Proceedings of the 20th Annual International Conference of the ZEEE Engineering in Medicine and Biology Society, Vol. 20, No 2,1998
DESIGN, IMPLEMENTATION AND BENCH EVALUATION OF A SYSTEM FOR AUTOMATIC SYNCHRONIZATION OF CHEST X-RAY RADIOGRAPHY WITH PEAK LUNG INFLATION Samsun LamPotang”’ **’***, Colin Cheung-Seekit****, Paul Langevin”’ *****
* Department of Anesthesiology, * * University of Florida Brain Institute, University of Florida College of Medicine, Box 100254, Gainesville, FL 32610-0254
* * * Department of Electrical and Computer Engineering, University of Florida * * * * Systems Engineering Design Department, University of Waterloo, Canada * * * * * Veterans Affairs Medical Center, Gainesville, Florida E-mail:
[email protected] ABSTRACT The clinical diagnostic data in a chest radiograph is enhanced if chest motion is minimal and X-ray beam exposure occurs at or near peak lung inflation (PLI). Currently, for patients who cannot voluntarily hold a maximal inspiration, beam exposure is manually “synchronized” with PLI and is a hit or miss proposition. We implemented a system for automatically synchronizing beam exposure with PLI during chest radiography. Pressure and bi-directional flow at the airway of patients undergoing positive pressure ventilation are monitored by a personal computer. An algorithm coded in C looks for a zero flow crossing and a peak lung pressure to determine PLI. Custom-built interface electronics allow (a) the system to detect, in real time, which buttons on the X-ray machine handswitch are being pressed by the operator and (b) X-ray beam exposure to be triggered via the software. During bench validation using a mechanical test lung, the system worked consistently, without false triggering. Tests with human patients are currently under way, with IRB approval and informed consent. Preliminary human data indicate that higher quality and more consistent chest radiographs are possible with automatic synchronization of X-ray beam exposure with PLI. Keywords: X-ray exposure synchronization, peak lung inflation, image quality
synchronization of the X-ray beam exposure with peak inflation of the lungs. At PLI, there is more gas in the lungs, resulting in a lower mean tissue density because the alveoli in the lungs are at maximum inflation. A lower mean tissue density allows more X-ray radiation to reach the X-ray film resulting in a darker background for regions with healthy lung alveoli. Against this darker background, lesions and tumors as well as atelectatic (collapsed) alveoli are more readily differentiated as paler regions, enhancing the radiograph contrast and quality. Additionally, at PLI, the chest is fleetingly motionless because the rate of change of volume (dv/dt) is instantaneously zero as dv/dt changes from positive (during inspiration) to negative (during expiration). Motion artifact is especially significant at longer beam exposure times. Current methods for synchronization of X-ray beam exposure with PLI include placing a paper cup on the chest of a supine patient to obtain a sharp, well-defined edge and visually anticipating PLI, for patients unable to voluntarily hold a maximal inspiration. Manual synchronization by visually anticipating PLI requires operator skill and experience and may result in highly variable chest radiograph quality, frequent need for repeat chest radiographs, increased X-ray radiation exposure to the patient, wasted material and personnel costs, delays in initiating treatment and reduced confidence in the diagnostic information conveyed by the radiograph. Our objective was to design, implement and evaluate a synchronization system that improves over the existing art and automatically triggers the X-ray beam exposure near PLI.
INTRODUCTION Chest radiographs must be of good quality such that accurate diagnostic information can be confidently interpreted and treatment initiated. Parameters influencing chest radiograph quality include: patient movement and position, consciousness and ability to comprehend and respond to instructions like holding a maximal inspiration, Xray beam penetration, beam exposure time and 0-7803-5164-9/98/$10.00 0 1998 IEEE
METHODS The synchronization system is interposed between the manufacturer-provided X-ray firing handswitch and the Xray machine (General Electric AMX-4+) and is transparent to the user except for the need to add a flow and pressure probe (FloTrak, Novametrix, Wallingford, CT) at the patient’s airway (see Figure 1). 711
The regular firing handswitch provided by the manufacturer normally attaches to the X-ray machine via a 10 ft telephone cord, allowing the operator to trigger X-ray beam exposure while standing away from the machine, usually behind a shielded enclosure. The 3 switches on the handle are for (a) a collimator light that projects a cross-hair on the site to be imaged and is used for aiming the X-ray beam, (b) a “rotor up’’ function which initiates the process of accelerating the rotor up to speed and (c) the “fire” button which actually fires the X-ray machine. The “rotor-up” and “fire” buttons are arranged such that one cannot press the “fire” button without first pressing the “rotor-up” button. Both pressure and flow are saiiipled at 60 Hz, from the processing module of the FloTrak, via a I O foot long RS-232 (19,200 baud, 8 data bits, no parity, 1 stop bit) serial cable which allows the PC (Pentium, 100 MHz) to be moved around, e.g., to a location shielded from X-ray radiation. Because there was some noise in both the pressure and flow signals which would have interfered with the PLI detection algorithm, we filtered the flow and pressure values by averaging each signal over the last 3 samples, Le., Qfilt(t)= [Q(t) + Q(t-I) + Q(t-2)]/3 where Qfilt(t) is the filtered flow signal at time t and Q is the actual flow. Gas flow towards, and away from, the patient are defined as positive and negative respectively.
