Radiomics and nomogram of magnetic resonance imaging for preoperative prediction of microvascular invasion in small hepatocellular carcinoma

Table 1 Demographics, alpha fetal protein, bilirubin, tumor characteristics, and microvascular infiltration of the patients with small hepatocellular carcinoma in data sets of three hospitals

Characteristics	Hospital A (n = 221)	Hospital B (n = 94)	Hospital C (n = 100)
Age (yr)	mean ± SD: 51.2 ± 10.9 (range 29-78)	mean ± SD: 53.04 ± 10.59 (range 28-77)	mean ± SD: 54.01 ± 10.82 (range 28-85)
Male/female	189 (85.5%)/32 (14.5%)	84 (89.4%)/10 (10.6%)	85 (85.5%)/15 (15.0%)
Causes of liver disease
Hepatitis B	162 (73.3%)	71 (75.6%)	82 (82.0%)
Hepatitis C	21 (9.5%)	10 (10.6%)	6 (6.0%)
Alcoholic hepatitis	29 (13.1%)	8 (8.5%)	2 (2.0%)
Others	9 (4.1%)	5 (5.3%)	10 (10.0%)
Cirrhosis
Present	173 (78.3%)	69 (73.4%)	70 (70%)
Absent	48 (21.7%)	25 (26.6%)	30 (30%)
AFP (ng/mL)	Median: 26.10 (range 0.98-25451.00)	Median: 9.34 (range 1.09-6740.42)	Median: 31.070 (range 0.713- 4587.000)
TBiL (μmol/L)	Median: 12.5 (range 2.9-64.6)	Median: 17.4 (range 3.6-446.9)	Median: 13.2 (range 4.4-47.2)
DBiL (μmol/L)	Median: 3.9 (range 1.0-17.0)	Median: 4.5 (range 1.0-218.9)	Median: 5.50 (range 1.50-26.84)
MELD scores	mean ± SD: 13.9 ± 4.8 (range 7.0-26.9)	mean ± SD: 13.8 ± 5.8 (range 6.0-35.5)	mean ± SD: 11.8 ± 4.5 (range 5.2-25.6)
Child-Pugh classes
A	205 (93.2%)	81 (86.2%)	86 (86.0%)
B	13 (5.9%)	12 (12.8%)	13 (13.0%)
C	2 (0.9%)	1 (1.1%)	1 (1.0%)
Edmondson-steiner grade
Grade I	32 (14.5%)	12 (12.8%)	8 (8.0%)
Grade II	142 (64.3%)	68 (72.3%)	71 (71.0%)
Grade III	47 (21.2%)	14 (14.9%)	21 (21.0%)
Tumor size (cm)	mean ± SD: 2.04 ± 0.67 (range 0.60-3.00)	mean ± SD: 2.17 ± 0.42 (range 0.80-3.00)	mean ± SD: 2.20 ± 0.41 (range 0.90-3.00)
MVI
Positive	64 (28.9%)	22 (23.4%)	16 (16.0%)
Negative	157 (71.1%)	72 (76.6%)	84 (84.0%)

TBiL: Total bilirubin; DBiL: Direct bilirubin; MELD: Model for end-stage liver disease; AFP: Alpha-fetoprotein; MVI: Microvascular infiltration.

Table 2 Age, gender, alpha-fetoprotein and radiologic features of patients with small hepatocellular carcinoma and relationship with microvascular infiltration

	Hospital A (n = 221)			Hospital B (n = 94)			Hospital C (n = 100)
	Positive	Negative	P value	Positive	Negative	P value	Positive	Negative	P value
Age (yr)	50.9 ± 10.7	51.3 ± 10.9	0.840	51.7 ± 12.3	53.1 ± 10.1	0.597	53.5 ± 9.1	53.7 ± 11.0	0.958
Gender			0.593			0.443			0.259
Female	8	24		1	9		1	14
Male	56	133		21	63		15	70
AFP (ng/mL)	33.91 (range 1.40-25451.00)	22.52 (range 1-18929)	0.026	38.97 (range 1.90-6740.40)	5.43 (range 1.10-2018.79)	0.010	109.72 (range 0.80-2278.00)	24.41 (range 0.71-4587.00)	0.032
Size (cm)	2.34 ± 0.56	1.92 ± 0.67	0.036	2.43 ± 0.57	2.11 ± 0.66	0.047	2.47 ± 0.43	2.21 ± 0.64	0.054
Nonsmooth tumor margin			0.019			0.003			0.004
Absent	37	116		13	64		6	62
Present	27	41		9	8		10	22
Capsule			0.020			0.062			0.756
Absent	39	120		21	55		10	49
Present	25	37		25	17		6	35
Peritumoral hypointensity			0.002			0.304			0.007
Absent	47	141		17	63		70	8
Present	17	16		5	9		14	8

Size: Large tumor size; AFP: Alpha-fetoprotein; MVI: Microvascular infiltration.

