Dosimetric experience with accelerated partial breast irradiation using image-guided interstitial brachytherapy

Methods and materials 

Twenty-eight consecutive patients treated at our institution with multi-catheter HDR BT alone between September 2004 and December 2006 were selected for the study. The treatments were performed in the framework of a Phase III multicentric APBI study developed by the Breast Cancer Working Group of the Groupe Européen de Curiethérapie-European Society for Therapeutic Radiology and Oncology (GEC-ESTRO) [1]. The patients had Stage I–II breast cancer and underwent breast-conserving surgery with wide tumor excision and axillary dissection or sentinel node biopsy. Study design and eligibility criteria for the GEC-ESTRO APBI trial have been described elsewhere [1]. The delivered dose was 7×4.3Gy, twice daily, in 4 days. All patients were treated with a microSelectron HDR afterloader (Nucletron, Weenendaal, The Netherlands).

Implant technique 

During the operation the excision cavity wall was marked with 4–6 surgical clips. Before the implantation CT imaging with 3mm slice thickness was performed with a plastic template on the breast of the patient. Before the removal of the template its position was marked on the skin with ink. The resection cavity was outlined in axial slices using the visualized seroma and surgical clips as the cavity boundaries, and the PTV was generated with a three-dimensional non-uniform volume expansion using a 3-D planning system (Philips-Pinnacle3 v7.6c). If the cavity was not directly visualized by any of the techniques mentioned before, the patient was not selected for APBI. The safety margin around the excision cavity was calculated in six directions along the three principal Cartesian axes individually taking into account the pathological tumor-free resection margin. In each direction the distance to be used for cavity volume expansion was determined as 20mm minus the tumor-free resection margin (Fig. 1). In this way it was guaranteed that the total margin (tumor-free surgical margin+normal/breast tissue in PTV) was 20mm. At target volume generation the PTV was limited to 5mm below the skin surface and to pectoral muscle. Then, using the needle’s-eye view of the 3-D rendered transparent skin and template with the projected PTV the specific holes to be used for catheter implantation were selected in order to ensure uniform geometrical coverage of the PTV (Fig. 2). Using this information the implantation was performed on the same or next day under local anesthesia using metal needles, which were replaced by flexible catheters. For implantation a more rigid template was placed on the breast in the skin-marked position. Then, the patient with plastic catheters in her breast underwent another CT scanning for planning purpose using 3mm slice thickness. Adjustments in catheter positions were made as necessary in order to get adequate PTV coverage. Catheters outside the PTV were not used for irradiation, and in a few cases additional catheters were inserted with free-hand technique.

View Large Image Download to PowerPoint

Fig. 1. Planning target volume (red) definition with seroma contouring (yellow) and use of individual safety margin around the lumpectomy cavity. Clips are at the boundary of the resection cavity.

View Large Image Download to PowerPoint

Fig. 2. Three-dimensional rendering of a patient’s skin and template on the breast from needle’s-eye view. PTV shown in wire frame is projected to the template for defining the holes (green dots) to be used for inserting the needles.

Planning and dosimetry 

The CT images were sent to the brachytherapy planning system (PLATO v.14.6, Nucletron B.V., Veenendaal, The Netherlands) and the PTV, ipsilateral lung and heart were outlined. The catheters were reconstructed, and reference dose points in the central plane were created between the catheters, similar to the basal dose points in the Paris dosimetry system [14]. Geometrical optimization was applied [15], and the dose was normalized (100%) to the mean dose in the reference dose points. Then, a prescription isodose was selected such that the target volume coverage by the reference dose was at least 90%, while keeping the dose non-uniformity ratio less than 0.35. By renormalizing the dose distribution the selected isodose line corresponds to the prescribed dose (100%). When the use of geometrical optimization did not result in acceptable dose distribution, it was followed by interactive graphical optimization in order to achieve both of our dosimetric constraints. The active lengths in the catheters were selected in such a way that the extreme source dwell positions in each catheter were on or close to the surface of the PTV. The source step size was 5mm. The patients were treated according to these optimized plans. A typical dose distribution is shown in Fig. 3. Retrospectively, another treatment plan was made for each patient with dose point optimization (conformal planning). In this case dose points were automatically placed on the surface of the PTV with 8mm separation, and optimization on all dose points on volume was performed. The dose was normalized and prescribed to the mean dose in dose points. In these plans all source positions were kept inside the PTV at a distance of 5mm from the surface.

