Deep Inspiration Breath Hold (DIBH) is a respiratory-gating technique employed in breast radiation therapy to lower the delivered radiation dose to the patient’s heart. When performing DIBH treatments, it is important to have a monitoring system to measure the quality of the patient's breath hold. In this retrospective study, we developed a system capable of monitoring DIBH breast treatments by using electronic portal imaging device (EPID) images acquired in each fraction. Three image-processing algorithms were evaluated on their ability to accurately measure the chest wall position (CWP) in EPID images. One of the algorithms is based on the use of the Canny filter while the other two analyze EPID profiles and identify the CWP as either the peak intensity or the inflection point at the lung-rib interface. The setup error and intrafraction motion were measured for all fractions of 20 left-sided breast patients, 10 treated with fields with no multi-leaf collimator (MLC) shielding and 10 with MLCs used for shielding. Setup error measurements were also collected for the first fractions of 20 additional patients. The algorithm showing the highest agreement (0.7 ± 0.5 mm) with manual measurements defined the CWP as the inflection point along the lung-rib interface on EPID profiles. This algorithm was then used to calculate intrafraction motion and setup errors. Intrafraction motion was found to be 0.7 mm on average for the group with no shielding and 0.9 mm on average for the shielded group. The largest intrafraction motion was recorded as +4.9 mm. Setup errors ranged from -7.1 mm to +5.1 mm for the unshielded group and from -4.3 mm to +8.9 mm for the shielded group. In the final group of patients where the setup errors of the first fractions were measured, these errors ranged from -5.2 mm to +7.5 mm. The most extreme motion and setup errors in all three groups of patients agreed within 2.1 mm of manual measurements. This algorithm agreement is smaller than the typical 3 mm action levels and future work could be conducted to evaluate its ability to automate tasks and/or replace surrogate-based monitoring systems.