TU‐E‐204B‐07: Real‐Time 3D Target Position Estimation Using a Single KV Imager Combined with an External Respiratory Monitor during Arc and Static Beam Delivery

B. Cho, P. Poulsen, D. Ruan, A. Sawant, P. Keall

Research output: Contribution to journalArticle

Abstract

Purpose: To experimentally investigate a real‐time 3D target position estimation method, using a single kV imager combined with an external respiratory monitor, integrated with a DMLC tracking. Method and Materials: The experimental DMLC tracking system employed a Varian Trilogy with kV imager and Respiratory Position Management (RPM). The internal/external correlation between 3D target position T(t) and RPM signal R(t) was modeled by a state‐augmented linear model: T(t)=a*R(t)+b*R(t‐ τ)+c. The model parameters were determined by solving least‐squares estimation of the error Σ∥pi ‐ P(θi)T(ti)∥2, where pi was the projected marker positions in kV images and P(θi) was the projection operator at the gantry angle θi. During arc and static 5‐field beam delivery, DMLC tracking was performed with ten patient lung tumor motion traces. A 3D motion stage with a gold marker and a separate 1D motion stage with an RPM marker‐block were used to reproduce the tumor motion and external respiratory signals, respectively. The gold‐marker positions were measured by 1‐Hz kV imaging, while the external marker positions were measure by RPM system at 30Hz. To initialize the correlation model rotational kV images over 120° of angular span were acquired. The estimated 3D marker positions were sent to the DMLC to reposition the beam. The tracking accuracies were quantified as “beam‐target” alignment mismatch in cine MV images acquired during the experiments. Results: With tracking, the average errors of the ten lung tumor traces were 0.6–1.4mm and 0.8–1.5mm during the arc and static 5‐field beam delivery, respectively. Without tracking, the average errors were 3.7–6.1mm and 3.0–7.3mm. Conclusion: The DMLC tracking system integrated with a new estimation method shows typical accuracy of 1mm for patient tumor motion traces during arc and static delivery, and thus has the potential for accurate motion management of thoracic and abdominal tumors. Conflict of Interest: Supported by NTH/NCI R0193626.

Original languageEnglish (US)
Pages (from-to)3402-3403
Number of pages2
JournalMedical Physics
Volume37
Issue number6
DOIs
StatePublished - 2010

Fingerprint

Neoplasms
Conflict of Interest
Lung
Gold
Linear Models
Thorax

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging

Cite this

TU‐E‐204B‐07 : Real‐Time 3D Target Position Estimation Using a Single KV Imager Combined with an External Respiratory Monitor during Arc and Static Beam Delivery. / Cho, B.; Poulsen, P.; Ruan, D.; Sawant, A.; Keall, P.

In: Medical Physics, Vol. 37, No. 6, 2010, p. 3402-3403.

Research output: Contribution to journalArticle

@article{1eec4365bfb3492daa451d500d7e84b2,
title = "TU‐E‐204B‐07: Real‐Time 3D Target Position Estimation Using a Single KV Imager Combined with an External Respiratory Monitor during Arc and Static Beam Delivery",
abstract = "Purpose: To experimentally investigate a real‐time 3D target position estimation method, using a single kV imager combined with an external respiratory monitor, integrated with a DMLC tracking. Method and Materials: The experimental DMLC tracking system employed a Varian Trilogy with kV imager and Respiratory Position Management (RPM). The internal/external correlation between 3D target position T(t) and RPM signal R(t) was modeled by a state‐augmented linear model: T(t)=a*R(t)+b*R(t‐ τ)+c. The model parameters were determined by solving least‐squares estimation of the error Σ∥pi ‐ P(θi)T(ti)∥2, where pi was the projected marker positions in kV images and P(θi) was the projection operator at the gantry angle θi. During arc and static 5‐field beam delivery, DMLC tracking was performed with ten patient lung tumor motion traces. A 3D motion stage with a gold marker and a separate 1D motion stage with an RPM marker‐block were used to reproduce the tumor motion and external respiratory signals, respectively. The gold‐marker positions were measured by 1‐Hz kV imaging, while the external marker positions were measure by RPM system at 30Hz. To initialize the correlation model rotational kV images over 120° of angular span were acquired. The estimated 3D marker positions were sent to the DMLC to reposition the beam. The tracking accuracies were quantified as “beam‐target” alignment mismatch in cine MV images acquired during the experiments. Results: With tracking, the average errors of the ten lung tumor traces were 0.6–1.4mm and 0.8–1.5mm during the arc and static 5‐field beam delivery, respectively. Without tracking, the average errors were 3.7–6.1mm and 3.0–7.3mm. Conclusion: The DMLC tracking system integrated with a new estimation method shows typical accuracy of 1mm for patient tumor motion traces during arc and static delivery, and thus has the potential for accurate motion management of thoracic and abdominal tumors. Conflict of Interest: Supported by NTH/NCI R0193626.",
author = "B. Cho and P. Poulsen and D. Ruan and A. Sawant and P. Keall",
year = "2010",
doi = "10.1118/1.3469298",
language = "English (US)",
volume = "37",
pages = "3402--3403",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "AAPM - American Association of Physicists in Medicine",
number = "6",

}

TY - JOUR

T1 - TU‐E‐204B‐07

T2 - Real‐Time 3D Target Position Estimation Using a Single KV Imager Combined with an External Respiratory Monitor during Arc and Static Beam Delivery

AU - Cho, B.

