Figure 8 is a flow diagram illustrating a general processing method for processing EIT lung data for lung health assessment in one embodiment.
Figure 9A contain plots illustrating the definition of lung ROI in one embodiment.
Figure 9B contain graphs illustrating regional EIT-derived indicators in one embodiment.
Figure 10 contain graphs illustrating measurements taken from shallow and deep breathing modes.
Figure 11 contain graphs of both training and testing sets of the EIT-derived spirometry indicators in some embodiments.
DETAILED DESCRIPTION
Some embodiments of the present invention are to measure EIT signals for three breathing modes: (1) guided breathing, (2) spontaneous breathing, and/or (3) forced expiration and inspiration.
In Figure 1, a data analysis pipeline for the guided breathing mode is illustrated. Once EIT measurements are acquired at one or multiple frequencies (102) , the tomographic conductivity images (106) , relative to a specific chosen reference average frame (104) , are computed by a standard EIT image reconstruction technique. The series of conductivity images are spatially filtered to remove spatial noise and temporally filtered to reject signals that are not related to breathing (108) . A voxel-wise breathing amplitude map is calculated from the filtered waveforms (110) . From the waveform at each voxel, a set of maximum conductivity changes are computed over a time-window that includes at least one breathing cycle. The breathing amplitude at a specific voxel is defined as an average of the set of maximum conductivity changes. A major breathing waveform (114) is extracted using feature selection (through filtering and signal extraction) over the breathing waveforms of the voxels with the largest breathing amplitude (112) . The correlation between the major waveform and the voxel’s waveform are computed, resulting in a correlation map (116) . The lung clusters are obtained by segmentation of the breathing amplitude map weighed with the correlation map (118) . The breathing waveforms at the lung clusters are then analyzed to extract indicators.
In Figure 2, the preferred data analysis pipeline for the spontaneous breathing mode is illustrated. The filtered series of the conductivity images (208) is obtained as in the guided breathing mode following the steps (202-206) . The lung clusters (118/210) obtained from the guided breathing test can be employed to analyze the waveforms within the lung regions. The analysis of the waveforms within the lung regions is used to infer respiratory related parameters including but not limited to breathing amplitude, relative phase delay of different lung segments (voxels) , and spatial lung homogeneity (212) . Quantitative measures of lung health are assessed by multiple order statistics of the aforementioned parameters at specific regions of interest (ROIs) (214) .
In Figure 3, the preferred data analysis pipeline for the forced expiration and inspiration mode is illustrated. The filtered series of the conductivity images (308) is obtained as in the guided breathing mode following the steps (302-306) . The lung clusters (118/310) obtained from the guided breathing test can be employed to analyze the waveforms within the lung regions. The conductivity waveforms at different voxels are regularized by fitting to a mathematical model, including but not limited to a generalized high order piecewise polynomial and a double logistic curve (312) . The regularized waveforms induce a better estimation of the respiratory parameters. The rate of change of conductivity obtained by the time-derivative of the regularized conductivity waveforms and is combined with the conductivity curve to obtain a conductivity vs rate of change of conductivity curves referred to as conductivity loop (314) . The analysis of the conductivity waveforms and conductivity loops within the lung regions is used to infer respiratory parameters related to the maximum lung capacity and speed to inhale and exhale air (316) . Preferably, the forced expiration and inspiration follows the standard spirometry test protocol set by the American Thoracic Society. With appropriate calibration of the EIT measurement, voxel-wise standard spirometry indicators can be inferred. Quantitative measures of lung health are assessed by multiple order statistics of the aforementioned parameters at specific regions of interest (ROIs) (318) .
The above method embodiment can be implemented using the system disclosed in U.S. Non-Provisional Patent Application No. 16/976,542, the entire contents of which is incorporated herein by reference (by choosing an appropriate form factor, the wrist band disclosed in US16/976,542 can be implemented as a wearable EIT band/belt/harness for lung EIT data collection) .
Overview
The inventors of the present invention have designed and devised, through research, experiments and trials, a portable EIT system for use in lung function assessment.
Figure 6 shows an embodiment of a portable EIT console in one embodiment of the invention. The EIT console generally includes a power management module for constant power supply, a current generation module for alternating current generation, a signal distribution and readout module for current injection and voltage readout, a data acquisition module for potential difference measurement, amplification and acquisition, and a control and output module for module coordination, data processing and cloud-server communication. As illustrated in Figure 6, the EIT console includes an EIT console having 5 major modules:
(A) Power management module that provides constant power supply to all other modules through the power socket or the Li-ion battery.
(B) Current generation module hat primarily includes a digitally programmable analog sine wave generator and a constant current generator successively to generate an alternating current of 1 mApp and a voltage amplitude of 1 Vpp. A low-pass filter is used to suppress total harmonic distortion and ambient electromagnetic interference (e.g., power line noise) .
(C) Signal distribution and readout module that introduces the generated current to the subject via the 16-electrode belt using a set of CMOS multiplexers (MUXs) . Four MUXs are used, in  which two MUXs are employed for current injection and the other two for voltage readout. The MUXs are configured into the adjacent-scan pattern through the microcontroller unit (MCU) .
(D) Data acquisition module that is the analog front-end (AFE) that acquires, measures and amplifies the potential differences from the electrodes. The AFE comprises a four-stage wide input differential amplifier with high common-mode rejection ratio (CMRR) , and a bandpass filter.
(E) Control and output module that includes an analog-to-digital converter (ADC) , a MCU and a wireless communication chip. The potential differences obtained from the data acquisition module are digitized by a 12-bit ADC, processed in the MCU unit, and transferred to the cloud server for image reconstruction and processing.
