Comparison of 2D-2D Versus 3D-3D Matching for Image Guided Setup of Head and Neck Cancer Patients

Purpose/Objective(s)Cone-beam CT (CBCT) allows direct imaging of the tumor volume to reposition patients with head and neck (H&N) tumors prior to treatment at the expense of increased patient dose and accelerator time. kV imaging of bony landmarks is faster at significantly lower patient dose. This study compared the two techniques to determine if kV imaging could be a suitable substitute for CBCT in patients with H&N cancers.Materials/MethodsTen patients with H&N cancers were immobilized using a thermoplastic mask and treated using IMRT on a Varian Trilogy™ linear accelerator using daily image guidance. Each day, prior to treatment and with the patient in the treatment position, orthogonal kV radiographs were obtained and compared to reference DRRs. Bony landmark-based 2D-2D matching was performed to determine translational table (patient) shifts along the A-P, L-R, and S-I axes. After physician review and approval, the shifts were performed and the patient was treated. Once or twice per week a CBCT was obtained prior to treatment. Depending on time constraints, the CBCT was either performed in addition to the orthogonal radiographs or in place of them. Each CBCT was compared to a reference CT obtained at the time of simulation and 3D-3D matching was performed. In contradistinction to the 2D-2D matching, the 3D-3D matching used soft tissue in and around the tumor volume. On the days that CBCT was performed, the table (patient) was shifted using the 3D-3D match after physician review and approval. Comparison of the table shifts determined using either 2D-2D or 3D-3D matching was made and the statistical significance evaluated. PTV margins were calculated using the formula PTV = 2.5Σ + 0.7σ where Σ and σ are the systematic and random errors as defined by the Netherlands Cancer Institute.ResultsAn average of 26.4 ± 7.8 2D-2D and 5.7 ± 1.2 3D-3D matches were made per patient. Average corrections are (L-R, A-P, S-I) (-0.11 ± .43, 0.00 ± .21, 0.01 ± .31) and (-0.03 ± .41, 0.12 ± .29, 0.04 ± .29) cm, respectively. The required PTV margins (in the absence of corrective table shifts) are (0.66, 0.33, 0.51) and (0.84, 0.55, 0.52) cm using 2D-2D or 3D-3D matching respectively. For a given direction, the 2D-2D systematic and random errors were approximately equal. Errors in the S-I direction were similar for 2D-2D and 3D-3D. In the A-P and L-R directions, the random errors computed from 3D-3D matching were similar to those obtained from the 2D-2D matching, but the 3D-3D matching derived systematic errors were 0.2 cm greater.ConclusionsThe slightly larger systematic errors seen with 3D-3D matching using CBCT imaging of the tumor may be due to tumor changes over time not seen using bony landmarks or difficulty in delineating the tumor volume. Detailed review of each patient, especially on days in which both 2D-2D and 3D-3D matching was performed indicates that kV imaging using bony landmarks may be an acceptable substitute for CBCT imaging of patients with H&N cancers where immobilization is quite good. An imaging schedule in which CBCT is performed once or twice per week to check for gradual changes in the tumor volume, shape or position with 2D-2D matching of KV images on the other days may be sufficient. This may not be necessarily the case for other sites. Purpose/Objective(s)Cone-beam CT (CBCT) allows direct imaging of the tumor volume to reposition patients with head and neck (H&N) tumors prior to treatment at the expense of increased patient dose and accelerator time. kV imaging of bony landmarks is faster at significantly lower patient dose. This study compared the two techniques to determine if kV imaging could be a suitable substitute for CBCT in patients with H&N cancers. Cone-beam CT (CBCT) allows direct imaging of the tumor volume to reposition patients with head and neck (H&N) tumors prior to treatment at the expense of increased patient dose and accelerator time. kV imaging of bony landmarks is faster at significantly lower patient dose. This study compared the two techniques to determine if kV imaging could be a suitable substitute for CBCT in patients with H&N cancers. Materials/MethodsTen patients with H&N cancers were immobilized using a thermoplastic mask and treated using IMRT on a Varian Trilogy™ linear accelerator using daily image guidance. Each day, prior to treatment and with the patient in the treatment position, orthogonal kV radiographs were obtained and compared to reference DRRs. Bony landmark-based 2D-2D matching was performed to determine translational table (patient) shifts along the A-P, L-R, and S-I axes. After physician review and approval, the shifts were performed and the patient was treated. Once or twice per week a CBCT was obtained prior to treatment. Depending on time constraints, the CBCT was either performed in addition to the orthogonal radiographs or in place of