Scholarly article on topic 'Automated Estimation and Analysis of Lung Function Test Parameters from Spirometric Data for Respiratory Disease Diagnostics'

Automated Estimation and Analysis of Lung Function Test Parameters from Spirometric Data for Respiratory Disease Diagnostics Academic research paper on "Medical engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia Computer Science
OECD Field of science
Keywords
{spirometry / spirometer / "respiratory rate" / "lung function test" / "data analysis" / "respiratory disease diagnostics" / "dead space volume" / regression.}

Abstract of research paper on Medical engineering, author of scientific article — Aruneema Das, David P Johns, Ritaban Dutta, Haydn E Walters

Abstract A spirometer is used for basic lung function test for preliminary diagnosis of respiratory diseases. There are significant amount of calculations and graphical analysis required to transform the raw spirometric data into meaningful parameters. This parameters and graphs help the physicians in preliminary patient diagnosis for respiratory disorders like asthma, chronic obstructive pulmonary disease, etc. This study was undertaken for the development of a software system which can be used with any spirometric instrument to automate the calculations of pulmonary dead space volumes and analysis of raw data. The clinician can feed the raw data from patient testing into the easy to use graphical user interface of the software which will be analyzed instantly and all the parameters, regression slopes, shape analysis plots and the results will be displayed graphically. The estimation of the vital parameters and regression slopes are based on standard protocols and equations. This system will eliminate presently practiced time consuming manual calculations and graphical analysis; will have increased precision, be considerably faster and more versatile.

Academic research paper on topic "Automated Estimation and Analysis of Lung Function Test Parameters from Spirometric Data for Respiratory Disease Diagnostics"

^jjj^ Procedia Computer Science

CrossMark Volume 29, 2014, Pages 2045-2054 ICCS 2014. 14th International Conference on Computational Science

Automated estimation and analysis of lung function test parameters from spirometric data for respiratory

disease diagnostics

112 1 Aruneema Das , David P Johns , Ritaban Dutta and Haydn E Walters

1 School of Medicine, University of Tasmania, Hobart, Australia 2CCI, CSIRO, Hobart, Australia aruneema.das@utas.edu.au, ritaban.dutta@csiro.au

Abstract

A spirometer is used for basic lung function test for preliminary diagnosis of respiratory diseases. There are significant amount of calculations and graphical analysis required to transform the raw spirometric data into meaningful parameters. This parameters and graphs help the physicians in preliminary patient diagnosis for respiratory disorders like asthma, chronic obstructive pulmonary disease, etc. This study was undertaken for the development of a software system which can be used with any spirometric instrument to automate the calculations of pulmonary dead space volumes and analysis of raw data. The clinician can feed the raw data from patient testing into the easy to use graphical user interface of the software which will be analyzed instantly and all the parameters, regression slopes, shape analysis plots and the results will be displayed graphically. The estimation of the vital parameters and regression slopes are based on standard protocols and equations. This system will eliminate presently practiced time consuming manual calculations and graphical analysis; will have increased precision, be considerably faster and more versatile.

Keywords: spirometry; spirometer; respiratory rate; lung function test; data analysis; respiratory disease diagnostics; dead space volume; regression.

1 Introduction

Spirometry is a basic Pulmonary Function Test (PFT) measured using a spirometer (Pierce & Johns, 2003). A spirometer is a precision differential pressure transducer based apparatus for measuring the volume of air inspired and expired by the lungs and the respiration flow rates. Many other key lung function parameters like vital capacity, tidal volume, peak expiratory flow (PEF), etc. can be calculated from these values. The most common ones are the forced vital capacity (FVC) and the forced expired volume in one second (FEV1) (Pierce & Johns, 2003). It is useful as a preliminary test of the health condition for patient's lung (Johns, Burton, Walters, & Wood-Baker, 2006). The

Selection and peer-review under responsibility of the Scientific Programme Committee of ICCS 2014 2045 (gi The Authors. Published by Elsevier B.V.

shapes and measurement of other vital parameters like respiratory rate (RR) and pulmonary dead space (VD) from the flow and volume curves (Schachter, Kapp, Maunder, Beck, & Witek, 1986) provides indices for the primary diagnosis of respiratory disease caused by obstruction in lungs like asthma and chronic obstructive pulmonary disease (COPD) (Wildhaber, et al., 2007). Other measures like the ratio of FVC and forced expiratory volume (FEV) (Nathel, Nathell, Malmberg, & Larsson, 2007), ratio of the forced expiratory flow (FEF) to the ideal value and regression slopes of dead space volumes at various flow volumes (Thamrin, Latzin, Sauteur, Riedel, Hall, & Frey, 2007) are also important factors to assess airway distensibility or dilation capacity of airways (Johns, Wilson, Harding, & Walters, 2000).