Figure 7 shows a lung health assessment system in one embodiment. The lung health assessment system includes, generally, a mobile app, a cloud server, and an EIT console, which may be the console of Figure 6. The mobile app, cloud server, and EIT console cooperate to facilitate lung health assessment. In one embodiment the EIT console is first connected to Wi-Fi, either automatically or manually. Users will then be instructed to wear the electrode belt, and electrode quality will be analyzed. Once a high signal-to-noise ratio (SNR) is achieved or a force-start is triggered, the lung function assessment will begin. Finally, the acquired raw data will be transferred to the image reconstruction and processing pipeline, and results will be displayed. In one embodiment, the mobile app guides users to connect the console and wear the electrode belt, instructs users to perform breathing paradigms, commands the console to acquire and transfer EIT-data for processing, and displays lung functional assessment results. The acquired data is then denoised, and time-difference EIT images are reconstructed and processed to generate functional maps, and EIT-derived indicators are extracted.
An exemplary operation pipeline (that makes use of the EIT system) is shown in Figure 4A and a general processing flow (that makes use of the EIT system) in shown in Figure 8. Specifically Figure 8 illustrates that global and regional indicators and functional maps are obtained through multiple processing stages. A reference is first chosen from the measured voltage data then the raw data is denoised. Time difference conductivity images are then reconstructed from all measured frames and from the reference data. Subsequently, spatial and temporal filtering is applied to the series of reconstructed images from which global functional indicators and waveform are extracted. The series of EIT images and the global waveform are used to define the regions of interest which are then used to extract regional indicators and functional maps.
To verify that the above system can assess lung function, in some examples simultaneous EIT and spirometry are applied with four different breathing paradigms to simulate a wide dynamic range, including a combination of full or mid capacity inhale, and fast or slow exhale (Figure 4B) . Maximal volume engaged (MVE) , exhaled volume in 1 second (EV1) , EV1/MVE ratio, maximum expiratory flow (MEF) , and expiratory flow at 25–75%of maximum volume engaged (EF25-75%) are calculated (further described below) . These EIT-derived indicators are significantly correlated with spirometry indicators across a wide dynamic range, demonstrating the portable EIT system has standard spirometry capabilities (Figure 4C) . Beyond global aerodynamics, EIT parametric maps  (Figure 4D) and regional EIT-derived indicators (Figures 9A and 9B) are consistent with the four corresponding breathing paradigms, signifying the portable EIT system allows regional lung function assessment.
Some embodiments of the invention also provide a guided breathing paradigm which includes a periodic inhalation and exhalation at a predetermined number of (e.g., 12) breaths per minute (Figure 5A) , and its corresponding processing pipeline (further described below) . The amplitude maps, total amplitude, conductivity-time curve, and frequency spectra exhibited higher amplitude during deep breathing compared to shallow breathing, demonstrating this close-to-effortless method can reflect global and regional changes in lung function (Figures 5B and 10) . It is observed that the right lung has significantly more activated voxels and significantly higher total amplitude compared to the left (Figure 5B) , likely due to the positioning of the heart within the left thorax. To further characterize the system with this close-to-effortless paradigm, a COVID-19 discharged subject is longitudinally monitored with two age-and gender-matched controls. The COVID-19 discharged subject have higher coefficient of variation (C.V.; Figure 5C) , suggesting lung function deterioration. Furthermore, C.V. significantly decreased across time in the left lung of the COVID-19 discharged subject (Figure 5D) , suggesting a functional deterioration at the beginning followed by a recovery. Regional analysis further pin-pointed the potential deterioration and recovery is in the anterior left lung (Figure 5E) .
More specifically Figure 10 illustrates EIT with guided breathing paradigm in some embodiments can quantify lung functional changes. Shallow and deep breathing modes were applied to test the sensitivity of this close-to-effortless breathing paradigm. The right lung has significantly more activated voxels and significantly higher total amplitude compared to the left in both shallow and deep breathing modes, likely due to the positioning of the heart within the left thorax. The total amplitude was higher during deep breathing compared to shallow breathing, while activated voxels remained similar. ***p < 0.001. Error bars indicate ± standard error of mean. Abbreviations: arbitrary unit (a.u. ) .
Further technical details of some embodiments of the invention are provided below.
Figures 4A-4D generally illustrate that EIT or EIT system enables global and regional lung function assessment. Figure 4A shows an EIT system in one embodiment including a portable console, electrode belt, mobile app interface, and cloud-based processing pipeline. Raw EIT data is first acquired followed by denoising, time-difference EIT image reconstruction, and lung function mapping. Further analysis is applied to extract lung functional indicators. Figure 4A shows four different breathing paradigms, including a combination of full or mid capacity inhale, and fast or slow exhale. Simultaneous EIT and spirometry were applied with these paradigms. Figure 4C shows that EIT-derived indicators are significantly correlated with spirometry indicators across a wide dynamic range, demonstrating the EIT system has standard spirometry capabilities. Figure 4D shows calculated parametric maps that allow regional lung function assessment. Abbreviations: Maximal volume engaged (MVE) , exhaled volume in 1 second (EV1) , maximum expiratory flow (MEF) , expiratory flow at 25-75%of maximum volume engaged (EF25-75%) .
