Accuracy of spirometric and flow-volume indices obtained by digitizing volume-time tracings

CR O'Donnell, SL Sneddon, M Schenker… - Am Rev Respir …, 1987 - atsjournals.org
CR O'Donnell, SL Sneddon, M Schenker, E Garshick, FE Speizer, J Mead
Am Rev Respir Dis, 1987atsjournals.org
Methods The data acquisition system consisted of a microcomputer (DEC LSI-B; Digital
Electronics Corp., Maynard, MA) with60k bytes of read/write memory, serial interface (DEC
DLV-IlJ), dual double density floppy-disk drives (Charles River data; Portage, MI) and a I5-
inch Bit-Pad-One digitizer (Summagraphics Corp.). A Tektronix Model 4013 terminal
(Tektronics Corp., Beaverton, OR) served as console and graphics display unit for plots of
volume versus time and flow versus volume. Tracings were placed on the digitizer …
Methods The data acquisition system consisted of a microcomputer (DEC LSI-B; Digital Electronics Corp., Maynard, MA) with60k bytes of read/write memory, serial interface (DEC DLV-IlJ), dual double density floppy-disk drives (Charles River data; Portage, MI) and a I5-inch Bit-Pad-One digitizer (Summagraphics Corp.). A Tektronix Model 4013 terminal (Tektronics Corp., Beaverton, OR) served as console and graphics display unit for plots of volume versus time and flow versus volume. Tracings were placed on the digitizer, andvolume-time curves were handtraced using a magnifying cursor. Individual curves were backextrapolated using thetechnique described by Smith and Gaensler (4), thus ensuring comparability to hand-read values of FEV. and FEV7 s-Beginning and end of the curves were signaled by pressing one of the buttons on the cursor assembly. Thedigitizer was connected to the serial interface of the LSI-B, and, at 19,200 baud, was capable oftransmitting70points/s with an underlying resolution of 0.005 inches (corresponding roughly to 5 ml or 5 ms on the spirometer paper).
Thecomputer routine for data acquisition accepted pointsfrom the digitizer tabletand used them to build a volume-time array in memory. Allof the volume values in the data stream corresponding to a discrete value of timewere averaged together to minimize the random error inherent in following a curve byhand. Thevolume-time array could beplotted for visual checks on accuracy. In addition, the program was designed to accept only curves that varied by less than 50ml in the lastsecond, inaccordance with accepted standards for the reliability of spirometric tracings (3). Flow was calculated byleast-squares regression of the volume-time signal, using data in an interval of about 0.06 s on either
ATS Journals