CPET - Automated Input Measurement

15O Arterial Time Activity Curve (TAC)

Questions about this project should be addressed to: Brian Murphy - brian@petnet.buffalo.edu.

Purpose

Develop an automated system for measuring arterial Time-Activity curves (TAC) during 15O-water studies. These curves are required in order to properly compute quantitative Cerebral Blood Flow using the autoradiographic technique (and most other techniques as well).

Prior to the development of this system, TACs were obtained manually by withdrawing arterial blood into weighed test tubes. These tubes were subsequently counted in an automated counting device to determine the activity in each sample. This approach has several significant drawbacks:

• sampling rate is low; maximum rate is one sample every 3-5 [sec]. This results in a large amount of dispersion (and poor sampling resolution) in the resulting TAC which in turn introduces errors into the calculated CBF values.
• sample withdrawal rate is variable (depends on varying arterial pressure).
• large amount of manual sample processing resulting in
  • significant complication of the study during the acquisition of the samples (requiring 1-2 people focused on that task alone)
  • significant amount of 15O decay between sample acquisition and sample counting
  • increased personnel radiation exposure

With the new system, blood is automatically withdrawn at a uniform rate (5-20[ml/min]) and almost instantaneously counted in much shorter sampling intervals (0.5[sec]). This eliminates all of the drawbacks present in the manual sampling method.

The beta detector approach was chosen because the high 511[keV] photon background levels from 15O present in the patient make TAC determination by measuring these photons difficult without a large amount of shielding. This shielding, in turn, would necessitate moving the detector assembly further from the patient - thus increasing the length of tubing through which the sample is acquired and adding dispersion to the resulting TAC. Coincidence detection is another approach, but was deemed unnecessary for this project - in large part due to increased detector/electronics complexity and the high cost (5 to 7 times that of current system costs) associated with commercial coincidence systems. With the beta detector, we are able to place the detector assembly only centimeters away from the arterial line catheter.

Several papers have been previously published on the use of a beta detector in this capacity. The following paper is a good place to start: A System for Cerebral Blood Flow Measurement Using an H215O Autoradiographic Method and Positron Emission Tomography I.Kanno, et. al.; J of Cerebral Blood Flow and Metabolism; Vol 7, No 2, 1987; pg. 143-153.

Thanks to:

• Joe Villani, University at Buffalo Dept. of Nuclear Medicine - for supplying the Beta Detector, some electronics/cables, and the shielding
• Alan Lockwood, University at Buffalo Dept. Nuclear Medicine and VAMC - for equipment funding
• Ed Bednarczyk, University at Buffalo Dept. Nuclear Medicine - for tubing, plumbing pieces parts and discussion of system he used in Cleveland.
• Raymond Raylman, University of Michigan Ann Arbor - for discussion of UM's beta detector design

Equipment

• Electronics
  • 486-33 SX acquisition computer w/ ISA card slot for MCS-plus(TM)
  • EG&G Ortec ACEMate(TM) Model 925-SCINT Amplifier and Bias Supply
  • EG&G Ortec MCS-plus(TM)
    • SCA sweep mode (10V / 128 channels) to emulate MCA acquisition
    • MCS sweep mode with independently adjustable lower and upper discriminator window thresholds
  • PMT base to signal/HV splitter and signal/HV cables (courtesy Joe Villani)
  • Ludlum Model 44-1 Beta Survey Detector (courtesy Joe Villani)
  • 0.25[cm] thick x 11.6[cm^2] surface area window (plastic scintillator)
  • 0.8 [mg/cm^2] metalized mylar over detector window for light tightness
  • 2[in] OD x 6.5[in] length self contained 10 stage PMT/detector package
• Shielding

• Octagonal machined lead casing with handle surrounding PMT/detector assembly; 25.5x10x10[cm] outside dimensions, 2.54[cm] thick (courtesy Joe Villani)

• Withdrawal Device

• Harvard '33' Syringe Pump

• Plumbing

• catheter assembly
• detector tubing: Teflon coated; OD=0.5[mm], ID=0.41[mm] (courtesy Ed Bednarczyk)

