Man or machine? Prospective comparison of the version 2018 EASL, LI-RADS criteria and a radiomics model to diagnose hepatocellular carcinoma

Jiang, Hanyu; Liu, Xijiao; Chen, Jie; Wei, Yi; Lee, Jeong Min; Cao, Likun; Wu, Yuanan; Duan, Ting; Li, Xin; Ma, Ling; Song, Bin

doi:10.1186/s40644-019-0266-9

Table 3 Diagnostic performances of the radiomics model, EASL and LI-RADS criteria

From: Man or machine? Prospective comparison of the version 2018 EASL, LI-RADS criteria and a radiomics model to diagnose hepatocellular carcinoma

All lesions	EASL^a	LI-RADS^a			Radiomics model^b	p value^‡
All lesions	EASL^a	L5/5V	L4/4V/5/5 V	p value^¶	Radiomics model^b	EASL vs Rad^d	EASL vs L5/5V^¶	Rad vs L5/5V^d
All nodules, n = 229
Sen	91% (95%CI: 0.85–0.95)	86% (95%CI: 0.80–0.91)	94% (95%CI: 0.89–0.97)	0.019	73% (95%CI: 0.61–0.83)	0.006	0.531	0.076
Spe	71% (95%CI: 0.58–0.83)	82% (95%CI: 0.70–0.91)	73% (95%CI: 0.60–0.84)	0.254	77% (95%CI: 0.55–0.92)	1.000	0.528	1.000
PPV	91% (95%CI: 0.87–0.94)	94% (95%CI: 0.89–0.86)	92% (95%CI: 0.87–0.94)	0.999	91% (95%CI: 0.8–0.97)	1.000	1.000	1.000
NPV	71% (95%CI: 0.60–0.80)	66% (95%CI: 0.56–0.74)	79% (95%CI: 0.67–0.87)	0.101	47% (95%CI: 0.3–0.65)	0.056	1.000	0.199
AUC	0.811 (95%CI: 0.754–0.859)	0.841 (95%CI: 0.787–0.886)	0.834 (95%CI: 0.780–0.880)	0.745	0.810 (95%CI: 0.690–0.931)	1.000	0.761	1.000
Cirrhotic liver, n = 136
Sen	92% (95%CI: 0.85–0.96)	86% (95%CI: 0.78–0.92)	93% (95%CI: 0.86–0.97)	0.082	70% (95%CI: 0.55–0.83)	0.007	0.408	0.112
Spe	63% (95%CI: 0.41–0.81)	79% (95%CI: 0.58–0.93)	71% (95%CI: 0.49–0.87)	0.503	62% (95%CI: 0.32–0.86)	1.000	0.588	0.797
PPV	92% (95%CI: 0.87–0.95)	95% (95%CI: 0.90–0.98)	94% (95%CI: 0.89–0.97)	0.999	87% (95%CI: 0.72–0.96)	1.000	1.000	1.000
NPV	63% (95%CI: 0.45–0.77)	54% (95%CI: 0.42–0.66)	68% (95%CI: 0.51–0.81)	0.275	36% (95%CI: 0.17–0.59)	0.199	1.000	0.530
AUC	0.772 (95%CI: 0.693–0.840)	0.824 (95%CI: 0.750–0.884)	0.818 (95%CI: 0.743–0.879)	0.849	0.715 (95%CI: 0.524–0.906)	1.000	0.612	0.897
Non-cirrhotic liver, n = 93
Sen	89% (95%CI: 0.78–0.95)	87% (95%CI: 0.76–0.94)	95% (95%CI: 0.86–0.99)	0.668	78% (95%CI: 0.56–0.93)	0.843	1.000	1.000
Spe	78% (95%CI: 0.60–0.91)	84% (95%CI: 0.67–0.95)	75% (95%CI: 0.57–0.89)	0.275	100% (95%CI: 0.66–1)	0.008	1.000	0.045
PPV	89% (95%CI: 0.80–0.94)	91% (95%CI: 0.82–0.96)	88% (95%CI: 0.80–0.93)	0.994	100% (95%CI: 0.81–1)	1.000	1.000	1.000
NPV	78% (95%CI: 0.63–0.88)	77% (95%CI: 0.63–0.87)	89% (95%CI: 0.72–0.96)	0.503	64% (95%CI: 0.35–0.87)	1.000	1.000	1.000
AUC	0.833 95%CI: 0.742–0.903)	0.856 (95%CI: 0.768–0.920)	0.850 (95%CI: 0.761–0.916)	0.852	0.923 (95%CI: 0.829–1)	0.631	1.000	0.981
≤2 cm, n = 34
Sen	86% (95%CI: 0.64–0.97)	76% (95%CI: 0.53–0.92)	95% (95%CI: 0.76–1)	0.067	38% (95%CI: 0.09–0.76)	0.030	1.000	0.141
