A Nomogram to Predict Benign/Malignant Mediastinal Lymph Nodes Based on EBUS Sonographic Features

Table 6

Ultrasound image features for predicting the malignancy outcomes of lymph nodes deemed negative following EBUS-TBNA in the training and validation sets.

Characteristics	Total (n = 317)	Training set (n = 224)	Validation set (n = 93)	value
Long-axis (median [IQR])	1.590 [1.300, 2.100]	1.600 [1.295, 2.185]	1.500 [1.300, 2.000]	0.395
Number of passes per node (median [IQR])	3.000 [2.000, 3.000]	3.000 [2.000, 3.000]	3.000 [2.000, 3.000]	0.184
Short-axis (%)				0.597
>1 cm	203 (64.0%)	146 (65.2%)	57 (61.3%)
≤1 cm	114 (36.0%)	78 (34.8%)	36 (38.7%)
Long-axis/short-axis ratio (%)				1.000
≥1.5	77 (24.3%)	54 (24.1%)	23 (24.7%)
<1.5	240 (75.7%)	170 (75.9%)	70 (75.3%)
Echo intensity (%)				0.033
Hyperechoic	11 (3.5%)	4 (1.8%)	7 (7.5%)
Hypoechoic	298 (94.0%)	215 (96.0%)	83 (89.3%)
Isoechoic	8 (2.5%)	5 (2.2%)	3 (3.2%)
Echogenicity (%)				0.720
Heterogeneous	123 (38.8%)	85 (38.0%)	38 (40.9%)
Homogeneous	194 (61.2%)	139 (62.0%)	55 (59.1%)
Margin (%)				0.498
Indistinct	264 (83.3%)	184 (82.1%)	80 (86.0%)
Distinct	53 (16.7%)	40 (17.9%)	13 (14.0%)
Blood flow (%)				1.000
Rich	243 (76.7%)	172 (76.8%)	71 (76.3%)
Lacking	74 (23.3%)	52 (23.2%)	22 (23.7%)
Fusion (%)				0.616
Absent	256 (80.8%)	183 (81.7%)	73 (78.5%)
Present	61 (19.2%)	41 (18.3%)	20 (21.5%)
Central hilar structure (CHS) (%)				0.442
Absent	131 (41.3%)	89 (39.7%)	42 (45.2%)
Present	186 (58.7%)	135 (60.3%)	51 (54.8%)