Table 3.

Results of elastic-net regression and linear modeling of the top texture features and their most closely associated clinical predictor.

Texture feature	Elastic net RMSE	Top predictor	Linear model coefficient	Linear model P-value	FDR P-value correction
0 GLNU	1.96 × 103	WBC	6.74 × 101	0.047	0.083
90 GLNU	1.92 × 103	WBC	8.15 × 101	0.022	0.049
0 RLN	2.63 × 103	AST	− 1.16 × 101	0.009	0.035
45 RLN	3.75 × 103	AST	− 1.65 × 101	0.006	0.029
90 RLN	3.25 × 103	AST	− 1.45 × 101	0.006	0.029
135 RLN	3.82 × 103	AST	− 1.49 × 101	0.019	0.049
135 cluster tendency	6.76 × 10–1	AST	− 3.02 × 10–3	0.003	0.020
Kurtosis	5.47	AST	2.71 × 10–2	0.001	0.011
LBM 0	4.40 × 10–3	WBC	1.28 × 10–4	0.095	0.126
LBM 1	2.10 × 10–3	AST	5.60 × 10–6	0.083	0.120
LBM 2	1.50 × 10–3	WBC	3.67 × 10–5	0.147	0.161
LBM 13	2.00 × 10–4	Albumin	− 1.14 × 10–4	0.012	0.040
LBM 18	1.00 × 10–4	Platelet count	− 4.62 × 10–6	0.226	0.226
LBM 20	3.00 × 10–4	Platelet count	− 1.48 × 10–5	0.102	0.126
LBM 21	6.00 × 10–4	WBC	3.52 × 10–5	0.110	0.126
LBM 22	1.30 × 10–3	WBC	1.28 × 10–5	0.215	0.225
LBM 23	2.10 × 10–3	AST	6.13 × 10–6	0.047	0.083
LBM 24	4.83 × 10–2	WBC	− 1.24 × 10–3	0.105	0.126
LBM 25	3.14 × 10–2	AST	8.04 × 10–5	0.072	0.111
Mean deviation	3.09 × 10–2	WBC	9.81 × 10–4	0.072	0.111
45 short run emphasis	6.94 × 10–2	AST	− 4.28 × 101	< 0.001	0.049
Size zone variability	3.95 × 102	Blood urea nitrogen	2.11 × 101	0.034	0.071
Inverse variance	5.87 × 102	Cirrhosis	2.58 × 102	0.021	0.049

Texture features with corrected p-values meeting statistical significance are bolded.

RMSE root mean square error, FDR false discovery rate, GLNU gray level non-uniformity, RLN run length non-uniformity, LBM local binary pattern matrix, AST aspartate aminotransferase, ALT alanine aminotransferase, WBC white blood cell count.