Purpose: To introduce an indicator which can show the deformable image registration (DIR) accuracy, and be used for commissioning DIR systems. Methods: Inaccurate DIR can produce inaccurate composite dose when we add doses from multiple treatment plans. Deformation Error Histogram(DEH) and QA procedure were proposed to overcome the shortfall of landmark approach. Our method conducted a DVM comparison approach instead of using the significant visible feature points of the landmark approach. To show the feasibility of using DEH, we used a Lung and a Liver cancer patient. Each patient has two CT image sets (Reference and Test image sets). We generated a truth DVM by performing DIR using R and T sets, and by applying the DVM to T, R' is generated. To quantifies the systematic error of our in‐house DIR system, we calculated a test DVM by performing DIR using R' and T image sets. Then, we calculated a difference vector in millimeter between the test DVM and the truth DVM. Finally, a DEH was made from the difference vector for showing the cumulative distribution of error per Structure‐Of‐Interest (SOI). Results: Both clinical cases, we visualized errors by displaying the calculated difference vectors. Relatively larger systematic errors are observed where large deformation occurs, such as skin, mobile joints, shoulders, diaphragm, and inferior Lung region. The DEH can show the measured confidence range in DIR for the selected SOI. For Liver patient, [Average error, Standard deviation, 95% Confidence] for SOIs were GTV [6.69,2.07, 2.94], R Kidney [4.13,1.65,1.76], Spinal Cord [3.77,0.30,3.27], and Liver [3.03,1.67,0.78] in millimeter. For Lung patient, the Result were GTV [3.24,1.37,1.12], Heart [3.16,1.63,1.02], Lung [4.93,1.95,2.00], and Spinal Cord [4.39,1.67,2.39]. Conclusion: DEH provides quantitative systematic error during DIR. DEH and our data can be a tool for estimating magnitude of errors per anatomical sites and commissioning DIR systems.
All Science Journal Classification (ASJC) codes
- Radiology Nuclear Medicine and imaging