The introduction of new treatment techniques (conformal radiotherapy, intensity modulated radiation therapy - IMRT) have underlined the importance to investigate dynamic beam characteristics like the startup behavior of a medical linear accelerator. As the number of treatment fields per session of a conformal treatment increases, the number of monitor units for each portal gets smaller. If the beam only stabilizes after a considerable fraction of time with respect to symmetry and dose rate, the startup phase may affect the dose distribution in the target region. The same holds for quasi-dynamic treatments (step-and-shoot technique), where a sequence of static fields with different shapes is delivered to the treatment volume with the gantry at a fixed position. Here the beam-on sequence is performed several times with small monitor unit values for the individual segments.
An influence on dose distribution is also imaginable in situations where the dose rate is regulated dynamically under control of a computer. With Varian's Enhanced Dynamic Wedge (EDW), in the first part of the EDW treatment a certain fraction of the total dose is delivered to the full field (open field phase). Then one of the upper jaws starts to sweep the field (collimator sweep phase) until the jaws are in an almost closed position. Beam is on during the whole treatment, but dose rate can vary quite significantly, especially at the transition from the open field phase to the collimator sweep phase.
The Varian Dynamic Multileaf Collimator (DMLC) is a dynamic treatment option that includes both the step-and-shoot technique and the continuous beam and motion technique. With the former, dose rate is servoed to a user defined value during the 'shoot' segments of the treatment, followed by move-only segments where the leaves move to the next shape. At the transition a sharp drop of dose rate to zero is achieved by delaying all gun pulses to prohibit beam output, but the beam stays actually on during the move-only segments. It is certainly interesting to measure beam quality under such rapid changes of dose rate. The continuous beam and motion technique of the DMLC has no such sharp changes in dose rate, since pulses are only dropped if one or more leaves would have to exceed maximum leaf speed. Generally, the MLC shape is changing constantly during beam-on. At each instant, the actual shape is calculated as the result of a linear interpolation between two shapes and their respective dose fractions (between 0.0 and 1.0).
Standing wave accelerators should perform better during startup compared to travelling wave machines. We investigated the startup behavior of our two Clinacs 2300C/D which are of the former type. The machines are equipped with Electronic Portal Imaging Devices (PortalVision), which require special dose rate control to ensure good image quality. During photon treatments, dose rate is controlled by a Pulse Length Servo, for Electron treatments a Pulse Drop Servo (PDS) is used. A bad startup performance could also have influence on the image quality of a portal image, which requires an average dose of 9 monitor units (MU) in our department. If the image is acquired before the beam stabilizes, image quality could be affected.
Coming to the point ...
This Winter 1998 PTW Freiburg introduced an extension to their LA48 Linear Array (Fig.1). The array consists of 47 ionization chambers (2.5 * 2.5 mm, 8 mm spacing). A new amplifier with a time constant of 10 ms (ME48F) allows measurement of 100 dose profiles per second. The acquisition of the profile itself takes 2 ms, after another 8 ms the next profile can be measured. The channels are read out sequentially, which is quite fast (42.5 microseconds per channel!), but this can still lead to artefacts, if the beam rises fast enough (as you will see below).
The measurement starts after a user-selected dose rate level (e.g., 0.02 Gy/min) is reached. A user selectable number of profiles (10 in my example) before reaching the level are also stored. Data can be exported to PTW's Mephysto software for further analysis.
The array was mounted in-plane in the PTW waterphantom (without water), SSD was set to 80cm, so the jaws could be set to 40 by 40 cm. The Clinac was set to HiX (15MV), Dose rate 600MU/min, and beamed-on. Scan time was 2ms, repetition time 10ms. Fig. 2 shows the first 250 profiles in the ME48F software.
Fig.2: Measuring window of ME48F software showing 250 profiles.
By pressing the 'Single Channel' button, one gets the time-dependent dose rate of a user selectable channel (Nr. 24 shows the central axis): see Fig. 3.
Fig.3: Dose rate at central axis
As can be seen in Fig. 3, at t=100ms Beam-on was pressed. The beam rises in no-time-at-all and reaches 95% of maximum dose rate (7.855 Gy/min) after 79 milliseconds (t=179ms).
The first true beam profiles above noise level were recorded at t=104ms and t=114ms. The following animation (Fig.4) shows all profiles (even the 10 before reaching the trigger level, so they are just noise) normalized to 100% at central axis (i'm sorry that the profile at t=188ms is missing, but there is nothing special about it -- and doing all these screen shots, converting them to GIFs and aligning the frames is a pain-in-the-ass!). The bad symmetry of the first few curves (especially at t=104ms and t=114ms) comes from the measurement system, not from the Clinac. This was checked by mounting the array the other way around because with the rotated setup curves were still rising to the right. PTW told me that they checked their measurement system on Siemens machines and found no such effect ...
So after measuring all combinations of Energy (also Electrons!) and dose rate from 1Gy/min to 10Gy/min I found out that the Clinac beam rises absolutely symmetrical. The measurement system is state-of-the-art and not cheap, but it gets to its limit at our machines. Nevertheless, it's a very fine system and I can recommend it to anyone who is interested in the startup behavior of his machines. To be honest: the problem of beam quality after 14 milliseconds is somewhat academic (dose rate at 14ms after beam-on is 18%).
Fig.4: Beam developement for the first 388 milliseconds. The 'off-axis' value to the right multiplied by ten gives the time (beam starts at t=100ms).
So if you find this interesting, stay tuned! I already performed some measurements of Dynamic MLC beams ...
back to Medical Physics home
Last modified: Feb.19 1999HK