Table 3 Diagnostic performance of alpha-fetoprotein and radiologic features for assessing microvascular infiltration of small hepatocellular carcinoma by receiver operating characteristic curve analysis

	AFP	Tumor size	Nonsmooth tumor margin	Incomplete capsule	Peritumoral hypointensity
Hospital A
AUC	0.597	0.675	0.580	0.577	0.582
95%CI	0.528-0.662	0.609-0.736	0.512-0.646	0.509-0.643	0.514-0.648
P value	0.024	< 0.001	0.024	0.027	0.007
Sensitivity	34.92	70.31	42.19	39.06	26.56
Specificity	81.82	61.15	73.89	76.43	89.81
Hospital B
AUC	0.683	0.639	0.649	0.595	0.551
95%CI	0.577-0.777	0.553-0.735	0.544-0.745	0.489-0.695	0.445-0.654
P value	0.006	0.035	0.008	0.005	0.304
Sensitivity	63.64	81.82	71.43	95.45	22.73
Specificity	72.06	44.44	88.89	23.61	87.50
Hospital C
AUC	0.669	0.576	0.682	0.521	0.667
95%CI	0.568-0.760	0.473-0.675	0.581-0.771	0.419-0.622	0.565-0.758
P value	0.016	0.213	0.007	0.759	0.014
Sensitivity	87.50	68.75	62.50	62.50	50.00
Specificity	52.38	54.76	73.81	41.67	83.33

DeLong test for comparison of receiver operating characteristic curve. AFP: Alpha-fetoprotein; AUC: Area under the curve; 95%CI: 95% confidence interval.

Table 4 Receiver operator characteristic curve analysis of radiomics scores with different sequences of magnetic resonance imaging for predict microvascular infiltration of small hepatocellular carcinoma

	T1WI	T2WI	DWI	AP	PVP	HBP
Training set
AUC	0.740	0.878	0.991	0.763	0.739	0.976
95%CI	0.661-0.808	0.814-0.926	0.958-0.999	0.695-0.823	0.661-0.807	0.940-0.991
P value	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
Sensitivity	77.08	82.98	95.74	71.43	56.25	89.83
Specificity	68.00	82.00	96.04	79.37	78.22	99.27
Testing set
AUC	0.776	0.813	0.971	0.788	0.790	0.979
95%CI	0.611-0.895	0.651-0.922	0.858-0.999	0.642-0.894	0.630-0.904	0.911-1.000
P value	0.0004	< 0.001	< 0.001	0.0014	0.0001	< 0.001
Sensitivity	91.67	58.33	100.00	64.29	84.62	100.00
Specificity	57.69	92.00	84.62	93.75	73.08	91.43
Validation Hospital B
AUC	0.834	0.825	0.816	0.810	0.847	0.970
95%CI	0.742-0.904	0.732-0.896	0.678-0.876	0.715-0.884	0.758-0.913	0.912-0.994
P value	< 0.001	< 0.001	0.0002	< 0.001	< 0.001	< 0.001
Sensitivity	63.64	95.45	71.43	57.14	54.55	95.45
Specificity	94.29	57.75	88.89	88.89	98.61	98.57
Validation Hospital C
AUC	0.766	0.761	0.801	0.824	0.833	0.803
95%CI	0.672-0.844	0.669-0.839	0.710-0.871	0.737-0.892	0.748-0.898	0.680-0.834
P value	0.0001	0.0003	< 0.001	< 0.001	< 0.001	0.007
Sensitivity	80.00	60.00	90.00	86.67	85.71	83.33
Specificity	70.45	89.01	68.89	68.54	72.83	77.67

T1WI: T1 weighted imaging; T2WI: T2 weighted imaging; DWI: Diffusion weighted imaging; AP: Arterial phase; PVP: Portal vein phase; HBP: Hepatobiliary phase; AUC: Area under the curve; 95%CI: 95% confidence interval.

Table 5 Predictive performance of the nomogram prediction model for estimating the risk of microvascular infiltration presence in patients with small hepatocellular carcinoma

	Training set	Testing set	Validation Hospital B	Validation Hospital C
T1WI
C-index	0.771	0.846	0.895	0.830
95%CI	0.695-0.836	0.594-0.883	0.775-0.925	0.667-0.850
P value	< 0.001	0.0014	< 0.001	0.0001
T2WI
C-index	0.895	0.917	0.886	0.808
95%CI	0.834-0.940	0.640-0.915	0.746-0.906	0.654-0.867
P value	< 0.001	0.0001	< 0.001	0.0015
DWI
C-index	0.990	0.970	0.843	0.869
95%CI	0.957-0.999	0.843-0.997	0.685-0.881	0.694-0.899
P value	< 0.001	< 0.001	0.0001	< 0.001
AP
C-index	0.774	0.794	0.886	0.874
95%CI	0.706-0.833	0.615-0.876	0.695-0.899	0.674-0.884
P value	< 0.001	0.0025	< 0.001	< 0.001
PVP
C-index	0.746	0.831	0.918	0.870
95%CI	0.668-0.814	0.650-0.916	0.791-0.934	0.732-0.887
P value	< 0.001	< 0.001	< 0.001	< 0.001
HBP
C-index	0.990	0.971	0.912	0.808
95%CI	0.944-0.993	0.892-0.999	0.918-0.996	0.635-0.892
P value	< 0.001	< 0.001	< 0.001	0.0081

T1WI: T1 weighted imaging; T2WI: T2 weighted imaging; DWI: Diffusion weighted imaging; AP: Arterial phase; PVP: Portal vein phase; HBP: Hepatobiliary phase; 95%CI, 95% confidence interval; C-index: Concordance index.