Figures 5A-5E generally illustrate a guided breathing paradigm that can quantify global and regional lung functional changes in some embodiments. Figure 5A shows that the paradigm is a periodic inhale and exhale pattern at 12 breaths per minute (bpm) . Figure 5B shows shallow and deep breathing modes were applied to test the sensitivity of this method. Activated voxels and total amplitude were extracted from the left and right lungs. The right lung has significantly more activated voxels and significantly higher total amplitude compared to the left. The amplitude maps, total amplitude, conductivity time curve, and frequency spectra exhibited higher amplitude during deep breathing compared to shallow breathing, while activated voxels remained similar. Figures 5C-E show measurements for a COVID-19 discharged subject (longitudinally monitored) along with two age-and gender-matched controls. Figure 5C shows the COVID-19 discharged subject had higher C.V., suggesting lung function deterioration. Note the right lung has significantly more activated voxels and significantly higher total amplitude. Figure 5D shows C.V. significantly decreased across time in the left lung of the COVID-19 discharged subject, suggesting a functional deterioration at the beginning followed by a recovery. Figure 5E shows regional analysis further pin-pointed deterioration and recovery in the anterior left lung. *p < 0.05, **p < 0.01, and ***p < 0.001. Error bars indicate ± standard error of mean. Abbreviations: arbitrary unit (a.u. ) ; coefficient of variation (C.V. ) .
Hardware
In one embodiment, as shown in Figure 4A, the EIT system includes an electrode lung belt for emitting and receiving electrical signals and a portable EIT console (Figure 4A) . In one example the electrode belt includes an elastic band with 16 equally spaced carbon gel electrodes (BJD-A, Bestpad, Shenzhen, China) . The elastic electrode belts come in sizes ranging from 65 cm –120 cm in length with extendable range of 10%of its original length. Sixteen 4 x 4 cm 2 electrodes are used for belts longer than or equal to 75 cm, while the same electrodes are trimmed to 3 x 4 cm 2 for belts shorter than 75 cm to avoid contact between adjacent electrodes.
In one embodiment, as shown in Figure 6, the portable EIT console includes the following functional modules. A power management module provides constant power supply to all other electronic modules through the power socket or the Li-ion battery. A current generation module primarily includes a digitally programmable analog sine wave generator and a current generator successively to generate an alternating current (a.c) of 1 mApp and a voltage amplitude of 1 Vpp. A lowpass filter is used to suppress total harmonic distortion and ambient electromagnetic interference (e.g., power line noise) . A signal distribution and readout module introduces the generated current to the subject via the 16-electrode belt using a set of CMOS multiplexers (MUXs) . Four MUXs are used, in which two MUXs are employed for current injection and the other two for voltage readout. The MUXs are configured into the adjacent-scan pattern through the microcontroller unit (MCU) . A data acquisition module is the analog front-end (AFE) that acquires, measures, and amplifies the differential voltage from the electrodes. The AFE comprises a four-stage wide input differential amplifier with high common-mode rejection ratio (CMRR) , and a bandpass filter. A control and output module includes an analog-to-digital converter (ADC) , an MCU and a wireless communication chip. The differential voltage obtained from the data acquisition module are digitized by three 12-bit ADCs using triple interleaved ADC method to  achieve higher sampling rate, processed in the MCU unit, and transferred to the cloud server for image reconstruction and processing.
In some embodiments, an alternating current (AC) of frequency 35kHz is injected sequentially between all adjacent electrode pairs and the potential differences are measured across other 13 adjacent electrode pairs. A data frame consisting of 208 (16 × 13) differential voltage measurements is acquired at a rate of 33 frames per second. The observed signal-to-noise ratio (SNR) ranged from 45-55 dB.
In some other embodiments, other frequency, frame rate (e.g., 50) , etc., could be used, depending on applications.
Breathing paradigms
In some embodiments two breathing paradigms are defined: a forced breathing paradigm and a guided breathing paradigm. The subjects performed all breathing modes in the upright position.
For the forced breathing paradigm in this embodiment, the subjects are guided by the curves shown in Figure 4B to perform three shallow breaths at a rate of 18 cycles per minute, followed by an inhalation to their maximum (or mid) lung capacity. Subsequently, they are instructed to exhale fast and abruptly (or slowly) during an exhalation segment lasting for 6 seconds. At the end of the exhalation phase, the subjects are guided to inhale again to their maximum (or mid) lung capacity, followed by an exhalation to their normal breathing exhalation volume, and lastly another three guided shallow breaths at a rate of 18 cycles per minute. The full paradigm lasted about 26 seconds. The middle section of the inhale and exhale have four variants resulting in four types of breathing efforts which we denoted as full inhale and fast exhale, full inhale and slow exhale, mid inhale and fast exhale, and mid inhale and slow exhale. At the same time, the subjects are instructed to breathe into and out of the spirometry device (Spirobank Smart, Medical International Research, Italy) via a mouthpiece to acquire the spirometry indicators and the volume-time curves during the forced exhalation phase.
For the guided breathing paradigm in this embodiment, the subjects are instructed to follow the breathing instructions on a screen for 60 seconds as shown in Figure 5A, which includes regular inhale-and-exhale cycles at a constant rate of 12 breaths per minute.
It should be appreciated that the forced breathing paradigm and the guided breathing paradigm can be modified in other embodiments to have other pattern and/or duration.
Subjects
A total of 26 subjects are recruited for testing the system/method embodiments of the invention. 14 subjects performed around 20 repetitions of forced breathing paradigm (Figure 4B-4D) , 9 subjects performed around 4 repetitions of guided breathing with deep and shallow breathing modes (Figure 5B) , and a COVID-19 discharged subject and two age-and gender-matched healthy controls  performed guided breathing across a span of 10 days (Figure 5C-5E) . The COVID-19 discharged subject conducted the tests by himself with a similar but separate system. The current generation module of this system included a 12-bit digital-to-analog converter (DAC) and a current generator successively to generate an alternating current (ac) of amplitude 1-2 mApp. The AFE, comprised two non-inverting amplifiers and two 12-bit ADC for signal acquisition.
