TABLE 1.

Algorithm chest wall data compared to manual measurements for EPID images

Patient	Canny algorithm (cm)	Peak‐finding algorithm (cm)	Inflection‐point algorithm (cm)	Manual measurement (cm)
P1	2.48± 0.04	2.88± 0.07	2.58± 0.09	2.61± 0.07
P2	2.30± 0.03	2.70± 0.04	2.45± 0.04	2.54± 0.03
P3	2.01± 0.02	2.35± 0.17	2.11± 0.16	2.19± 0.06
P4	1.62± 0.04	1.91± 0.04	1.72± 0.05	1.74± 0.04
P5	1.81± 0.34	2.30± 0.07	1.96± 0.10	2.14± 0.06
P6	2.30± 0.07	2.65± 0.05	2.45± 0.06	2.48± 0.08
P7	1.32± 0.06	1.67± 0.06	1.47± 0.08	1.50± 0.09
P8	2.35± 0.59	2.74± 0.07	2.50± 0.06	2.56± 0.08
P9	1.91± 0.33	2.30± 0.08	2.06± 0.06	2.12± 0.04
P10	1.96± 0.06	2.30± 0.04	2.06± 0.11	2.17± 0.06

Chest wall measurements from the three different image‐processing algorithms compared to manual measurements for the 10‐patient dataset. The first algorithm used the Canny filter to outline the chest wall, while the other two used either the peak position within the breast or the inflection point leading up to the peak as the chest wall position. For each algorithm, the median chest wall position within the region of interest was compared to the median of ten manual measurements within the same region. Uncertainties for both the algorithm and manual measurements are given by the standard deviation of chest wall measurements within the region of interest.