Spe	69% (95%CI: 0.39–0.91)	77% (95%CI: 0.46–0.95)	62% (95%CI: 0.32–0.86)	0.389	88% (95%CI: 0.47–1)	0.876	0.606	1.000
PPV	82% (95%CI: 0.66–0.91)	84% (95%CI: 0.66–0.94)	80% (95%CI: 0.67–0.89)	0.992	75% (95%CI: 0.19–0.99)	1.000	1.000	1.000
NPV	75% (95%CI: 0.50–0.90)	67% (95%CI: 0.47–0.82)	89% (95%CI: 0.53–0.98)	0.166	58% (95%CI: 0.28–0.85)	1.000	1.000	1.000
AUC	0.775 (95%CI: 0.599–0.900)	0.766 (95%CI: 0.589–0.893)	0.784 (95%CI: 0.610–0.906)	0.788	0.859 (95%CI: 0.658–1)	1.000	1.000	1.000
>2 cm, ≤5 cm, n = 83
Sen	85% (95%CI: 0.74–0.93)	80% (95% CI: 0.68–0.89)	95% (95% CI: 0.86–0.99)	0.011	68% (95%CI: 0.45–0.86)	0.354	1.000	0.829
Spe	82% (95%CI: 0.60–0.95)	91% (95% CI: 0.71–0.99)	82% (95% CI: 0.60–0.95)	0.375	82% (95%CI: 0.48–0.98)	1.000	1.000	1.000
PPV	93% (95%CI: 0.84–0.97)	96% (95%CI: 0.87–0.99)	94% (95% CI: 0.86–0.97)	0.997	88% (95%CI: 0.64–0.99)	1.000	1.000	1.000
NPV	67% (95%CI: 0.51–0.79)	63% (95%CI: 0.50–0.74)	86% (95%CI: 0.66–0.95)	0.043	56% (95%CI: 0.30–0.80)	1.0001	1.000	1.000
AUC	0.835 (95%CI: 0.738–0.908)	0.856 (95%CI: 0.762–0.924)	0.885 (95%CI: 0.796–0.944)	0.466	0.806 (95%CI: 0.617–0.994)	1.000	1.000	1.000
>5 cm, n = 112
Sen	96% (95%CI: 0.89–0.99)	92% (95% CI: 0.85–0.97)	92% (95% CI: 0.85–0.97)	1.000	83% (95%CI: 0.67–0.93)	0.063	1.000	0.514
Spe	62% (95%CI: 0.38–0.82)	76% (95% CI: 0.53–0.92)	71% (95% CI: 0.48–0.89)	0.725	67% (95%CI: 0.09–0.99)	1.000	0.933	1.000
PPV	92% (95%CI: 0.86–0.95)	94% (95%CI: 0.89–0.97)	93% (95%CI: 0.88–0.97)	0.999	93% (95%CI: 0.81–0.99)	1.000	1.000	1.000
NPV	76% (95%CI: 0.54–0.90)	70% (95%CI: 0.52–0.83)	68% (95%CI: 0.50–0.82)	0.920	22% (95%CI: 0.09–0.45)	1.000	1.000	1.000
AUC	0.788 (95%CI: 0.700–0.859)	0.842 (95%CI: 0.762–0.904)	0.819 (95%CI: 0.735–0.885)	0.317	0.746 (95%CI: 0.590–0.866)	0.303	0.915	0.627

Abbreviations: EASL European Association for the Study of the Liver, LI-RADS Liver Imaging Reporting and Data System, L4 LR-4, 4 V LR-TIV in the presence of LR-4 lesions, L5 LR-5, 5 V LR-TIV in the presence of LR-5 lesions, Rad radiomics model, Sen sensitivity, Spe specificity, PPV positive predictive value, NPV negative predictive value, AUC area under the receiver operating characteristic curve
^aDiagnostic measures were evaluated in the combined cohort comprising all patients
^bDiagnostic measures were evaluated in the validation cohort
^dComparisons were made in the validation cohort
‡P values were corrected with the Bonferroni method
¶Comparisons were made in the combined cohort comprising all patients

Back to article page

ISSN: 1470-7330

Contact us

Submission enquiries: alyzza.javate@springernature.com
General enquiries: info@biomedcentral.com

Cancer Imaging

Contact us