For the forced breathing paradigm, subjects performed variable efforts to simulate different forced vital capacity (FVC) , forced expiration volume in 1 second (FEV1) , FEV1/FVC ratio, peak expiratory flow (PEF) and forced expiratory flow at 25–75%of forced vital capacity (FEF25–75%) . From here-on the simulated spirometry indicators are referred to as maximal volume engaged (MVE) , exhaled volume in 1 second (EV1) , EV1/MVE ratio, maximum expiratory flow (MEF) , and expiratory flow at 25–75%of maximum volume engaged (EF25-75%) , corresponding to FVC, FEV1, FEV1/FVC, PEF, FEF25-75%respectively. These subjects followed 4 different types of instructions and performed 5 repetitions for each type of instruction. The first type is the spirometry-like forced breathing paradigm, i.e., inhale to full capacity, and exhale as fast as possible. This is to obtain the largest possible values of MVE, EV1, EV1/MVE, MEF and EF25-75%. The second type is to inhale to the full capacity but exhale slowly. This is to simulate a low EV1, EV1/MVE, MEF and EF25-75%, while MVE remains high. The third type is to inhale to mid capacity, and exhale as fast as possible. This is to simulate a low MVE, and a high EV1/MVE. The fourth type is to inhale to mid capacity and exhale slowly. This is to simulate a low MVE, EV1, EV1/MVE, MEF and EF25-75%.
Image reconstruction and processing
Image reconstruction and processing steps in some embodiments are as follows.
The reference voltage data frame for time-difference EIT image reconstruction is set to the mean data frame across the full time-series after gain normalization. Individual data frames are denoised by setting outlier voltage values above and below system thresholds to the corresponding value in the reference data frame. The EIT images are reconstructed from the denoised data frames and reference voltage data frame using one-step linear Gauss-Newton solver, with a regularization matrix based on the Newton's one-step error reconstructor (NOSER) prior with p = 0.35 and a regularization hyperparameter of λ 2 = 0.005. In some other embodiments, other p or other regularization hyperparameter, could be used, depending on applications.
As the sampling rate is not strictly consistent during data acquisition, the initially reconstructed time-series EIT images are interpolated into a new time-grid of regular sampling rate, resulting in a new series of timestamp-corrected EIT images. Further spatiotemporal filtering is applied on every 3D spatiotemporal image series to reduce temporal noise and spatial artifacts.
For the forced breathing paradigm, a 0.8s-wide moving average is used to filter the temporal waveforms at each voxel. For the guided breathing paradigm, a 3rd order Butterworth filter of 0.083 –0.5 Hz passband is used to filter out the irrelevant signals embedded in temporal waveforms at each voxel, such as the cardiac related signals (i.e., 60 –80 beats per minute) , but to keep breathing related signals (i.e., 5 –30 breaths per minute) . The filtered image series is then  transformed from corresponding initial triangular simplices into a 64×64-voxel rectangular grid by using interpolation weights defined by a sigmoid function.
For each time-series of images, a global conductivity curve, an amplitude map and a correlation map are calculated. The global conductivity curve represents the dominant conductivity signal due to breathing and is obtained by a weighted sum of the conductivity curves at all voxels. The amplitude map represents the conductivity variation at different voxels. The time series at each voxel is first partitioned into segments of duration T d seconds and a stride of T s seconds. For forced breathing, T d=6s and T s=6s, whereas for guided breathing T d=15s and T s=3s. The amplitude of the corresponding voxel is evaluated as the 50 th and 100 th percentile of the maximum change of conductivity at all time-segments for guided breathing and forced breathing, respectively. The correlation map represents the voxel-wise correlation between the global conductivity-time curve and the respective conductivity-time curve.
For the forced breathing paradigm, additional global EIT indicators and functional maps corresponding to MVE, EV1, MVE/EV1, MEF and EF25-75%are calculated from the global conductivity curve and the voxel-wise conductivity curve. The conductivity curve corresponding to the forced exhale and inhale is first extracted by excluding the first and last segments corresponding to the guided shallow breathing. The starting time of the forced exhale is obtained using the back-extrapolation method as in standard spirometry data analysis 1. EIT indicators corresponding to MVE, EV1, MVE/EV1, MEF and EF25-75%are calculated as follows, (1) MVE is obtained by the difference between the maximum and minimum conductivity changes; (2) EV1is obtained by the difference between the conductivity change at the starting time of the exhalation and one second after; (3) EV1/MVE is the ratio of EV1 and MVE; (4) MEF is the maximum value of the time derivative of the conductivity curve; and (5) EF25-75%is the average flow during the expiration from 25%of MVE to 75%of MVE.
Regional analysis
In some embodiments regional analysis is performed. In some embodiments, regions of interest (ROIs) are defined using a threshold-based segmentation of the amplitude-correlation product map. The largest cluster with the most selected voxels localized in the left-half and right-half of the product image are considered as the left and right lung clusters, respectively. For forced breathing, a group threshold is applied onto all repetitions performed by the same subject in a single trial. The group threshold value is determined by finding the mean of threshold values which selected the top 35%voxels with the largest amplitude-correlation product across repetitions of the fourth (weakest) breathing effort. This thresholding strategy is used to compare the activated voxels for different breathing effort. For guided breathing, individual thresholds are applied to each repetition to select the top 35%voxels with the largest amplitude-correlation product. The lung clusters are further divided into four ROIs, namely the anterior left, posterior left, anterior right and posterior right lungs. The anterior-posterior division is defined by the horizontal line passing through the midpoint between the topmost and bottommost voxels in both lung clusters.
The left and right conductivity waveforms (Figure 5B) are obtained by averaging the temporal signals of voxels within both lung clusters. The frequency spectra of the left-right waveforms are obtained by computing their fast Fourier transforms. Regional EIT indicators including the number of activated voxels, total amplitude, and the coefficient of variation (CV) shown in Figure 5B-5E are computed from amplitude image. The number of activated voxels is the count of voxels in the corresponding ROI. The total amplitude is the sum of the voxel amplitude within different ROIs. The CV is the ratio of the standard deviation over mean in the top 75%voxels within each ROI. CV are compensated for potential system-dependent SNR. CV serves to monitor homogeneity of amplitudes as a significant consistent loss in homogeneity could imply decreased lung function. For the forced breathing paradigm, additional regional indicators (MVE, EV1, EV1/MVE ratio, MEF and EF25-75%) are computed for each lung ROI, by averaging the corresponding functional indicator maps within each ROI.