I
power S“uPp*,
li FIGURE 1. The schematic of the system layout The software is a DOS-based application with a graphical user interface (GUI), written in C using the Borland C v 3.1 compiler (Scotts Valley, CA). The PLl detection algorithm looks for a falling zero flow crossing (instantaneous change in gas flow direction from positive to negative) AND a peak pressure value, to positively identify peak lung inflation, during positive pressure ventilation. Each method (falling zero-flow crossing and peak pressure), can individually by itself, detect PLI. Both conditions need to be satisfied in our algorithm, for added robustness in detecting PLI. In addition
712
to low pass filtering, a presumed PLI is only valid if the pressure at the presumed PLI is more than 2 cm H,O above the baseline pressure. The baseline pressure is updated every time a rising, zero flow crossing occurs, i.e., end of exhalation and start of inspiration, automatically compensating for the positive-end expiratory pressure (PEEP) level, if used. The PLI detection algorithm was also designed to work with a sigh breath which is sometimes used to hyperinflate the lung, for only one single breath, while the radiograph is taken. The graphical user interface (GUI) displays, in real time, separate time plots of flow and pressure to allow the operator to follow on the GUI, the operation of the algorithm. The actual firing of the X-ray machine is indicated by a vertical red line on both the flow and pressure time plots. The GUI also emulates the functionality of the firing handswitch, allowing the operator to “depress” the collimator, rotor-up and fire buttons, using a mouse. During initial trials, the GUl emulation of the firing handswitch was not well received by X-ray machine operators. Thus, the standard firing handswitch was retained as a preferred and familiar user interface so that the operators would not have to learn a new user interface. The three switches on the standard firing handswitch were interfaced such that each toggled a bit on the parallel port of the PC, enabling the software to determine, in real time, which button the operator was pressing. Three single-pole, double-throw, electromechanical relays (TKI -5V, Aromat Corporation) corresponding to the collimator, rotor-up and fire functions were controlled by the software, using another 3 bits on the parallel port. Opto-isolation is provided for the electromechanical relays so that the X-ray machine and the synchronization system are not in direct electrical contact. The circuitry for the interface electronics is shown in Figure 2. During standard use, beam exposure will not occur if the “fire” button is depressed within 2.5 seconds after the “rotorup” button has been pressed, i.e., the rotor has not had enough time to accelerate to speed [I]. Beam exposure will occur once the rotor is up to speed but by that time the lungs might no longer be at peak inflation. This required delay between pressing the rotor-up and the fire buttons renders manual synchronization even harder. It may also explain the “variability” in the delay between pressing the fire button and the machine actually firing reported by some operators. To ensure that the rotor is fully up to speed before X-ray beam exposure occurs, our software imposes a 3 s minimum delay between closing of the “rotor-up” and the “fire” electromechanical relays. In other words, if a peak lung inflation happens to occur within 3 seconds of the “rotor up” electromechanical relay being closed, the “fire” electromechanical relay will not close, but instead will fire on the next peak lung inflation after the 3 seconds delay has elapsed. The automated synchronization system allows the operator to “fire and forget” by pressing the regular firing
button whenever he or she is ready instead of having to visually anticipate peak lung inflation. A buzzer will sound and a green LED will light after a successful X-ray beam exposure to inform the operator that it is safe to leave the shielded enclosure or remove the lead apron.
the tidal volume (VT) exhaled, dividing the plate separation by 1.65 gives %VT exhaled which is listed in Table 1
RESULTS The synchronization system worked 100% of the time during the bench validation, without any false triggering. All 8 radiographs triggered by the synchronization system clearly show the metal top plate of the test lung almost touching the marker. The quality of the radiographs was also clear with sharp, well-defined edges. Figure 3 is a representative radiograph from the bench validation experiments.
