Fig. 2. The Integrated radiotherapy imaging system (IRIS) shown mounted on the linear accelerator with VisionRT camera pod superimposed in photograph. A commercially available 3D surface imaging system (AlignRT, VisionRT, London, UK) was also tested as an image-guidance Fig. 1. Setup, imaging, and analysis workflow. Fig. 3. (a) Example digitally reconstructed radiograph for X-ray clip registration. (b) Example daily X-ray image with marked clips. Fig. 4. (a) Sample surface image with highlighted region of interest. (b) Sample video texture image. (c) Surface generated from plan- ning computed tomography scan. Table 2. Target registration error comparison Abbreviations: TRE = target registration error; SurfRef-CT = computed tomography-based reference surface; SurfRef-fx 1 = refer- ence surface generated from first fraction; kV = kilovoltage. The TRE for surface-based setup using the CT-based reference surface was not significantly better than the chest Note, values for laser, chest wall, and kilovoltage clips include both gated and nongated data; with data for surface imaging (Surf- Ref-CT) given as gated only. Fig. 5. Color wash comparison of two surface images after alignment with AlignRT. Daily surface image was aligned tc reference surface from first fraction. Shading shows goodness of fit in region of interest. particularly for a site such as the breast. Chest wall alignment also was not expected to be as accurate as clip-based align- ment, because variations in arm position make it difficult to establish an exact correlation between the bony anatomy and the tumor position. Fig. 6. Target registration error results in box-plot format. SUR- FACE = results with use of first fraction reference surface (SurfRef-fx1); SURFACE-CT = results with computed tomography (CT)-based reference surface. Results shown for each imaging modality, except for surface imaging, include both gated and nongated data. wall alignment and was significantly worse than surface setup using the reference surface from the first fraction (Surf- Ref-fx1). Differences between CT-based reference surfaces and reference surfaces directly captured by 3D cameras were also noted by Bert et al. (19). These differences can be attributed to several factors. First, the patient was breath- ing lightly during the CT scan, which can cause surface be attributed to several factors. First, the patient was breath-