them. Each CBCT was compared to a reference CT obtained at the time of simulation and 3D-3D matching was performed. In contradistinction to the 2D-2D matching, the 3D-3D matching used soft tissue in and around the tumor volume. On the days that CBCT was performed, the table (patient) was shifted using the 3D-3D match after physician review and approval. Comparison of the table shifts determined using either 2D-2D or 3D-3D matching was made and the statistical significance evaluated. PTV margins were calculated using the formula PTV = 2.5Σ + 0.7σ where Σ and σ are the systematic and random errors as defined by the Netherlands Cancer Institute. Ten patients with H&N cancers were immobilized using a thermoplastic mask and treated using IMRT on a Varian Trilogy™ linear accelerator using daily image guidance. Each day, prior to treatment and with the patient in the treatment position, orthogonal kV radiographs were obtained and compared to reference DRRs. Bony landmark-based 2D-2D matching was performed to determine translational table (patient) shifts along the A-P, L-R, and S-I axes. After physician review and approval, the shifts were performed and the patient was treated. Once or twice per week a CBCT was obtained prior to treatment. Depending on time constraints, the CBCT was either performed in addition to the orthogonal radiographs or in place of them. Each CBCT was compared to a reference CT obtained at the time of simulation and 3D-3D matching was performed. In contradistinction to the 2D-2D matching, the 3D-3D matching used soft tissue in and around the tumor volume. On the days that CBCT was performed, the table (patient) was shifted using the 3D-3D match after physician review and approval. Comparison of the table shifts determined using either 2D-2D or 3D-3D matching was made and the statistical significance evaluated. PTV margins were calculated using the formula PTV = 2.5Σ + 0.7σ where Σ and σ are the systematic and random errors as defined by the Netherlands Cancer Institute. ResultsAn average of 26.4 ± 7.8 2D-2D and 5.7 ± 1.2 3D-3D matches were made per patient. Average corrections are (L-R, A-P, S-I) (-0.11 ± .43, 0.00 ± .21, 0.01 ± .31) and (-0.03 ± .41, 0.12 ± .29, 0.04 ± .29) cm, respectively. The required PTV margins (in the absence of corrective table shifts) are (0.66, 0.33, 0.51) and (0.84, 0.55, 0.52) cm using 2D-2D or 3D-3D matching respectively. For a given direction, the 2D-2D systematic and random errors were approximately equal. Errors in the S-I direction were similar for 2D-2D and 3D-3D. In the A-P and L-R directions, the random errors computed from 3D-3D matching were similar to those obtained from the 2D-2D matching, but the 3D-3D matching derived systematic errors were 0.2 cm greater. An average of 26.4 ± 7.8 2D-2D and 5.7 ± 1.2 3D-3D matches were made per patient. Average corrections are (L-R, A-P, S-I) (-0.11 ± .43, 0.00 ± .21, 0.01 ± .31) and (-0.03 ± .41, 0.12 ± .29, 0.04 ± .29) cm, respectively. The required PTV margins (in the absence of corrective table shifts) are (0.66, 0.33, 0.51) and (0.84, 0.55, 0.52) cm using 2D-2D or 3D-3D matching respectively. For a given direction, the 2D-2D systematic and random errors were approximately equal. Errors in the S-I direction were similar for 2D-2D and 3D-3D. In the A-P and L-R directions, the random errors computed from 3D-3D matching were similar to those obtained from the 2D-2D matching, but the 3D-3D matching derived systematic errors were 0.2 cm greater. ConclusionsThe slightly larger systematic errors seen with 3D-3D matching using CBCT imaging of the tumor may be due to tumor changes over time not seen using bony landmarks or difficulty in delineating the tumor volume. Detailed review of each patient, especially on days in which both 2D-2D and 3D-3D matching was performed indicates that kV imaging using bony landmarks may be an acceptable substitute for CBCT imaging of patients with H&N cancers where immobilization is quite good. An imaging schedule in which CBCT is performed once or twice per week to check for gradual changes in the tumor volume, shape or position with 2D-2D matching of KV images on the other days may be sufficient. This may not be necessarily the case for other sites. The slightly larger systematic errors seen with 3D-3D matching using CBCT imaging of the tumor may be due to tumor changes over time not seen using bony landmarks or difficulty in delineating the tumor volume. Detailed review of each patient, especially on days in which both 2D-2D and 3D-3D matching was performed indicates that kV imaging using bony landmarks may be an acceptable substitute for CBCT imaging of patients with H&N cancers where immobilization is quite good. An imaging schedule in which CBCT is performed once or twice per week to check for gradual changes in the tumor volume, shape or position with 2D-2D matching of KV images on the other days may be sufficient. This may not be necessarily the case for other sites.