Retrospective analysis of reduced energy switching and room switching times on throughput efficiency of a multi-room proton therapy center

Abstract

Objective:

To quantify how a control software upgrade changed beam delivery times and impacted efficiency and capacity of a multiroom proton therapy center.

Methods:

A four-room center treating approximately 90 patients/day, treating for approximately 7 years with optimized operations, underwent a software upgrade which reduced room and energy switching times from approximately 30 to 20 s and approximately 4 s to ~0.5 s, respectively. The center uses radio-frequency identification data to track patient treatments and has software which links this to beam delivery data extracted from the treatment log server. Two 4-month periods, with comparable patient volume, representing periods before and after the software change, were retrospectively analyzed.

Results:

A total of 16,168 and 17,102 fields were analyzed. For bilateral head and neck and prostate patients, the beam waiting time was reduced by nearly a factor of 3 and the beam delivery times were reduced by nearly a factor of 2.5. Room switching times were reduced more modestly. Gantry capacity has increased from approximately 30 patients to 40–45 patients in a 16-h daily operation.

Conclusions:

Many proton centers are striving for increased efficiencies. We demonstrated that reductions in energy and room switching time can significantly increase center capacity. Greater potential for further gains would come from improvements in setup and imaging efficiency.

Advances in knowledge:

This paper provides detailed measured data on the effect on treatment times resulting from reducing energy and room switching times under controlled conditions. It helps validate the models of previous investigations to establish treatment capacity of a proton therapy center.

Introduction

Proton therapy has undergone significant growth in the past few years. In the USA, the number of operating centers has increased from 10 in 2011 to approximately 30¹ in 2019 despite the higher operating and capital costs. Globally, there are nearly 90 centers operating in 2019.² The growth in the number of rooms has come from the construction of several multiroom centers and more recently, from installation of single room facilities. While technological and reimbursement strategies have been proposed,^3,4 current practice is constrained by existing technologies and cost structures. Among these constraints, in multiroom centers, there is one proton source (cyclotron or synchrotron) that is switched between different rooms. Hence, treatment of a field in one room cannot start until the current field in another room is finished. In a perfectly orchestrated system, patient setup could be coordinated to minimize different rooms concurrently requesting beam, however, in practice, this is difficult for a variety of reasons including variable patient setup times, changes in patient schedules and variable beam delivery times. This may lead to delays and reduce the number of patients that can be treated. Treatment capacity of any multiroom center is affected by multiple factors including number of rooms, treatment plan, case mix, room configuration, room switching times, dose rate, energy switching time, downtime and patient/therapist workflow.⁵ The interactions between these components are difficult to optimize a priori and while many business plans for multiroom centers attempt to project capacity and reimbursement; they are fraught with uncertainties.

Pencil beam scanning⁶ (PBS) or spot scanning with protons uses magnets to sweep protons over a volume. Specifically, individual layers are set for a particular energy and spots of different intensities and locations are delivered for each layer according to the treatment plan. In our cyclotron system (IBA Proteus Plus 235, Louvain-La-Neuve, Belgium), the protons are produced at fixed beam energy of 232 MeV and the energy is reduced by rotating a degrader (low atomic number wedge) to a thicker position to reduce the energy and hence, the penetration depth of the Bragg peak, or range. Different layers are delivered in sequence and the process is repeated to sculpt the treatment volume. Delivery time is affected by energy switching, spot layer scanning and spot dwell times. It is difficult to project the influence of any one of these parameters on a center’s capacity owing to the inherent complexity of patients, plans, delivery mechanism and staff experience. Moreover, typically analysis of the data provide retrospective information to determine how efficiencies may be improved in future systems, since upgrades of existing systems may not be feasible as the center continues treating. In contrast, we underwent a change in the energy and room switching times which motivated our retrospective analysis. Specifically, a software change resulted in significant reductions in these two variables while holding all other variables nearly constant. We report on the resulting changes in patient throughput and capacity.

