ESTRO 36 Abstract Book

S909

ESTRO 36

_______________________________________________________________________________________________

were the DVH metrics used during the treatment planning

for each considered OAR (e.g. D

, V

) and TVs (e.g D

Dose constraints were also defined according to the tumor

site (e.g. D

mean

Parotid < 30 Gy). Two levels of warning

were considered:

•

red flag: a 10% deviation of the clinical indicator

relative to the planned value (e.g. for the

parotid ΔD

mean

(cumulated)>10% D

mean

(planned))

AND

a violation of a dose constraint (e.g. for the

parotid D

mean

(cumulated) >30 Gy)

•

orange flag: a 10% deviation of the clinical

indicator relative to the

dose constraint

(e.g.

for the parotid ΔD

mean

(cumulated) >3 Gy).

Both adaptive software evaluated the dose to TVs using

deformed PTVs. This approach is questionable because the

PTV corresponds to a geometrical (not anatomical) safety

margin. Therefore, we reported the dose on rigidly

registered PTVs.

Results

Deformed contours were judged acceptable for all H&N

and lung cases. However, registrations failed for most

pelvic cases, for which large anatomical deformations

occurred (see figure 1). Consequently, pelvic cases were

excluded.

Dose calculation of both analytical engines were in good

agreement with TomoPen (around 1.5% mean difference

on PTV D

Results are reported in Table 1. For TVs, only 6 flags (out

of 62 patients) were reported for the rigidly registered

PTV, which was considered as the only relevant volume.

The flags reported for lung cases were irrelevant because

of the blurring of the tumor density leading to large dose

calculation deviations. For the H&N case, the red flag was

rejected after analysis (wrong doses in part of the PTV out

of the external contour). For the OARs, one H&N was

flagged (true flag) with an increase of 11% of the mean

parotid dose that exceeded the dose constraint (30 Gy).

Conclusion

Considering a constant PTV, the impact of treatment

adaptation on the quality of delivered plans is minor for

the included patients. The conclusion might be different

for pelvic cases due to the larger anatomical

deformations. Conclusions might also differ for an adapted

PTV, but such strategy must address clinical

considerations before implementation.

EP-1670 Couch shifts in NAL protocols: ¿Which is the

optimal threshold?

A. Camarasa

, V. Hernández

Hospital Universitari Sant Joan de Reus, Servei de

Protecció Radiològica i Física Mèdica, Reus, Spain

Purpose or Objective

The NAL protocols applied to patient positioning in

treatments evaluated by CBCT use a threshold regarding

couch shifts. If the CBCT demands shifts over the

threshold, the patient must be moved, while shifts below

the threshold remain as residual errors. The aims of this

study are: (1) to determine the relation between the

treatment positioning uncertainty and the corresponding

workload, and (2) to obtain the optimal threshold for

couch shifts in prostate treatments.

Material and Methods

The quadratic sum of the uncertainties associated with

patient positioning is calculated. If the proposed shifts

remain below the threshold, the uncertainties are related

to the CBCT matching procedure and to the distribution of

residual errors. If the shifts are over the threshold, the

uncertainties are due to the couch movement accuracy

and, again, the CBCT matching procedure. The

relationship between treatment positioning uncertainty

and workload was optimized using the threshold for couch

shifts as an independent variable. Partial uncertainties

were computed based on 811 CBCT clinical cases, together

with the historical QA matching results from OBI’s

equipment and measurements from Varian’s couch

accuracy. The total positioning uncertainty with K = 2 was

calculated for VMAT treatments delivered in 28 sessions

with daily CBCT. The workload was estimated from the

probability of couch shifts, which was derived from the

statistics of the 811 clinical cases.

Results

The positioning uncertainty and the probability of couch

shifts as a function of the chosen threshold are shown in

Figure 1. As expected, if a high threshold is used (greater

than 12 mm) the workload is minimized but uncertainty is

stabilized at an excessively high value. On the contrary, if

a very low threshold is used, i.e. between 0 and 2 mm, the

probability of couch shifts is very high (between 97% and

100%). In this case, interestingly, the total uncertainty is

not significantly reduced due the contribution of the

remaining factors. Thus, the chosen threshold should be

between 2 and 12 mm. To facilitate the determination of

the optimal threshold, the derivations of both functions

are shown in Figure 2. It can be observed that uncertainty

has a maximum increase when the threshold is raised from

5 to 8 mm. However, if the same procedure is applied to

the probability distribution of couch shifts, the maximum

decrease takes place for a threshold between 4 and 5 mm.