Figures 9A and 9B more specifically illustrate that regional EIT-derived indicators were consistent with the four corresponding breathing paradigms which is the combination of full or mid capacity inhale and fast or slow exhale. Figure 9A shows that individual functional image was used to define the regions of interest (ROIs) , including anterior left, posterior left, anterior right and posterior right. Figure 9B shows that all regional indicators followed a similar trend, i.e., the MVE is higher in paradigms involving full capacity inhale, the EV1 is highest in full capacity inhale with fast exhale and is lowest in half capacity inhale with slow exhale, the EV1/MVE ratio is higher in paradigms involving fast exhale, the MEF is highest in full capacity inhale with fast exhale and is lowest in half capacity inhale with slow exhale, and the EF25-75%is highest in highest in full capacity inhale with fast exhale and is lowest in slow exhale. *p < 0.05, **p < 0.01, ***p < 0.001. Error bars indicate ±standard error of mean. Abbreviations: Maximal volume engaged (MVE) , exhaled volume in 1 second (EV1) , EV1/MVE ratio, and maximum expiratory flow (MEF) .
Data exclusion
In some implementations trials are excluded by inspecting the correlation map and the global conductivity curve. The accepted trials showed two positively correlated clusters in the correlation map (corresponding to two lung regions) and a global conductivity curve which followed the instructed curve.
Regression analysis
For the forced breathing paradigm, a functional mapping to predict spirometry indicators from the global EIT indicators and anthropometrics (including chest circumference, weight, height, weight/height, gender and age) is learnt from a sub-set (training set) of the collected data, then evaluated with the remaining sub-set (test set) . The test-set is obtained by randomly excluding all data from two participants and another 10%of the data from the remaining participants. The training-set is the remaining data from the global set after excluding the test-set. The proportion of the test-set is 24%. This splitting strategy is adopted to verify whether the trained model is capable to generalize unseen data and subjects with different anthropometrics. Inspection of the scatter plots between the EIT-derived indicators and the spirometry indicators for different subjects  showed a high degree of correlation (Figure 11, Table 1) , demonstrating the linear model can be used to predict spirometry indicators. The coefficients of the multiple linear regression model are computed using least squares estimation and then evaluated the statistical significance of different regression coefficients. For MVE, EV1 and EF25-75%, only the coefficients associated with the chest circumference and weight/height are significant, whereas for MEF, only the coefficient associated with the chest circumference is significant. These findings are consistent with the previous study that explores the relationship between the change in lung volume and EIT measurement as they are indicative of the total volume of the conductive medium under inspection as well as the distribution of the body fat. For the EV1/MVE, all the coefficients except the EIT indicator are non-significant. This is explained by the fact that both EV1/MVE and its corresponding EIT indicators are ratios between quantities of the same type. After excluding the non-significant parameters, the coefficients of the multiple linear regression are obtained using the least squares estimation and are shown in Table 2. The scatter plots comparing the predicted indicators and the spirometry indicators are shown in Figure 4C. The model performance is evaluated on both the training set and the test set using two metrics: the correlation (ρ) between the predicted indicator and the target (spirometry) indicator, and the mean absolute error percentage (MAE%) . The evaluation results are presented in Table 2.
More specifically Figures 11 illustrate that spirometry indicators are predicted from EIT over a wide dynamic range for both train and test sets. These results demonstrated that the EIT system has standard spirometry capabilities. The blue dots are the samples from the training set and the red dots are the samples from the testing set.
More specifically for Table 1, the test-set is obtained from the global set by randomly excluding all data from two subjects and another 10%of the data from the remaining subjects. The train-set is the remaining data from the global set after excluding the test-set. The performance of the regression model is evaluated with the correlation coefficient (ρ) and mean absolute error percentage (MAE%) . The results show that the performance on the train-set and test-set are very similar which suggest that the model has a good generalization performance.
Table 1: Evaluation of the regression model used to predict the spirometry indicators.
  ρ (train) ρ (test) MAE% (train) MAE% (test)
MVE 0.80 0.83 15 13
EV1 0.78 0.83 26 23
EV1/MVE 0.86 0.91 13 15
MEF 0.71 0.72 49 50
EF25-75% 0.75 0.79 52 41
Table 2: Coefficients associated with the different variables of the multiple linear regression model used to predict the spirometry indicators.
Figure PCTCN2022082511-appb-000001
Figure PCTCN2022082511-appb-000002
Comparison of functional maps and regional indicators
All functional maps are masked with the lung clusters to remove irrelevant values in non-lung voxels. At each breathing effort, the average normalized functional maps across subjects are obtained (Figure 4D) . Voxels less than 0.1 are additionally masked in the maps to elucidate the relative sizes of the lung clusters. The mean functional indicators in each ROI are used to predict the spirometry indicators at each region. A mapping function is obtained by training a multiple linear regression model for each indicator with the input features used in the previous section but using EIT indicators computed from the average waveform within the lung clusters (instead of the extracted global waveform) . The trained models are then used to compute the predicted spirometry indicators at each ROI (Figure 11) . Two-way ANOVA followed by Bonferroni multiple comparisons is applied to compare the predicted regional indicators across breathing efforts and ROIs.
