Do maximum intensity projection images truly capture tumor motion?

PURPOSE: For the treatment of patients with lung cancer, internal target volume frequently is determined by using maximum intensity projection (MIP) images generated by means of four-dimensional computed tomography (4D-CT). To check the accuracy of MIPs for various target motions, especially for targets moving irregularly, we performed phantom studies using a programmable dynamic lung phantom. METHODS AND MATERIALS: A custom-built programmable lung phantom was used to simulate irregular target motions along the superior-inferior direction. After scanning in helical mode using 4D-CT, reconstructed phase and MIP images were imported into the Pinnacle 8.0 treatment planning system for image analysis.
RESULTS: For all regular periodic target motions with constant amplitude and period, the measured MIP target span along the superior-inferior direction was accurate within 2-3 mm of the real target motion span. For irregular target motions with varying amplitudes and periods, the measured MIP target span systematically underrepresented the real range of target motion by more than 1 cm in some cases. The difference between measured MIP target span and real target span decreased as the target moved faster. We associate these discrepancies with the fact that current reconstruction algorithms of commercial 4D-CT are based on phase binning.
CONCLUSIONS: According to our phantom measurements, MIP accurately reflects the range of target motion for regular target motion. However, it generally underestimates the range of target motion when the motion is irregular in amplitude and periodicity. Clinical internal target volume determination using MIP requires caution, especially when there is breathing irregularity.