Figure 1(a) shows Spiroson (Easy-on-PC model), a hand held spirometer (developed by ndd Medizintechnik AG) used in this study (www.ndd.ch). It measures air flow using an ultrasonic transducer. Figure 1(b) shows the cross-section of the Spiroson (Easy-on-PC model). The sensor consists of a flow tube and an oblique channel in which two ultrasonic transducers are mounted on opposite sides as shown in Figure 1(b). At short intervals ultrasonic pulses are emitted and received by the transducers in alternating directions.

Transduce« l

Transducer 2

Temperature

Figure 1: (a) Spiroson Easy-on-PC model developed by ndd Medizintechnik AG. (b) Cross-section of the

Spiroson Easy-on-PC model (www.ndd.ch)

The observed up and down-stream transit times are determined with a resolution of at least 10 ns. Mean flow velocity of the gas flow through the breathing tube can be derived from the transit times. At a given tube geometry, flow and volume can be calculated and are independent of gas composition, humidity and temperature. Unlike many other types of spirometer, the Spiroson (Easy-on-PC) has no moving parts; therefore accuracy is independent of any mechanical function and measurement of variables such as pressure or displacement volume. Providing the tube geometry (cross-sectional area of the gas stream) is fixed, the only variable requiring accurate measurement is the transit-time of the ultrasonic pulses between the two transmitters to receivers making this a stable system. In addition to flow, molecular mass of the test gas can be derived in the same measuring cycle. Offered as an option to flow/volume measurement the molecular gas analysis requires no additional hardware but an upgrade to the signal processor software. Based on a purely digital working principle Spiroson needs a once in a life-time calibration only (www.ndd.ch).

The Spiroson can be connected to the PC via USB as shown in Figure 2 and its own data logger called WBreath acts as an interface for all the raw data collection. It records raw data (e.g. flow, volume, etc.) when the subject breath in and out through the white mouthpiece. The data files provided by WBreath after completion of a testing cycle consists of a text file with the raw data of volume, flow and molar mass (MM) against sampling time and a Microsoft Excel file with the other estimated key parameters.

There are some problems with WBreath:

a. Display of the raw data is uncalibrated so there is no information about any values of time, flow, volume and MM.

b. The manufacturer produces no information about the method/formula used for estimating parameters like dead space volume hence there is no means for cross validation.

c. It has no scope of any other analysis or plot generation specific to research needs like regression slopes of VD against flow volumes etc.

Figure 2: Spiroson Easy-on-PC model connected to the PC and the raw data displayed in the data logger.

VD Estimation Regression

.txt File >

rp_breath1.txt

0.067075 0 05786 0.073808 0.10586 0.071217 0.089895 0.083519 0.069756 0.088856 0.11335 0 08381 0 10299 0.099162 011102 0.089742

0 066719 0.061406 0 073464 0 099842 0.056429 0.061513 0.074527 0.056726 0.070857 0.10833 0.076 0.D81043 0 077807 0.083562 0.070302

Flow and Volume

Figure 3: Spiroson Easy-on-PC model connected to the PC and the raw data displayed in the data logger.

This study aims to develop a user friendly software system providing a solution to the above problems. The main purpose will be to semi-automate the calculations of airway distensibility and flow-dependence of VD. The next section describes the details of the software development and the final section contains discussion and further work.

2 Software Development

The software for calibrated data display of flow, volume & molar mass and VD estimation and analysis is developed in MATLAB (www.mathworks.com) for pc. Figure 3 shows the software with the plot of raw flow, volume and molar mass data from Spiroson after a 30 second patient breath testing. It also displays the estimated VD in two different methods namely VD 10% (Thamrin, Latzin, Sauteur, Riedel, Hall, & Frey, 2007).