Method and materials

Our center consists of four rooms: one gantry and one fixed beam at IEC gantry angle 90°, treating with PBS and two inclined beam lines (IBL). The IBLs consist of two fixed beams with gantry angles of 90° and 30° which treat using uniform scanning (US)⁷ which mimics double scattering using compensators and apertures to shape the beam. The center has been operational for approximately 7 years at nearly full capacity over much of the time. Consequently, center operations are mature to maximize efficiency without compromising quality. The center capacity varies with case mix and has often been bottlenecked by gantry patient capacity where the most complex cases are treated. Through the use of prone setups and a treatment chair,⁸ the IBLs are able to accommodate many patients. Historically, the center has treated over 110 patients in a single day, although typical patient loads are approximately 90 patients in 16-h treatment day.

Data were extracted from a commercial software system (Transeo, San Francisco, USA) that interfaces with the vendor’s treatment software, the oncology information system (OIS) and a radio-frequency identification (RFID) patient tracking system. The software obtains several metrics directly from the vendor’s treatment control software. These include the beam on time, the room switching time, defined as the time the cyclotron is in the switching state, and the beam waiting time, defined as the time between a beam request and the start of tuning. Patient scheduled times were obtained directly from the OIS and actual treatment times were measured using the RFID system. The data were sampled in time resolutions of 0.1 min and binned into histograms for analysis.

On 11 June 2018, the vendor upgraded the software control system in the center to a new version called ‘R8’. The software upgrade resulted in a reduction in the energy switching times for PBS only. This was accomplished through software updates that permitted reduced computation time for each layer, running more software processes in parallel and minimizing delays for equipment setpoint stabilization. US deliveries (IBLs) were not affected by the software upgrade. The software enabled the treatment system to reduce the energy switching time by as much as a factor of 10 from approximately 5 to 0.5 s although the beam switching times did vary depending upon the proton beam energy and the difference in energy layers. R8 also automated the process for room switching rather than requiring the main control room operator to keyboard the change.

The impact on the center operations is evaluated on a retrospective basis using the metrics listed above. The initial month after the software upgrade was excluded as operations and scheduling were not optimized while the center assessed the stability of the changes and implemented changes. The subsequent four months (July 1 to October 31) were chosen for analysis. Review of the total population of patients was performed; however, since changes to delivery and treatment times could be masked by changes in case mix, we also examined subpopulations of patients undergoing similar treatments over the two time periods. Specifically, we selected (1) prostate and (2) bilateral head and neck patients for this analysis for several reasons. First, we have a substantial volume of these patients providing good statistics. Second, the PTV do not vary dramatically between patients and third, the patients are treated with standard beam geometries. Specifically, bi-lateral head and neck patients are treated with two posterior obliques treating the neck nodes and an anterior–posterior field treating the primary disease. Low and intermediate risk prostates are treated with alternating single lateral fields. High-risk prostates, treated with bilateral fields to cover the nodes and exclude bowel, are excluded from this analysis. Finally, we performed the same analysis for US breast patients to act as a control variable.

The overall treatment efficiency is governed by a variety of factors including beam on time. The scheduled time includes the time required for the patient to enter the treatment room, be greeted, set up, imaged, treated and to leave the room. The overall beam on time as a fraction of the scheduled appointment time from our OIS was calculated before and after the software change. The absolute changes in the scheduled time reflect the efficiency gains that could be achieved while the fractional gains establish the upper limit that delivery times have on patient throughput. We also calculated an additional parameter called slot efficiency which is the difference between the scheduled time in the OIS and the actual treatment time from the RFID measurements.

Downtime also affects throughput. Downtime in our center is recorded using the same software system. The overall availability, which is tabulated on a weekly basis, is defined as:

$${\displaystyle \mathrm{A}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}=\mathrm{U}\mathrm{p}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}/(\mathrm{u}\mathrm{p}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}+\mathrm{d}\mathrm{o}\mathrm{w}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e})}$$ (1)

Results

The average system availability was 98.0 and 98.3% for the 2017 and 2018 time periods, respectively; so its effect is ignored in the subsequent analysis.

Table 1 provides a summary of the patient volume and total number of fields treated for the two time periods for overall, prostate and bilateral head and neck patients. In all cases, the values are comparable but not identical, reflecting differences in both volume and case mix. Additionally, mean and standard deviations of the beam on times and beam hold times for both sites are shown.

Table 1.