For each guided breathing paradigm, the amplitude maps are masked with the lung clusters. For shallow and deep guided breathing, the amplitude maps are normalized by the maximum voxel amplitude across all repetitions performed. The average amplitude maps at shallow and deep breathing are obtained by computing the mean normalized amplitude maps at each breathing depth (Figure 5B) . For the COVID-19 case study, each amplitude map is normalized by its maximum voxel amplitude, and the average amplitude maps are obtained by averaging across different subject groups (patient and healthy controls) (Figure 5C) . Two-way ANOVA followed by Bonferroni multiple comparisons is applied to compare the number of activated voxels and total amplitude between the left and right lungs across breathing depth or patient/control group (Figure 5B-5C) . For shallow and deep guided breathing, the regional waveforms and frequency spectra are normalized by their maximum value across all trials, and the mean normalized breathing waveforms and frequency spectra are presented with ± SEM (Figure 5B) . For the COVID-19 case study, the mean left and right CV in the patient and healthy control groups are presented (Figure 2C). Two-way ANOVA followed by Bonferroni multiple comparisons is applied to compare the left and right lung CV between subject groups. To show the trends in CV across time, linear regression models are fit for the left and right lung CV against day of data collection for each subject (Figure 5D). For the COVID-19 discharged subject, the CV against time in each ROI are also fit with linear regression models to further localize the reduction in CV across the period of recovery (Figure 5E) .
It will be appreciated that where the methods and systems of the invention are either wholly implemented by computing system or partly implemented by computing systems then any appropriate computing system architecture may be utilized. This will include stand-alone computers, network computers, dedicated or non-dedicated hardware devices. Where the terms “computing system” and “computing device” are used, these terms are intended to include (but not limited to) any appropriate arrangement of computer or information processing hardware capable of implementing the function described.
It will be appreciated that numerous variations and/or modifications may be made to the invention as shown in the specific embodiments to provide other embodiments of the invention. The described embodiments of the invention should therefore be considered in all respects as illustrative, not restrictive.
The invention has provided a computer-implemented method for lung health assessment, comprising: receiving EIT data associated with a lung of a subject; and processing the received EIT data to determine a health condition of the lung of the subject. Optionally, the EIT data is raw EIT data. Optionally, the EIT data is obtained from the subject or the subject’s lung during a guided breathing operation. Optionally, in the guided breathing operation, the subject periodically or regularly inhales and exhales, optionally with a predetermined number of breaths per minute. Optionally, the EIT data is obtained from the subject or the subject’s lung during a forced breathing operation. Optionally, in the forced breathing operation, the subject first takes a predetermined number of shallow breaths, then inhales to maximum or medium lung capacity, then exhales abruptly or slowly, then inhales to maximum or medium lung capacity, then exhales to normal breathing exhalation volume, and finally takes a predetermined number of shallow breaths. Optionally, the EIT data is obtained from the subject or the subject’s lung during a spontaneous breathing operation. Unlike the guided breathing operation and the forced breathing operation in which the subject breaths according to predetermined patterns, the spontaneous breathing operation is an unguided operation in which the subject can breathe unguided. Optionally, the processing comprises: denoising the received EIT data; and reconstructing time-difference EIT images based on the denoised EIT data. Optionally, the denoising comprises: filtering the EIT data; and/or adjusting outlier measurements in the EIT data based on reference voltage data. Optionally, the reconstruction is performed using a Gauss–Newton method, such as a one-step linear Gauss-Newton solver optionally with a regularization matrix based on the Newton's one-step error re-constructor prior. Optionally, the processing further comprises, for each time series of the time-difference EIT images: determining a global conductivity measure or curve that represents a dominant conductivity signal in the time series due to breathing of the subject; determining an amplitude measure or map that represents a local conductivity variation at different voxels of the images, and/or determining a correlation measure or map that represents voxel-wise correlation between the global conductivity-time curve and the respective local conductivity-time curve. Optionally, the global conductivity measure or curve is determined based on feature selection through filtering and signal extraction. Optionally, the processing further comprises analyzing the global conductivity curve, the amplitude map, and/or the correlation map to determine the health condition of the lung of the subject. Optionally, the processing further comprises, for each time series of the time-difference EIT images: determining one or more parametric functional indicator measures or maps, each of which is related to a respective indicator measure. Optionally, the one or more parametric functional indicator measures or maps are determined based on the global conductivity curve and a voxel-wise conductivity curve. Optionally, the indicator measure is spirometry indicator measure; and the one or more parametric functional indicator comprise: maximum volume engaged (MVE) corresponding to forced vital capacity (FVC) measure; exhaled volume in 1 second (EV1) corresponding to forced expiration volume in 1 second (FEV1) measure; exhaled volume in 1 second (EV1) to maximum volume engaged (MVE) ratio corresponding to  forced expiration volume in 1 second (FEV1) to forced vital capacity (FVC) ratio measure; maximum expiration flow (MVE) corresponding to peak expiratory flow (PEF) measure; and/or expiratory flow at 25-75%of maximum volume engaged (EF25-75%) corresponding to forced expiratory flow at 25–75%of forced vital capacity (FEF25–75%) measure. Optionally, the processing further comprises: processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine a spirometry indicator measure of the subject. This spirometry indicator measure is determined for EIT data obtained using the forced breathing operation. Optionally, the processing further comprises: processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine an indicator of the subject indicative of lung condition, the indicator comprises amplitude, activated voxels, and /or coefficient of variation, which are non-spirometry indicator. The non-spirometry indicator is determined for EIT data obtained using the guided breathing operation and spontaneous breathing operation. The invention has also provided a system and a non-transitory computer-readable medium for implementing the above method.
The method of the invention can be implemented using other EIT signal measurement system, which may or may not be portable, and may or may not be the same as the system specifically illustrated in the above embodiments.