View Large Image Download to PowerPoint

Fig. 3. Relative dose distribution of a four-plane implant on an axial CT slice. The reference isodose line (thin red) conforms to the shape of the PTV (thick red).

Results 

In all 28 patients the treatment was finished according to the treatment plan. The implant characteristics are shown in Table 1. Three-, four- and five-plane implants were used in 10 (36%), 13 (46%) and 5 (18%) patients, respectively. The median number of implanted catheters was 14 (range 8–22). The mean volume receiving the reference dose was 75.3cm3 (range 26.6–137.4cm3). The average MCD was 135% (range 125–145%), whereas the volume irradiated by 1.5 times the MCD (high dose volume according to ICRU 58) was 8.3cm3 (range 4.5–21.4cm3).

Table 1. Implant characteristics and implant-related parameters (n=28)

	Dosimetric characteristic	Mean	Range
	Number of catheters	14±3.8	8–22
	Number of implant planes	4±0.7	3–5
	Vref	75.3±32.5cm3	26.6–137.4cm3
	V1.5×ref	24.4±9.7cm3	9.2–42.4cm3
	V1.5×MCD	8.3±3.7cm3	4.5–21.4cm3
	DNR	0.33±0.04	0.25–0.41
	TRAK	0.20±0.06cGy at 1m	0.11–0.30cGy at 1m

Abbreviations: Vref, volume receiving at least the reference dose; V1.5×ref, volume receiving at least 1.5 times the reference dose; V1.5×MCD, volume receiving 1.5 times the mean central dose; DNR, dose non-uniformity ratio; TRAK, total reference air kerma.

In eleven cases out of 28 the geometrical optimization resulted in acceptable dose distribution regarding target volume coverage and dose non-uniformity, while in seventeen cases graphical optimization was applied after geometrical optimization in order to improve the target coverage. The use of graphical optimization increased the mean V100 from 86% (range 73–91%) to 91% (range 90–94%), p=0.0010 and simultaneously the DNR changed from 0.31 (range 0.20–0.42) to 0.33 (range 0.28–0.41), p=0.0091. Since our primary aim was to reach at least 90% target volume coverage by the reference dose, this slight deterioration in the dose uniformity was accepted. The mean absolute volumes of V1.5×ref, V150 and V200 were 24.4cm3, 20cm3, and 7cm3 with maximum values of 42.4cm3, 36.8cm3 and 17cm3, respectively.

On average, 74% of the mean central dose (range 69–80%) was selected for dose prescription in order to achieve our dosimetric goals for target coverage and homogeneity. Volume and dose parameters for PTV and critical structures are presented in Table 2. The mean volume of the PTV was 63.1cm3 (range 17.2–124cm3) and on average, 91% (range 90–96%) of the PTV received the reference dose. The mean and median D90 was 102% and 101%, respectively, with a range of 99–107%. The D100, which corresponds to minimal dose in PTV, was 69% (range 53–92%). The dose homogeneity in the PTV was characterized with a DHI of 0.64 (range 0.50–0.76), while the conformal index was equal to 0.68 (range 0.51–0.88). On average, normal tissue volume around PTV receiving the reference dose was close to one-third of the volume of PTV. The mean EI was 0.32 with a wide range of 0.14–0.69.

Table 2. PTV and critical structures dosimetry (n=28)

	Dosimetric characteristic	Mean	Range
	VPTV	63.1±29.0cm3	17.2–124cm3
	PTVref	57.2±26.8cm3	15.2–118.1cm3
	PTV coverage
	V90	96±1.5%	93–100%
	V100	91±1.6%	90–96%
	V150	33±5.4%	23–45%
	V200	12±3.6%	7–22%
	D90	102±2.1%	99–107%
	D100	69±9.7%	53–92%
	Homogeneity
	DHI	0.64±0.06	0.50–0.76
	Conformity
	COIN	0.68±0.07	0.51–0.82
	EI	0.32±0.13	0.14–0.69
	Skin
	Dmax	53±14.1%	18–75%
	Ipsilateral lung
	Dmax	42±17.0%	7–75%
	V5Gy	42.6±40.7cm3	0–160cm3
	V10Gy	4.8±8.9cm3	0–39.5cm3
	V15Gy	0.5±1.5cm3	0–7.9cm3
	Heart (left sided lesions, n=13)
	Dmax	21±10.8%	4–40%
	V5Gy	8.0±12.7cm3	0–33.5cm3
	V10Gy	0.1±0.3cm3	0–0.3cm3