Number of treatment fields overall, prostate and bilateral head and neck patients. The mean beam on time for all fields were extracted only for the PBS rooms

	Total		Prostate		Head and Neck
	2017	2018	2017	2018	2017	2018
Treatment Fields	1,6168	1,7102	1296	1089	2272	2218
Mean ± SD Beam on time PBS [min]	2.63 ± 1.36	0.81 ± 0.67	2.10 ± 0.96	0.61 ± 0.24	2.04 ± 1.03	0.79 ± 0.35
Mean ± SD Beam waiting time PBS [min]	2.17 ± 2.39	0.76 ± 1.06	1.92 ± 2.20	0.66 ± 0.82	1.7 ± 2.04	0.58 ± 0.69

Figure 1 (a) shows the normalized histogram of the beam waiting time and (Figure 1b) shows the corresponding beam on time for head and neck patients. Similarly, (Figure 1 c,d) show the corresponding values for prostate patients. The distributions are significantly different by inspection for both cases, which was confirmed by student’s t-test which found p << 0.05 for all cases.

Figure 1.

(a) Relative histogram of field beam on times for bilateral head and neck patients before (2017) and after (2018) the software change. (b) Corresponding beam waiting times, defined as the time between request and first tuning, for head and neck patients. (c) Relative histogram of field beam on times for prostate patients before (2017) and after (2018) the software change. (d) Histogram of corresponding beam waiting times for prostate patients. (e) Histogram of beam on times for uniform scanning breast patients showing no changes in beam on times.

There were 1024 breast treatment fields for 2017 and 584 for 2018 on the IBL. These breast patients were treated with two matched superior–anterior obliques and two matched isiplateral anterior obliques. Figure 1(e) shows that the IBLs were not affected by the software upgrade and act as a control variable to the analysis (p = 0.06).

Figure 2 shows a histogram overlay of the patient time in the treatment rooms for a) prostate patients and b) head and neck patients. The mean and standard deviation are for head and neck patient was 27.5 ± 8.5 min in 2017 compared to 23.8 ± 8.8 min in 2018. For prostate patients the corresponding values are 12.2 ± 3.5 min and 10.3 ± 2.4 min. The ratio is approximately 0.84 in both cases.

Figure 2.

(a) Histogram of patient in room time for (a) prostate and (b) head and neck patients.

The mean room switching time decreased from 0.51 ± 0.49 min to 0.32 ± 0.37 min. Room switching times were reduced in part because a number of steps that are operator-dependent were reduced or eliminated. Comparing these values with Table 1, the impact of room switching is smaller than the impact of energy switching times. Specifically, the overall beam delivery time was reduced by a factor of over three with the software upgrade, whereas the room switching time was reduced by a factor of 1.6. Moreover, the total time for delivery is per field, whereas the room switching time is per patient. Hence, the effects of delivery time on overall treatment time are amplified further by the number of fields.

Table 2 shows the room distribution before and after the software upgrade. Capacity has increased in the PBS rooms to the point where approximately 75% of patients instead of 50%. The distribution of patients between patients changed substantially with the increased gantry capacity such that in early October; the center had only one patient in one of the IBLs and eventually closed one of the US rooms, reducing beam competition between rooms.⁹ The number of patients treated in the gantry was 25.3 ± 3.5 (max 31) in 2017 vs 32.8 ± 4.9 (max 44) in 2018. Our experience demonstrates that gantry room capacity has increased from approximately 30–35 patients to 40–45 patients depending upon case mix.

Table 2.

Average fraction of patients treated in room 1 (PBS Fixed beam), Rooms 2 and 3 (US IBL) and Room 4 (PBS Gantry) for two periods retrospectively studied

	2017	2018
Fixed beam (PBS)	24	37
Both IBLs (US)	52	27
Gantry (PBS)	24	36

Table 3 shows the slot utilization for all patients, prostate and head and neck subsets for both 2017 and 2018 periods. On average, there is an extra 5 min between each patient appointment regardless of the site or time period. This time permits the therapist to greet, queue in and escort the patient in and out of the room. Since the workflow did not change between the two periods, the time is constant.

Table 3.