Claims (20)

  1. A computer-implemented method for lung health assessment, comprising:
    receiving EIT data associated with a lung of a subject; and
    processing the received EIT data to determine a health condition of the lung of the subject.
  2. The computer-implemented method of claim 1, wherein the EIT data is raw EIT data.
  3. The computer-implemented method of claim 1 or 2, wherein the EIT data is obtained from the subject or the subject’s lung during a guided breathing operation.
  4. The computer-implemented method of claim 3, wherein in the guided breathing operation, the subject periodically or regularly inhales and exhales, optionally with a predetermined number of breaths per minute.
  5. The computer-implemented method of claim 1 or 2, wherein the EIT data is obtained from the subject or the subject’s lung during a forced breathing operation.
  6. The computer-implemented method of claim 5, wherein in the forced breathing operation, the subject first takes a predetermined number of shallow breaths, then inhales to maximum or medium lung capacity, then exhales abruptly or slowly, then inhales to maximum or medium lung capacity, then exhales to normal breathing exhalation volume, and finally takes a predetermined number of shallow breaths.
  7. The computer-implemented method of claim 1 or 2, wherein the EIT data is obtained from the subject or the subject’s lung during a spontaneous breathing operation.
  8. The computer-implemented method of any one of claims 1 to 7, wherein the processing comprises:
    denoising the received EIT data; and
    reconstructing time-difference EIT images based on the denoised EIT data.
  9. The computer-implemented method of claim 8, wherein the denoising comprises:
    filtering the EIT data; and/or
    adjusting outlier measurements in the EIT data based on reference voltage data.
  10. The computer-implemented method of claim 8 or 9, wherein the reconstruction is performed using a Gauss–Newton method, such as a one-step linear Gauss-Newton solver optionally with a regularization matrix based on the Newton's one-step error re-constructor prior.
  11. The computer-implemented method of any one of claims 8 to 10, wherein the processing further comprises, for each time series of the time-difference EIT images:
    determining a global conductivity measure or curve that represents a dominant conductivity signal in the time series due to breathing of the subject;
    determining an amplitude measure or map that represents a local conductivity variation at different voxels of the images, and/or
    determining a correlation measure or map that represents voxel-wise correlation between the global conductivity-time curve and the respective local conductivity-time curve.
  12. The computer-implemented method of claim 11, wherein the global conductivity measure or curve is determined based on feature selection through filtering and signal extraction.
  13. The computer-implemented method of claim 11 or 12, wherein the processing further comprises analyzing the global conductivity curve, the amplitude map, and/or the correlation map to determine the health condition of the lung of the subject.
  14. The computer-implemented method of any one of claims 11-13, wherein the processing further comprises, for each time series of the time-difference EIT images:
    determining one or more parametric functional indicator measures or maps, each of which is related to a respective indicator measure.
  15. The computer-implemented method of claim 14, wherein the one or more parametric functional indicator measures or maps are determined based on the global conductivity curve and a voxel-wise conductivity curve.
  16. The computer-implemented method of claim 14 or 15, wherein the indicator measure is spirometry indicator measure; and wherein the one or more parametric functional indicator comprise:
    maximum volume engaged (MVE) corresponding to forced vital capacity (FVC) measure;
    exhaled volume in 1 second (EV1) corresponding to forced expiration volume in 1 second (FEV1) measure;
    exhaled volume in 1 second (EV1) to maximum volume engaged (MVE) ratio corresponding to forced expiration volume in 1 second (FEV1) to forced vital capacity (FVC) ratio measure;
    maximum expiration flow (MVE) corresponding to peak expiratory flow (PEF) measure; and/or
    expiratory flow at 25-75%of maximum volume engaged (EF25-75%) corresponding to forced expiratory flow at 25–75%of forced vital capacity (FEF25–75%) measure.
  17. The computer-implemented method of claim 14, wherein the processing further comprises:
    processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine a spirometry indicator measure of the subject.
  18. The computer-implemented method of claim 14, wherein the processing further comprises:
    processing the one or more parametric functional indicator measures or maps with a model, e.g., a regression model, to determine an indicator of the subject indicative of lung condition, the indicator comprises amplitude, activated voxels, and /or coefficient of variation.
  19. A system for lung health assessment, comprising one or more processors arranged to:
    receive EIT data associated with a lung of a subject; and
    process the received EIT data to determine a health condition of the lung of the subject.
  20. A non-transitory computer-readable medium comprising instructions which, when executed by one or more processors, causes the one or more processors to perform the method of any of claims 1 to 17.
PCT/CN2022/082511 2021-03-23 2022-03-23 Electrical impedance tomography based lung assessment WO2022199619A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
HK32021027897.6 2021-03-23
HK32021027897 2021-03-23

