Poster — Wed Eve—46: Quantitative Evaluation on the Accuracy of Image Registration Methods in SPECT Guided Radiation Therapy for Lung Cancer Patients

• Published in: Medical Physics Vol. 36; no. 9; p. 4312
• Main Authors: Yin, L, Tang, L, Hamarneh, G, Moiseenko, V, Celler, A, Shcherbinin, S, Fua, T‐F, Thompson, A, Liu, M, Duzenli, C, Gill, B, Sheehan, F, Powe, J, Worsley, D
• Format: Conference Proceeding Journal Article
• Language: English
• Published: American Association of Physicists in Medicine 01-09-2009
• Subjects: Cancer; Computed tomography; Image registration; Image scanners; Lungs; Medical imaging; Medical treatment planning; Radiation therapy; Single photon emission computed tomography
• ISSN: 0094-2405; 2473-4209
• DOI: 10.1118/1.3244150

Purpose:
To quantitatively evaluate the accuracy of several SPECT/CT image registration methods in recent studies and its impact on the functional lung volume segmentation in SPECT guided radiation therapy (RT) treatment planning.
Methods and Materials:
Five lung cancer patients were consented to have a perfusion SPECT scan with 99mTc‐macroaggregated albumin. During the scan, a low resolution CT image was acquired using the SPECT/CT scanner. This CT scan was co‐registered to the patient's planning CT scan through four rigid and deformable image registration programs (rigid registration, skin/lung control points based registration and B‐spline deformable registration). After the CT to CT co‐registration, original SPECT reconstructions were warped and co‐registered to the planning CT scan. The functional lung volumes were segmented from each deformed SPECT using 10, 20, …, 90% of maximum pixel value as a threshold. The differences in the size and contours of each functional volume were calculated.
Results:
Based on the evaluation of registered CT images, the result from B‐spline registration demonstrated the smallest intensity difference. Using the warped SPECT images obtained from this registration method as a reference, the smallest difference in the size and contour of functional volumes was found using rigid registration. In the point‐based registrations, a better result was found when the control points were placed on lung volume instead of body contour.
Conclusion:
Apply B‐spline based image registration method in SPECT‐guided RT studies was shown to be accurate. Point‐based image registration using skin markers with a standalone SPECT scanner was found least accurate.