Purpose: To simulate phase-based 4DCT tumor size and motion measurement errors.
Methods: We conducted one-dimensional 4DCT simulations of imaged tumor size and motion. Tumor motion was modeled using patient breathing traces, obtained in the clinic during 5DCT acquisition for 98 patients. Breathing traces spanned >5-minute time periods and were normalized to the 5th to 85th percentile amplitudes. Commercial 4DCT protocols were simulated. For each simulation, the detector was incrementally moved simultaneous to the breathing motion (tumor) and CT detectors overlaps recorded and analyzed as a function of breathing phase, defined between successive inhalation peaks. Overlaps at each slice of >50% of imaged times were considered to have visualized the tumor at those slices and phases. The imaged tumor sizes (first to last imaged slice) at each phase, as well as the midpoint imaged tumor motion phase-to-phase were analyzed and compared to the input tumor size and motion to determine the fraction of time for each patient that artifacts incorrectly measured each quantity.
Results: We calculated the mean imaged tumor width and motion across all simulations and all patients and compared these values to the set ground truth values. The mean tumor motion and tumor size were found to reproduce the original to within 1% and 10%, respectively (the tumor size reflects the decision to use 5th to 95th percentile range). Specific attention was paid to outliers. The worst 10% over- and under-estimations of motion and size for the worst 30% of patients were 0.70 and 1.26 and 0.90 and 1.60 (inhalation phase), respectively.
Conclusion: 4DCT was able to measure the tumor motion and size in most patients and even for irregular patients most of the time. However, there were unpredictable outliers that caused large motion and tumor position errors. Further studies will examine tumor image discontinuities.