Abbreviations: PTV, planning target volume; VPTV, volume of PTV; PTVref, volume within PTV receiving at least the reference dose; V90, V100, V150 and V200, percentage volume of PTV receiving 90, 100, 150 and 200% of reference dose, respectively; D90 and D100, relative dose received by 90 and 100% of PTV, respectively; DHI, dose homogeneity index for target volume; COIN, conformal index; EI, external volume index; Dmax, maximum dose; V5Gy, V10Gy and V15Gy, volume receiving more than 5, 10 and 15 Gy, respectively.

The mean maximum dose to skin, lung and heart was 53%, 42% and 21%, respectively. The volume of ipsilateral lung receiving 5Gy, 10Gy and 15Gy was 42.6cm3, 4.8cm3 and 0.5cm3, respectively. The dose values in Gy correspond to 17%, 33% and 50% relative doses, respectively. For left sided lesions the heart volume receiving 5Gy was 8.6cm3, and the 10Gy isodose reached the heart only in 2 of 13 patients.

Correlation analysis was performed between V100, D90 and D100. The correlation coefficient of R2 was 0.867 between V100 and D90, and 0.42 between V100 and D100. The correlation between D90 and D100 was very poor (R2=0.323).

The TRAK was calculated for each dose plan and related to the volumes of the reference dose and PTV. Fig. 4 shows the association between TRAK and VPTV for the patients. As evident from the graph, the correlation between the two parameters is high with R2 of 0.974. The calculated values agree within ±7% of the predicted values of the linear regression. The TRAK also correlated well with Vref (R2=0.98).

View Large Image Download to PowerPoint

Fig. 4. Correlation between total reference air kerma (TRAK) and volume of the PTV for 28 patients.

A comparison was made in selected parameters between treatment plans of the patients and hypothetical plans made by conformal planning. The results are shown in Table 3. The target volume coverage in conformal plans was 3% lower than in the treatment plans and was close to that obtained when only geometrical optimization was used. The volume receiving 1.5 times the reference dose was significantly larger for the conformal plans versus treatment plans in both the whole implant (V1.5×ref) and inside the PTV with 35.5cm3 vs. 24.4cm3 and 34.8cm3 vs. 20.5cm3, respectively (p<0.0001 for both). This larger volume resulted in the worse dose uniformity and homogeneity with higher DNR and lower DHI values. In the conformal plans the D90 and D100 values were significantly lower, and the volume receiving the reference dose was very close to the volume of PTV (64.3 and 63.1cm3). Since after dose point optimization the reference isodose generally followed the contour of PTV tightly, the conformality of the dose distribution was very good (COIN=0.77), much better than in the treatment plans (COIN=0.68), p<0.0001. This was the consequence of the fact that considerably less normal tissue volume was irradiated by the reference dose when conformal planning was used (7.6cm3 vs. 18.2cm3, p<0.0001). With conformal planning the skin dose was significant less (48% vs. 53%, p=0.0011), but the doses to heart and lung were practically the same as in the treatment plans.