Slot utilization: the difference between the RFID measured time and the scheduled time in the OIS

	Overall (mins)	Prostate (mins)	Bilateral Neck (mins)
2017	−5.5 ± 13.3	−4.74 ± 9.1	−4.5 ± 8.6
2018	−5.74 ± 6.32	−3.10 ± 4.6	5.8 ± 5.9

Discussion

The cost of proton centers is substantially higher than photon centers. Capital costs remain a substantial component of this cost but operating costs¹⁰ are also higher. Improvements in efficiency for quality assurance through new hardware and software approaches,^11–15 the replacement of double scattering with PBS where possible, and integration of established workflows from external beam treatments may result in more efficient treatments. Here, we have demonstrated that improvements in delivery speed can significantly affect treatment capacity in a multiroom center.

The overall beam on time for treatment has had a significant impact on the center and its ability to treat patients. The overall efficiency of the center has increased to the point where we are able to treat a comparable number of patients in fewer rooms. However, we caution that other factors, such as changing out apertures and compensators, may make US rooms less efficient so a one-to-one relationship likely does not exist. Moreover, the ability to treat more patients on the gantry may cause the less efficient, but equally effective treatments, to be under utilized. For instance, the center has a treatment chair that is used for treating cranial patients. With the increased gantry capacity, it is easier to use the gantry for a posterior beam rather than rotating the patient. We caution that energy switching time one of many parameters that affect throughput. Another factor is the choice of room configurations, hence a center with three gantries instead of one gantry and two IBLs could have a different capacity. Our goal was to quantify the effect on a gantry which has high demand rather than to produce a model that could be applied globally.

We compared our data to published models of throughput capacity. Specifically, we compared our data with the model of Aitkenhead et al⁹and the model of Suzuki et al⁵ . We caveat that our rooms do differ from those modeled and that our treatment approaches may differ. We reduced our energy switching time by a factor of approximately 10. We found that overall, the average number of patients that could be treated in the gantry increased by a factor of 1.3. The beam on times were reduced by a factor of approximately 2.5 to 3 based upon the head and neck and prostate data and over the entire patient population by a factor of 3.2. While the models are not identical, our experience provides some empirical comparison with the model. Aitkenhead et al estimate that if there are increases in beam switch and delivery time by 50%, the center capacity decreases by approximately 15%. On a per patient basis, this is reasonably consistent with our reductions of approximately 15% patient in room time.

Suzuki et al⁵ reported energy switching times of 2.1 s and found that beam delivery times were about 30–40% of the total treatment time for 80% of 64 patients studied. To generate comparable data, we calculated Suzuki’s ‘equipment time’, defined as the sum of the time total beam on time and beam waiting time. For N fields,

$$\begin{gathered} {\displaystyle \mathrm{E}\mathrm{q}\mathrm{u}\mathrm{i}\mathrm{p}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}=\mathrm{N}\times \mathrm{b}\mathrm{e}\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e} \\ +(\mathrm{N}-1)\times \mathrm{b}\mathrm{e}\mathrm{a}\mathrm{m}\mathrm{w}\mathrm{a}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}} \end{gathered}$$ (2)

where the N-1 results from the fact that there is no waiting once the final field has been delivered. Despite the differences in energy switching times between our system and the one at MD Anderson, for the time period studied in 2017, our equipment time was comparable to theirs, at 34 and 37% for prostate and head and neck data respectively. However, for 2018, the values had dropped to 14 and 12%, reflecting increased efficiency. Hence, while the software upgrade did not reduce the beam waiting time to zero, further improvements in efficiency through beam delivery are limited. Improving setup or imaging efficiency may lead to more significant gains.¹⁶

While the measurable effect of faster treatments are in terms of throughput and efficiencies, there are other less quantifiable benefits as well. For instance, the shorter time on the treatment table in an uncomfortable immobilization system may improve the patient experience. The reduction in beam competition also reduces stress between the treatment personnel in different rooms who are trying to stay on schedule.

Conclusions

To our knowledge, this is the first empirical report of the effect of changes energy switching time in a capacity constrained multiroom proton center. Reductions in beam on times through reduction of energy switching time have resulted in substantial gains in throughput and efficiencies. Vendors are expected to continue to add contributions to efficiency since this reduces their costs and makes their products more competitive. Further improvements in efficiencies are possible with faster delivery times, but greater gains may be obtained from reductions in setup and imaging times.