AU - Poulsen, P.

AU - Ruan, D.

AU - Sawant, A.

AU - Keall, P.

PY - 2010

Y1 - 2010

N2 - Purpose: To experimentally investigate a real‐time 3D target position estimation method, using a single kV imager combined with an external respiratory monitor, integrated with a DMLC tracking. Method and Materials: The experimental DMLC tracking system employed a Varian Trilogy with kV imager and Respiratory Position Management (RPM). The internal/external correlation between 3D target position T(t) and RPM signal R(t) was modeled by a state‐augmented linear model: T(t)=a*R(t)+b*R(t‐ τ)+c. The model parameters were determined by solving least‐squares estimation of the error Σ∥pi ‐ P(θi)T(ti)∥2, where pi was the projected marker positions in kV images and P(θi) was the projection operator at the gantry angle θi. During arc and static 5‐field beam delivery, DMLC tracking was performed with ten patient lung tumor motion traces. A 3D motion stage with a gold marker and a separate 1D motion stage with an RPM marker‐block were used to reproduce the tumor motion and external respiratory signals, respectively. The gold‐marker positions were measured by 1‐Hz kV imaging, while the external marker positions were measure by RPM system at 30Hz. To initialize the correlation model rotational kV images over 120° of angular span were acquired. The estimated 3D marker positions were sent to the DMLC to reposition the beam. The tracking accuracies were quantified as “beam‐target” alignment mismatch in cine MV images acquired during the experiments. Results: With tracking, the average errors of the ten lung tumor traces were 0.6–1.4mm and 0.8–1.5mm during the arc and static 5‐field beam delivery, respectively. Without tracking, the average errors were 3.7–6.1mm and 3.0–7.3mm. Conclusion: The DMLC tracking system integrated with a new estimation method shows typical accuracy of 1mm for patient tumor motion traces during arc and static delivery, and thus has the potential for accurate motion management of thoracic and abdominal tumors. Conflict of Interest: Supported by NTH/NCI R0193626.

AB - Purpose: To experimentally investigate a real‐time 3D target position estimation method, using a single kV imager combined with an external respiratory monitor, integrated with a DMLC tracking. Method and Materials: The experimental DMLC tracking system employed a Varian Trilogy with kV imager and Respiratory Position Management (RPM). The internal/external correlation between 3D target position T(t) and RPM signal R(t) was modeled by a state‐augmented linear model: T(t)=a*R(t)+b*R(t‐ τ)+c. The model parameters were determined by solving least‐squares estimation of the error Σ∥pi ‐ P(θi)T(ti)∥2, where pi was the projected marker positions in kV images and P(θi) was the projection operator at the gantry angle θi. During arc and static 5‐field beam delivery, DMLC tracking was performed with ten patient lung tumor motion traces. A 3D motion stage with a gold marker and a separate 1D motion stage with an RPM marker‐block were used to reproduce the tumor motion and external respiratory signals, respectively. The gold‐marker positions were measured by 1‐Hz kV imaging, while the external marker positions were measure by RPM system at 30Hz. To initialize the correlation model rotational kV images over 120° of angular span were acquired. The estimated 3D marker positions were sent to the DMLC to reposition the beam. The tracking accuracies were quantified as “beam‐target” alignment mismatch in cine MV images acquired during the experiments. Results: With tracking, the average errors of the ten lung tumor traces were 0.6–1.4mm and 0.8–1.5mm during the arc and static 5‐field beam delivery, respectively. Without tracking, the average errors were 3.7–6.1mm and 3.0–7.3mm. Conclusion: The DMLC tracking system integrated with a new estimation method shows typical accuracy of 1mm for patient tumor motion traces during arc and static delivery, and thus has the potential for accurate motion management of thoracic and abdominal tumors. Conflict of Interest: Supported by NTH/NCI R0193626.

UR - http://www.scopus.com/inward/record.url?scp=85024782742&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85024782742&partnerID=8YFLogxK

U2 - 10.1118/1.3469298

DO - 10.1118/1.3469298

M3 - Article

AN - SCOPUS:85024782742

VL - 37

SP - 3402

EP - 3403

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 6

ER -