Table 3. Dosimetric findings of dose plans with different optimizations

	Dosimetric characteristic	Geometrical and graphical optimization	Conformal dose point optimization	p-value
	Implant related
	Vref	75.3±32.5cm3	64.3±30.3cm3	<0.0001
	V1.5×ref	24.4±9.7cm3	35.5±17.2cm3	<0.0001
	DNR	0.33±0.04	0.54±0.03	<0.0001
	PTV related
	V90	96%±1.5%	93%±4.9%	0.0017
	V100	91%±1.6%	88%±5.3%	0.0013
	V150	33%±5.4%	54%±4.2%	<0.0001
	D90	102%±2.1%	96%±8.9%	0.0028
	D100	69%±9.7%	60%±10.6%	0.0005
	DHI	0.64±0.06	0.38±0.03	<0.0001
	COIN	0.68±0.07	0.77±0.07	<0.0001
	Skin
	Dmax	53%±14.1%	48%±13.5%	0.0011
	Ipsilateral lung
	Dmax	42%±17%	41%±15.1%	0.1952
	V5Gy	42.6±40.7cm3	35.4±35.8cm3	0.0868
	Heart (left sided lesions, n=13)
	Dmax	21%±10.9%	21%±11.6%	0.3522
	V5Gy	8.0±12.7cm3	6.1±13.1cm3	0.6679
	TRAK	0.20±0.05cGy at 1m	0.20±0.06cGy at 1m	0.4901

Abbreviations as in Table 1, Table 2. Data are means±standard deviation.

Discussion 

Historically, most of the studies of interstitial breast BT report implant-related dosimetric parameters, only. However, there are a few recent publications in which results of target-oriented dose–volume assessments are published (Table 4). In the table the studies are divided into two groups. In the first group, the catheters were inserted using standard fluoroscopy-guided techniques followed by conventional planning, and CT scanning was done after the implantation for plan evaluation purpose, only. Whereas, the second group contains studies in which the implantation was performed by image-guided technique using real 3-D anatomical information of the target volume and the plan evaluation was target oriented.

Table 4. Clinical studies reporting dose–volume parameters of high-dose-rate interstitial breast brachytherapy

	Author	n	V90	V100	D90	DHIc
	Standard technique
	Vicini (10)	8	NR	68%	69%	0.89
	Kestin (11)	11	NR	68%	NR	0.83
	Cuttino (28)	15	89%a	96%b	NR	0.77
	Major (9)	17	76%	70%	72%	0.65
	Weed (24)	10	68%	58%	NR	NR
	Image-guided technique
	Kolotas (29)	42	NR	90%	NR	NR
	Das (25)	50	NR	95%	NR	0.73
	Cuttino (28)	14	95%a	98%b	NR	0.82
	Present study	28	96%	91%	102%	0.67

Abbreviations: n, number of patients; NR, not reported. Other abbreviations as in Table 2.

a For PTV 2cm. b For PTV 1cm. c Dose homogeneity index for the implant.

