Statistical analysis
Tumor size and volume measurements are summarized by mean and standard deviation, and compared between different measurement types using paired t-tests. The sample sizes needed to achieve 80% and 90% power for detecting differences in means are calculated based on the observed means, standard deviations, and correlation coefficients between measurement types. P values <.05 are considered statistically significant. All analyses are performed using R version 4.3.1 (R Core Team, 2023).
Tumor size
Pathologist vs. segmentation methods
Tumor size measurements are compared between pathologist and segmentation methods. Figure 1 below illustrates the distribution of tumor sizes for each measurement type, with individual measurements connected by lines to show variability within patients. Table 1 provides numerical and test results.
Power analysis
Based on the observed means and standard deviations from Table 1, and a calculated correlation coefficient of 0.52, we need \(n=\) 6 and 8 patients for a paired t-test to achieve 80% and 90% power, respectively.
Tumor volume
Segmentation vs. oblique vs. straight methods
Power analysis
Based on the observed means and standard deviations from Table 2, and calculated correlation coefficients of 0.9, 0.87, and 0.99 for the three pairs of measurement types, we need \(n=\) 9459 and 12662 patients for Segmentation vs. Oblique, \(n=\) 911370 and 1220066 patients for Segmentation vs. Straight, and \(n=\) 9913 and 13271 patients for Oblique vs. Straight to achieve 80% and 90% power, respectively.