Operating Design/Characteristics

• tubing
  • detector to patient tubing length (excluding catheter): 20[cm]. This may be reduced in the future (7-10 [cm] is feasible)
  • tubing is coiled (~0.1[ml]) in front of detector; an additional ~0.1[ml] of entry/exit tubing that will contribute to some of the counts seen by the system.
  • efficiency*geometry*volume correction factor for converting to actual [µCi/ml] from [µCi] measured by the detector was empirically determined to be 51.1. Given that the detector sees only 0.1-0.2[ml] of sample at any given time and is only looking at roughly 50% of the positrons escaping the tubing, this translates to an approximate beta detection efficiency of 18-36%. The detector has a quoted efficiency of 15% for 14C and 60% for 32P.
• Dispersion in the detector system: External dispersion of the input system (e.g. that occurring outside the body and due to the catheter/detector assembly) can be determined by measuring the rising/falling edges of the detector system's response to an input step function (Error Analysis of a Quantitative Cerebral Blood Flow Measurement Using H2-15O Autoradiography and Positron Emission Tomography, With Respect to the Dispersion of the Input Function, H. Iida, et.al., J of Cerebral Blood Flow and Metabolism; Vol 6, No 5, 1986; pg. 536-545). Using the data from the tail portion of the subject raw TAC below, one can see that the time to rise or fall to half-maximum in the tail section of the curve when the 3-way stopcock was being switched between heparinized saline and the arterial line is always <= 1[sec]. Approximating the rising/falling curve as a mono-exponential of the form f(t)=exp(-t/tau) to gives us a tau of < (1.0[sec]/-ln(0.5)) or tau < 1.5[sec], which can be considered negligible.
• Linear activity range: Approximately 30 to < 0.5[µCi/ml].
• Background: measurements show that the background contribution of 511[keV] photons doesn't exceed 0.15[µCi/ml] in a typical 15-O water study with an initial injection of 70[mCi].

Example Data Curves

Representative TAC

These curves show the raw TAC, a quantitative TAC derived from the raw TAC, and the typical background underlying the quantitative TAC.

• subject raw TAC (14,412 bytes) This image shows the raw Time Activity Curve (TAC) as it appears on the MCS acquisition display. It was acquired during a ~10[sec] slow bolus injection of 70[mCi] of 15O -water. Sampling was begun at the start of injection and was performed at 0.5[sec] intervals for 3[min] at a withdrawal rate of 12[ml/min]. Initial activity onset to initial curve peak (indicated by vertical line, counts in this channel listed at bottom of display) occurs over 18[sec]. In the >120[sec] section of the TAC, the input to the detector was switched several times via a 3-way stopcock between the arterial line and a pressurized IV line filled with heparinized saline (with the withdrawal pump running continuously). This permits us to measure the 511[keV] gamma background and provides the raw data necessary for us to determine dispersion in the detector system.
• subject quantitative TAC (6,219 bytes) This plot shows the same TAC as above in [µCi/cc]/[sec] for the first 120[sec] after a) decay correction of 15-O for count duration and back to time of injection, b) correction for volume of 15-O in front of the detector, and c) correction for detector geometry/efficiency factors.
• subject quantitative background (7,129 bytes)This plot shows the background counts seen by the detector system over the first 120[sec] (same conditions as above). It was acquired by flushing the detector tubing with heparinized saline and counting as above following the subject's 2nd 15-O injection (no withdrawal pump used and no arterial blood in front of the detector). You can see from the plot that this background never exceeds a contribution of ~0.15[µCi/cc] to the overall TAC.
• hand sampled quantitative TAC (5,465 bytes) This plot shows a TAC acquired by the old hand-sampling technique in [nCi/cc]/[sec]. In particular, note the severe undersampling on the curve's peak (basically 4 points covering the entire rise and initial fall).

15O Calibration

The data in the following curves were derived by drawing 1.0[mCi/cc] of 15O -water into the detector assembly and initiating counting measurements after the activity had decayed to a level at the high end of the range normally observed in patient TAC curves (in this case, we waited 4 x 123[sec] half-lives). Data was then acquired for 6 half-lives to determine the response characteristics of the detector system over the entire range of TAC levels expected.