Perera et al. [21] used CT films to determine the volume of the lumpectomy site, the number of implant planes necessary, and the orientation of the implant planes, but treatment planning was still based on orthogonal radiographs. Vicini et al. [22] used CT imaging at implementation of 3-D virtual BT in the management of breast cancer. They used the preimplant images to define the positions of the needles, and following the implantation a second set of CT images were taken for comparison of the actual target volume coverage with the virtual implant generated preoperatively. In another paper, Vicini et al. [10] presented their results of dose–volume analysis for quality assurance of interstitial breast BT. Eight patients treated with HDR breast BT were selected for the analysis in order to evaluate the 3-D relationship between delivered dose and target volume. In the postimplant CT images the lumpectomy cavity and target volume (lumpectomy cavity+1cm margin) were outlined, and following geometrical optimization different volumetric and dose parameters were calculated. For five selected patients, on average, V100, D90 and DNR were 72%, 73% and 0.11, respectively. However, it has to be noted that the implantation was not CT-based. Kestin et al. [11] performed a retrospective CT-based 3-D dose–volume analysis of HDR breast implants to evaluate the dose coverage of lumpectomy cavity and target volume. They used standard geometric optimization, and in the actual implants of eleven patients the median proportion of the lumpectomy cavity and target volume (cavity with 1cm margin) receiving at least the prescription dose was 85% and 68%, respectively. This means that even the lumpectomy cavity coverage was suboptimal. With a simple dwell-time adjustment algorithm they improved the lumpectomy cavity coverage significantly. However, with the improvement of coverage the dose homogeneity worsened. Polgar et al. [23] justified the superiority of conformal BT planning over 2-D treatment planning with better conformality parameters, but in the study no anatomical DVH-s were applied. Weed et al. [24] reported only 58% of V100 for their ten interstitial BT patients. In a previous study we reported 70% target volume coverage by the reference dose for 17 patients treated with fluoroscopy based implantation technique [9]. Das et al. [25] reported their experience with CT-based interstitial breast implants, and they demonstrated the technical feasibility of this approach along with improved target volume delineation and optimal coverage. On evaluating 50 patients the target volume coverage by the prescribed dose ranged between 75% and 100%, with a mean and median value of 95% and 96%, respectively. The conformality of dose distribution, however, has not been discussed. In a number of studies it was found that even with a three-plane implant the target volume coverage by the catheters was insufficient in regions superficial and sometimes deep to the implant [9], [10], [11]. Our data presented here reinforce these findings. Although, almost half of our patients were treated with four-plane and 5 with five-plane implant, we occasionally experienced underdosed region, mainly in anterior direction. In order to cover these areas with the reference isodose we used the graphical optimization, but the improvement in target coverage went with a slight deterioration of dose homogeneity (DNR, 0.33 vs. 0.30). The DNR values reported here are higher than values published by others. However, the high-dose volumes represented by V1.5×ref (24.4cm 3) are much lower than the upper limit (70cm3) used in the NSABP B-39/RTOG 0413 protocol for multi-catheter system [26]. In another study from our institution with comparable high-dose volumes and DNR values no association was found between V1.5×ref or dose homogeneity and the incidence of fat necrosis [27]. It has to be also noted that the high dose volumes occur mostly in the resection cavity with no clinical consequences. The DNR value could be decreased by selecting a higher isodose line for prescription. However, then the PTV coverage would be lower. We intend to have at least 90% PTV coverage by the prescribed dose, while in the NSABP B-39/RTOG 0413 protocol the 90% of the prescribed dose should cover at least the 90% of the PTV [26]. With the latter constraint a lower DNR value can be obtained.

The quality of dose distributions in interstitial BT can be improved with the use of different dose optimization methods and image based 3-D information for planning the geometry of the catheter positions [9], [10], [11]. With conformal dose planning the target volume coverage can be significantly increased compared to the traditional dosimetry systems, but the dose homogeneity may worsen [9]. The use of 3-D imaging before or during the implantation can improve the dose delivery regarding both conformality and homogeneity, which may turn into improved clinical results. Cuttino et al. [28] reported that the percentage of patients satisfying their dosimetric goals of target coverage and dose homogeneity increased from 42% to 93% when CT-guided technique was used instead of fluoroscopic-guided free-handed catheter insertion technique. In their study not only the coverage but also the DHI was better with the CT planning (Table 4). As evident from Table 4, much better target volume coverage can be achieved with image-based implant technique than with conventional one. The question of homogeneity is still debatable.

Different image-guided interstitial implant techniques exist [10], [28], [29], [30]. To outline the lumpectomy cavity, and to determine the number and location of catheters to be implanted can be determined by using image information of digital mammography, ultrasonography or CT scanning. Clinical examination and pathology reports can provide additional information. The patient can be in prone or supine position, the catheter insertions can be accomplished by free-hand technique or under template guidance [25], [28], [29], [30], [31]. At our institution we developed a technique, which is based on preimplant CT imaging and template-guided insertion followed by postimplant CT imaging and 3-D treatment planning dosimetry. This technique seems to be much more superior to the conventional fluoroscopic image-based planning. Compared to the results published previously by our group all dosimetric parameters have been improved by the CT-based planning (70% vs. 91% for V100, 0.35 vs. 0.33 for DNR, 0.40 vs. 0.68 for COIN) [9].

The placement of catheters and dose prescription we used for our patients are based on the principles of the Paris system [14]. But, instead of 85% of mean central dose we selected a lower isodose (mean 74%) for prescribing the dose in order to achieve acceptable target volume coverage and dose uniformity. Originally, the Paris system was developed for wire sources, which however can be imitated by using uniform dwell times in stepping source BT [32]. When dose optimization algorithm is applied the source dwell times in the catheters will not be uniform, and regarding dose prescription clinicians can depart from the rules of Paris system [33]. In papers about image-based interstitial brachytherapy the concept of MCD is hardly ever used, although the ICRU Report 58 recommends its use for reporting [20]. Therefore, we were not able to compare our results with published data. It also has to be noted that in modern image-based BT, having 3-D anatomical information, the dosimetry must rely on target and not on implant [6]. Probably this is the explanation for not relating the dose prescription to the MCD in image-guided interstitial brachytherapy.