Input hom Finlg Handle
Tidal volume exhaled (“A)
Tidal volume exhaled (“A)
Mid-exhalation 7 7 Full deflation 7 7 Mid-inhalation 7 7 Mean 7.25 7.25 II 0.5 I 0.5 Std. deviation .. TABLE 1. The percentage of the tidal volume exhaled at the time of X-ray beam exposure
I
FIGURE 2. T h e interface electronics
I
~
Bench Evaluation Protocol The bench evaluation was performed with a test lung (compliance set at 0.1 L/cm H,O, flow resistance 20 cm H,O/L/s, Michigan Instruments TTL, Grand Rapids, MI) under positive pressure ventilation (tidal volume 700 mL, respiratory rate 10 bpm, inspirat0ry:expiratory time ratio 1 :2) and a portable X-ray machine (AMX-4+, General Electric). At peak lung inflation, the top plate of the test lung momentarily touched a radio-opaque metal marker. The cross hairs of the collimator were aimed at the point where the top plate and marker touch to avoid parallax error. Eight exposures were taken with X-ray machine settings of 60 kVp and 0.8 mAs. One set of 4 exposures was taken by pressing the rotor-up and firing buttons simultaneously at peak inflation, mid-exhalation, maximal deflation and midinspiration. Another set of 4 radiographs was taken at roughly the same points in the breathing cycle; however after pressing the rotor-up button, the fire button was depressed only after the “X-ray ready” message appeared on the display of the X-ray machine. The separation between the marker and the top plate of the test lung was recorded from the radiograph (Figure 3). Two notches one inch apart were cut on the marker to provide a reference distance. The distance of the top plate from the marker when all the tidal volume has been exhaled was 1.65 inches. Assuming that the distance of the marker from the top plate is proportional to
713
DISCUSSION
The original design specification called for the radiograph to be taken with at least 90% of the tidal volume still in the lungs. Our experimental results indicate that on average, only 7.25% of VT has been exhaled at the time of beam exposure, meeting the original design specifications. The results were also reproducible between exposures with a low standard deviation of only 0.5% VT. The current software supports only detection of peak lung inflation during positive pressure ventilation because the majority of patients who might benefit most from the synchronization system are under positive pressure ventilation. Modification of the software to detect peak lung inflation during spontaneous ventilation, and temporary application of a face mask, will allow the system to be used with non-intubated, spontaneously breathing patients. The system should also work with other airway devices like a Combitube and a laryngeal mask airway. Our approach used both pressure and flow to determine peak lung inflation. Other devices like belts with stretch sensors [2], temperature sensors [2], humidity sensors [3], acoustic sensors [4] and in-line capnographs might provide alternative means of respiratory gating. The reliability of an
I
algorithm that uses only pressure or only flow to determine peak lung inflation also deserves to be investigated. Initially, we investigated the possibility of determining peak lung inflation by reading the serial port of those electronic ventilators whose serial output indicates in real time the end of inspiration. This initial approach had the disadvantages that (a) it would not have worked with ventilators which do not provide serial data that indicate in real time the end of inspiration, like ail-pneumatic ventilators and (b) different serial communication protocols would have been required for different brands, models and versions of ventilators, requiring correct identification of the brand, model and software version of each ventilator, information which may not always be available.
flowmeter into the breathing circuit, especially if the ventilator’s anti-disconnect, low pressure alarm starts sounding. An alternative, permanent location for the flowmeter would be the connection between the ventilator and the inspiratory limb of the breathing circuit. This site would require a change in the peak lung inflation detection algorithm as the monitored flow would no longer be bidirectional; it would have the advantage that the flowmeter could be permanently left on the ventilator, assuming no backflow from the patient that could cause crosscontamination. The user would simply connect the serial cable from the flowmeter into the synchronization unit, without a need to break the breathing circuit integrity. Respiratory gating has also been used in magnetic resonance imaging [2]. Our technique might also be applicable to MRI because the flowmeter probe and tubing is made of plastic.
CONCLUSIONS The bench test data suggest that automatic synchronization of X-ray beam exposure with peak lung inflation is feasible and should provide more consistent and higher quality radiographs in a clinical setting. Preliminary results from human trials appear to support our expectation of enhanced imaging of the chest.
ACKNOWLEDGMENTS The authors would like to thank Nikolaus Gravenstein, Craig Bakuzonis, Kevin Tharp, Justin Sanchez and Beverly Hoyle for their assistance and support during this project.
REFERENCES
FIGURE 3. A typical radiograph showing the PLI marker, with two notches one inch apart, almost touching the top plate of the test lung. On the other hand, one could argue that a weakness of the design using the flowmeter is that it requires the operator of the X-ray machine to temporarily break the integrity of the breathing circuit to insert a flowmeter at the Y-piece. Some X-ray operators might not be comfortable inserting a
714
[I] GE Medical Systems: Technical Publications. Direction 2 166913-100, Revision 1. A M X d + Operation (Model 2 169360 Series). General Electric Co., 1997 [2] R.L. Ehman, M.T. McNamara, M. Pallack, H. Hricak and C.B. Higgins, “Magnetic resonance imaging with respiratory gating: Techniques and advantages,” AJR 143:1175-1182, 1984. [3] T. Tatara and K. Tsuzaki, L‘An apnea monitor using a rapid-response hygrometer,” J Clin Monit 13:5-9, 1997. [4] J. Werthammer, J. Krasner, J. DiBenedetto and A.R. Stark, “Apnea monitoring by acoustic detection of airflow,” Pediatrics, V O ~ .71, 1983.