Publications (1)

Publication Number Publication Date
WO2022199619A1 true WO2022199619A1 (en) 2022-09-29

Family

ID=83398120

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/082511 WO2022199619A1 (en) 2021-03-23 2022-03-23 Electrical impedance tomography based lung assessment

Country Status (1)

Country Link
WO (1) WO2022199619A1 (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100228143A1 (en) * 2009-03-09 2010-09-09 Drager Medical Ag & Co. Kg Apparatus and method to determine functional lung characteristics
US20140221865A1 (en) * 2013-02-05 2014-08-07 Dräger Medical GmbH Electric impedance tomography device and method
US20160235333A1 (en) * 2013-09-27 2016-08-18 Drägerwerk AG & Co. KGaA Electrical impedance tomography device and electrical impedance tomography method
CN109381187A (en) * 2017-08-02 2019-02-26 德尔格制造股份两合公司 For determining the device and method of difference characteristic number based on EIT- data
US20190125277A1 (en) * 2016-04-25 2019-05-02 Oliver C. Radke User interface of a medical diagnosis system, and computer program therefor

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100228143A1 (en) * 2009-03-09 2010-09-09 Drager Medical Ag & Co. Kg Apparatus and method to determine functional lung characteristics
US20140221865A1 (en) * 2013-02-05 2014-08-07 Dräger Medical GmbH Electric impedance tomography device and method
US20160235333A1 (en) * 2013-09-27 2016-08-18 Drägerwerk AG & Co. KGaA Electrical impedance tomography device and electrical impedance tomography method
US20190125277A1 (en) * 2016-04-25 2019-05-02 Oliver C. Radke User interface of a medical diagnosis system, and computer program therefor
CN109381187A (en) * 2017-08-02 2019-02-26 德尔格制造股份两合公司 For determining the device and method of difference characteristic number based on EIT- data

Similar Documents

Publication Publication Date Title
JP6093421B2 (en) Electrical impedance tomography imaging instrument and electrical impedance tomography imaging method
US10945628B2 (en) Apparatus and method for processing electromyography signals related to respiratory activity
Gomez-Laberge et al. A unified approach for EIT imaging of regional overdistension and atelectasis in acute lung injury
Estrada et al. Improvement in neural respiratory drive estimation from diaphragm electromyographic signals using fixed sample entropy
US20120157866A1 (en) System and method for determining physiological parameters based on electrical impedance measurements
Van Leuteren et al. Processing transcutaneous electromyography measurements of respiratory muscles, a review of analysis techniques
US20160135715A1 (en) Method for respiratory measurement
US11229403B2 (en) System and method for assessing physiological signal quality
Tabata et al. A novel method for detecting airway narrowing using breath sound spectrum analysis in children
Estrada et al. Onset and offset estimation of the neural inspiratory time in surface diaphragm electromyography: a pilot study in healthy subjects
Lee et al. Respiratory rate extraction from pulse oximeter and electrocardiographic recordings
Gazi et al. Robust estimation of respiratory variability uncovers correlates of limbic brain activity and transcutaneous cervical vagus nerve stimulation in the context of traumatic stress
US20240008759A1 (en) Method and system for monitoring cardiopulmonary function using electrical impedance tomography
Milagro et al. Electrocardiogram-derived tidal volume during treadmill stress test
Strodthoff et al. Inferring respiratory and circulatory parameters from electrical impedance tomography with deep recurrent models
Maarsingh et al. Airflow limitation in asthmatic children assessed with a non-invasive EMG technique
Sello et al. Respiratory sound analysis in healthy and pathological subjects: A wavelet approach
WO2022199619A1 (en) Electrical impedance tomography based lung assessment
Młyńczak et al. Flow parameters derived from impedance pneumography after nonlinear calibration based on neural networks
Grossman et al. Reliability of respiratory tidal volume estimation by means of ambulatory inductive plethysmography
Liu et al. Non-invasive respiration and ventilation prediction using a single abdominal sensor belt
Seppä et al. Assessment of breathing parameters during running with a wearable bioimpedance device
Ge et al. Research on denoising algorithm of thoracic impedance signal for respiratory monitoring during running exercise
Seppä Development and clinical application of impedance pneumography technique
Zouari et al. Standalone electrical impedance tomography predicts spirometry indicators and enables regional lung assessment

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22774275

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22774275

Country of ref document: EP

Kind code of ref document: A1