Compared to external beam radiotherapy of breast the lung doses are much less using multi-catheter BT alone [24]. The dose to lung is highly related to the location of the resection cavity in the breast and its distance from the lung–chest wall interface. This is well demonstrated by the large range of lung volume receiving 5Gy (0–160cm3). Similar conclusions can be drawn for heart, but due to the anatomical location the doses to heart are much less.

In our correlation analysis we found high correlation between V100 and D90 (R2=0.867). D100 did not correlate with either V100 or D90. Due to the irregular shape of the target volume the D100, which is basically the minimum dose in PTV, does not represent a valuable dosimetric measure. Instead, D90 is however a volume-dependent parameter in which the small geometrical irregularities at the boundaries of the PTV are not taken into account, and this is the reason for its good correlation with target coverage (V100).

Since the TRAK showed a very good correlation with the size of the target volume the graph shown in Fig. 4 can be used for quality assurance (QA) purpose. A quick and simple way of checking the complex computer calculation is not so easy. Das et al. [25], [34] provided a recipe to predict the time needed to deliver a prescribed dose to a given prescription volume for an implant with known number of catheters. In their study they used a parameter similar to the TRAK to check the total time calculated by the computer plan as a part of a QA program.

Interestingly, the geometrical optimization followed by graphical optimization resulted in better target volume coverage in the treatment plans than the conformal planning (91% vs. 88% for V100). The volumes irradiated by 150% and 200% of the reference dose were significantly higher in conformal planning, and the consequence was the worse dose homogeneity, which is characterized with high DNR and low DHI values (Table 3). In dose point optimization algorithm the dose homogeneity is not taken into account, and this explains the occurrence of large volumes of high doses. However, the conformality of dose distributions was better when optimization on dose points placed on the surface of the PTV was used (COIN=0.77). The volume of normal tissue around the target receiving the reference dose is only 13% related to volume of PTV at conformal planning, and 32% in the treatment plans. Since the V100 values do not differ much, the less irradiated normal tissue volume resulted in the higher COIN value in the conformal plans. The TRAK values were practically equal in the treatment and conformal plans. This means that the total irradiation times were also equal, and only the distributions of individual dwell times in the catheters were different, which is the consequence of the different optimization algorithms.

There is no consensus on what degree of conformality is acceptable in interstitial BT and its measure is hardly ever reported. Pieters et al. [35] obtained a conformation number of 0.48 at a typical breast implant with geometrical optimization. van’t Riet et al. [36] found the average conformation number to be 0.72 for prostate implants, Kolotas et al. [29] reported COIN value of 0.76 for 42 breast implants, while Baltas et al. [19] aimed to achieve a COIN value above 0.64 by using their CT-based planning system. For our patients the COIN was 0.68 (range 0.51–0.82). But with the use of conformal planning this value has increased to 0.77 (0.55–0.88), which is almost the same as reported by Kolotas [29]. Conformality can depend considerably on the shape of the target volume. For a highly irregular PTV, which frequently occurs in clinical situations, perfect conformality cannot be expected, while for target volumes of regular shape the reference isodose surface can be tailored more easily to the PTV.

Conclusions 

In breast BT, preimplant CT-based target volume definition and implant simulation can be effectively used to obtain acceptable dose distributions regarding target volume coverage, dose homogeneity and conformality, and dose to critical structures. The use of graphical optimization can improve the dose distributions obtained by geometrical optimization. Dose point optimization performed on points placed on the surface of the PTV (conformal planning) can result in highly conformal dose distributions, but only at the cost of deterioration of dose homogeneity. Complete description of dose distributions with reporting the volumetric parameters and quality indices shown in this paper is recommended in order to establish their associations with treatment outcome and complications when clinical data are available. To decide the importance of and find the best compromise between the target coverage, conformality and dose homogeneity requires further clinical investigations with more patients and long follow-up.