Sensitivity and specificity of time resolved point dose measurements for patient specific quality assurance

This study aimed to quantify the sensitivity and specificity of time-resolved point dose measurements. Criteria were defined to assess whether errors would cause a clinically relevant dose deviation during patient treatment. The sensitivity and specificity were determined based on verification measurements of five error-free plans and 84 intentional error plans. Receiver operator characteristic analysis was conducted to quantify the efficiency of the method. In addition, the specificity of the method was investigated in more detail by assessing its ability to identify different error modes. For measurements made at planning target volume locations, a moderate sensitivity (65%  ±  13%), specificity (76%  ±  12%), and an area under the curve (AUC) equal to 0.77 were obtained for a quality control (QC) acceptance criterion of 2%. Measurements made at organ at risk (OAR) locations had high sensitivity (80%  ±  20%), but low specificity (54%  ±  13%), and an AUC equal to 0.70. The low specificity for OAR locations could be traced to the impact of a small couch tilt on measurement locations at larger distances from the isocentre, resulting in increased shielding by multi-leaf collimator (MLC) leaves. Further analysis showed that output errors and errors affecting the penumbra region can be resolved on a per measurement basis with moderate to high sensitivity (100% and 67% for errors in output, and in the penumbra region, respectively) and high specificity (77% and 85% for errors in output, and in the penumbra region, respectively). This can potentially result in saving time investigating failing QC measurements.