VD Estimation Regression

Load (.txt) for VD

Save VD 1 -

VD Results (L) VD10 VDP 0.6 0.4 0.2 -

Time (s

Figure 4: The VD estimation menu.

Flow and Volume

5.5 Time (sec) 6 6.5 7 7.5 8

Figure 5: Dead space volume estimation from the flow, volume and MM curves.

The raw data file from Spiroson can be selected by the user from the 'VD estimation' menu as shown in Figure 4 and the software then displays the data and also displays the VD values after estimation. The system also pre-processes the data by filtering the noise out. The VD values can be saved using the 'Save VD' option as shown in . The name of the selected data file is displayed as well.

2.1 Dead space volume estimation

Dead space volume is a key parameter to determine any obstruction in the small airways of the patient's lungs (Bouhuys, 1964). There is a defined normal range for male and female subjects. VD has been estimated by the software from the flow and volume data as described below and shown in Figure 5.

Expiration starts at red dot on line 'a' when the volume starts falling in downward direction and flow goes below the 0 axis Expiration ends at line 'b' when the volume starts rising in upward direction and flow goes above 0 axis.The 10% rise point (from the start of expiration to the next peak) is the red dot for each breath on the MM plot as shown in in Figure 3and marked by line 'c' in Figure 5. The minimum value (after start of expiration) is the black dot for each breath on the MM plot as shown in Figure 3 and marked by line 'd' in Figure 5.

VD 10% is the area under the Flow curve between line 'a' to line 'c' and the red line passing through the minimum value of the flow curve for that expiration cycle. The Alveolar dead space (VDP) is the area under the Flow curve in Figure 5 between line 'a' to line 'd' and the red line passing through the minimum value of the flow curve for that expiration cycle. An inspiration and expiration together forms a complete breath cycle. A value of VD and VDP is estimated for each breath and displayed in the table as shown in Figure 3.

Figure 6 : The "Regression" menu.

2.2 Regression plots

The user can select the regression option from the 'Regression' menu in the software as shown in Figure 6. The 'select file for regression' option allows the user to choose the data file containing the RR and PEF values corresponding to the flow and volume data used for VD estimations.

- ^^REGRESS_rp_brea

Home Insert Page Layout Formulas Data

Al * £ 1.433

A B C D E

1 1.433 0.061308

2 3 4 1.62 0.066719 1.5 0.061406 1.513 0.073464 1.647 0.099842 1.43 0.056429 1.613 0.061513 1.627 0.074527 1.243 0.056726 1.407 0.070857 1.443 0.10833 1.577 0.076 1.617 0.081043 1.723 0.077807 1.673 0.083562 1.4 0.070302

13 14 15

M < ► >l RR VD10 RR VDP PEF VD10 PEF VDP

Figure 7: The Excel file format for the saved regression data points.

Figure 8: Dead space volume estimation from the flow, volume and MM curves.

The estimated values of VD and VDP are plotted against RR and PEF respectively and a linear regression is performed. Data points from each of the regression plots are saved as separate labeled sheets in an Excel file automatically as shown in Figure 7, in the same folder where the previous data files belong and named as the corresponding raw data text file. The file name is displayed under Excel file as shown in Figure 8. All the plots are displayed in the software as shown in Figure 8. The equation for the linear regression and R2 value is also displayed for each of the plots. If some of the data points are far apart from the regression line in any of the plots, the user can use the 'open & edit regression' option as shown in Figure 6 to open the saved Excel file as shown in Figure 7 and delete the problematic data points. The 'repeat regression' option as shown in Figure 6 can then be used for repeating the regression calculation and re-plotting with modified R2 values and regression equations. The total process can be repeated multiple times until the user is satisfied with the regression values in reaching a meaningful conclusion about the patient's lung condition.

Figure 9: Flow volume curve shapes.

3 Further Development

Other measures of lung obstruction estimation are being developed based on the shape of the flow volume curve for each breath sample. They will be added to the software along with the traditional measures and put to patient testing. Figure 9 shows the shape of ideal flow volume curve and also for a person with obstructive lungs. The forced expiratory flow at 75% of flow volume (FEF 75%) and forced expiratory flow at 50% of flow volume (FEF50%) are reference points for deviation of the curve from the ideal curve. In the global concave pattern curve, both FEF75% and FEF50% are reduced than the ideal values. In the peripheral concave pattern curve, FEF50% is normal but FEF75% is reduced from the ideal values. Another measure of lung obstruction is being developed based on the ratios of deviation from the ideal point.