• raw data (18,689 bytes) As it appears on the MCS acquisition screen. The main graph is integrated counts over the 0.5[sec] counting interval for each channel. The small graph in the upper right hand corner is this same graph with ln() scaling on the Y-axis.
• [µCi/ml] (5,218 bytes) Data after applying analytical corrections for decay during the 500[ms] counting interval and empirically determined correction factors for volume of sample in front of the detector, geometry of sample/detector, and efficiency of detector. The actual data between the two vertical lines (bounded with triangles) is fit to determine the half-life of the sample and its actual activity in [µCi/ml]. In this case, the half-life was found to be 123.099[sec] which is in good agreement with the known half-life of 15O. An ideal curve is then generated using this data and overlaid on the raw data to illustrate count rates where dead time may pose a problem (e.g. where ideal curve and actual data deviate).
• ln([µCi/ml]) (5,367 bytes) Same as [µCi/ml] graph, but after taking the natural log of the activity
• Percent Error (19,347 bytes) [µCi/ml] fit vs. measured. This graph plots the percent error of theoretical (fit) data vs. actual (measured) data based on the two curves plotted in the [µCi/ml] plot above.

Sensitivity to 511[keV] background from patient

The data in the following curves are data as acquired from a study where the subject was injected with 70[mCi] of 15O -water. The "Marker Channel" is the current channel who's data is displayed at the bottom of the screen. This channel is represented visually on the display by a vertical white line running the full vertical width of the plot.

• signal & background (14,915 bytes) Tail end of TAC acquisition (from approximately 120-180[sec] post injection of quantitative TAC above) where, at the end of the acquisition, the detector line was switched three times between the arterial line and a heparinized saline IV line (third switch is at last few points of the acquisition). You can see the rapid drop in counts which occurs when the saline reaches the detector. At this point, until the line is switched back to the arterial line, all counts (or lack thereof) are coming from background 511[keV] photons.
• background (14,347 bytes) Acquisition of background data only (from 0-120[sec] post injection, no blood drawn through the detector system following injection). Compare the maximum count rate seen here ~126[cps] with ~17,660[cps] seen at the 20[µCi/ml] level and ~800[cps] seen at the 1[µCi/ml] level.
• SCA sweep (11,075 bytes) This image shows an SCA sweep from our detector system while measuring background from a subject (no activity in the detector's tubing). In Single Channel Analyzer (SCA) sweep mode, a very narrow acceptance window (a "single channel") is rapidly and repeatedly swept across the entire voltage range (0-10[V]) of the detector's amplified output, dividing that range into 128 separate channels. Events detected which result in a particular output voltage are then accumulated in the appropriate channel. Note that this is the "poor man's" equivalent of a Multi-Channel Analyzer (MCA) acquisition (where all channels are observed simultaneously) since SCA sweep only records an event at a particular voltage if that event's voltage matches the current SCA sweep channel. Our hardware does not support MCA acquisitions, however, the SCA sweep is sufficient for acquiring an energy spectrum of incoming events from a relatively stable source. The position and width of the voltage acceptance window utilized in Multi-Channel Scalar (MCS) mode, which records the number of events falling within the voltage acceptance window over multiple time points during TAC acquisition, can therefore be determined and set so that background and other signals we aren't interested in can be omitted from the TAC. For our studies, the window is set to exclude the voltages at the low end of the spectrum (more easily seen as the larger peak at the leftmost edge of the inlaid semi-log plot), and to include everything above that up to the 10[V] limit.

Other sites and their solutions

coincidence detector - R. Paul Maguire; PET Program, Paul Scherrer Institute

Future Work

• possibly incorporate dead time corrections to permit measurement of higher [µCi/ml] levels
• link detector system, Harvard pump, and PET camera so that acquisition start may be accomplished with a single button press
• incorporate 15O bolus infusion via Harvard pump
• see if we can see any of the lower-energy betas produced by 18F, and see if signal/background is high enough to make this device useful for 18FDG TAC determinations.
• See if we can develop a non-invasive system for capturing 15O input curves.