Table 2.

The prognostic significance of the examined box-count features.^a,b

Dbin	Dout	SFR	Λ
Coronal section
0.51	0.74	0.04	0.02
0.42; 0.16–0.68	0.54; 0.29–0.79	0.30; 0.08–0.53	0.21; 0.01–0.40
Sagittal section
0.05	0.92	0.03	0.67
0.30; 0.07–0.49	0.49; 0.24–0.74	0.31; 0.08–0.47	0.45; 0.19–0.71
Transversal section
0.43	1.0	0.20	0.51
0.40; 0.15–0.65	0.50; 0.24–0.76	0.34; 0.11–0.57	0.58; 0.34–0.83
Average
0.08	0.95	0.02	0.32
0.30; 0.06–0.51	0.49; 0.23–0.75	0.20; 0–0.41	0.38; 0.14–0.62

ROC analysis was used for evaluation of predictive significance. Each result indicates a P-value followed by AUC and its 95% confidence interval.

^aAUC values and 95% CI are bootstrap corrected.

^bObtained with continuous data.

Dbin, fractal dimension obtained on binarized images; Dout, fractal dimension obtained on binarized and outlined images; SFR, ratio of the Dbins obtained on images in which tumor ROI was filled with foreground and background pixels; Λ, lacunarity.

Boldface values indicate significance with p < 0.05.