The angle B (Schachter, Kapp, Maunder, Beck, & Witek, 1986) is a measure of the degree of concavity which signifies the obstruction. We have developed another angle a as shown in Figure 10 which in addition to angle B will produce a more accurate measure of the degree of concavity and thus the obstruction. Here X is volume expired to PEF.

Angle a is estimated from Figure 10 as follows:

Tan FBE = (PEF - FEF50%)/ (0.5 x FVC)

Tan ABE = (PEF - FEF50%)/ ((0.5 x FVC) - X)

ABF = ABE - FBE

= tan-1[(PEF - FEF50%)/ ((0.5 x FVC) - X)] - tan-1 [(PEF - FEF50%)/ (0.5 x FVC)]

Angle a = Angle CBD ~ Angle ABF

= tan -1 [(PEF - FEF50%)/ ((0.5 x FVC) - X)] - tan -1 [(PEF - FEF50%)/ (0.5 x FVC)]

VOLUME

/2 FVC

Figure 10: Angle a estimation.

The disadvantages of using the traditional FEV1 (forced expiratory volume at 1 second) to FVC ratio as a measure of concavity is its dependence on FVC. If a patient fails to blow out fully during the test then will be overestimated and airflow obstruction underestimated. Another new index is under development to overcome this disadvantage. All the additional measures will be added to the software for a comparative study of the effectiveness and validity of all of them.

There are various indices in spirometry to measure airway obstruction in lungs and the superiority of one index over the other is still debated. The estimation of VD and the linear regression plots obtained from the developed software will be used for patient trial to measure airway distensibility and flow-dependence of VD. People with asthma or COPD have limited airway distensibility compared to normal subjects. As the estimations and plots are automated it will generate results much faster after completion of patient sampling. The software has been compiled to an application package, hence it is free to use without requirement of any paid license. As the software interface is easy to operate, anyone can start using it with very little training. The software is also quite generic and can be made compatible for any spirometer with some basic modifications. Other new indices based on the shape of flow and volume curve is still under development as described in Section 3. These additional indices and other medical tests will run in parallel for cross validation of the results.

The authors would like to thank Paula Fottrell for her efforts in conducting the trial data collection experiments which helped in the software development process.

4 Discussion

5 Acknowledgement

References

Bouhuys, A. (1964). Respiratory dead space. In A. Bouhuys, W. O. Fenn, & H. Rahn (Eds.), Handbook of Physiology (Vol. 1, p. Section 3: Respiration). Washington: American Physiological Society.

Johns, D. P., Burton, D., Walters, J. A., & Wood-Baker, R. (2006). National survey of spirometer ownership and usage in general practice in Australia. Respirology, 11(3), 292-298.

Johns, D. P., Wilson, J., Harding, R., & Walters, E. H. (2000). Airway distensibility in healthy and asthmatic subjects: effect of lung volume history. Journal of Applied Physiology, SS(4), 1413-1420.

Nathel, L., Nathell, M., Malmberg, P., & Larsson, K. (2007). COPD diagnosis related to different guidelines and spirometry techniques. Respiratory research, S(1), 89-95.

Pierce, R., & Johns, D. P. (2003). Pocket guide to Spirometry. McGraw-Hill Australia.

Schachter, E. N., Kapp, M. C., Maunder, L. R., Beck, G., & Witek, T. J. (1986). Smoking and cotton dust effects in cotton textile workers: an analysis of the shape of the maximum expiratory flow volume curve. Environmental health perspectives, 66, 145-148.

Thamrin, C., Latzin, P., Sauteur, L., Riedel, T., Hall, G. L., & Frey, U. (2007). Deadspace estimation from CO2 versus molar mass measurements in infants. Pediatric pulmonology, 42(10), 920928.

Wildhaber, J. H., Sznitman, J., Harpes, P., Straub, D., Moller, A., Basek, P., et al. (2007). Correlation of spirometry and symptom scores in childhood asthma and the usefulness of curvature assessment in expiratory flow-volume curves. Respiratory Care, 52(12), 1744-1752.