A Novel Clinical Radiomics Nomogram to Identify Crohn’s Disease from Intestinal Tuberculosis

Figures & data

Figure 1 Three-dimensional segmentation of the CD and ITB. (A and D) CD and ITB venous ileocecal lesions. (B and E) Layer by layer delineation of lesions. (C and F) 3D ROI of ileocecal lesions.

Table 1 Clinical Features of Patients with Crohn’s Disease and Intestinal Tuberculosis

Characteristics	CD (n=93)	ITB (n=67)	P value
Gender (%)			0.028
Male	67 (72.0)	37 (55.2)
Female	26 (28.0)	30 (44.8)
Age (years ± SD)	30.16±12.201	42.15±15.673	0.001
Abdominal pain (%)	61 (65.6)	47 (70.1)	0.544
Diarrhea (%)	52 (55.9)	23 (34.3)	0.007
Fever (%)	7 (7.5)	9 (13.4)	0.219
Weight loss (%)	36 (38.7)	21 (31.3)	0.337
Bloody stools (%)	12 (12.9)	11 (16.4)	0.532
Perianal lesion (%)	33 (35.5)	2 (3.0)	0.001
CRP (%)	68 (73.1)	37 (55.2)	0.019
ESR (%)	70 (75.3)	38 (56.7)	0.013
T-sport (%)	20 (21.5)	60 (89.6)	0.001
Comb sign (%)	56 (60.2)	23 (34.3)	0.001
Target sign (%)	16 (17.2)	3 (4.5)	0.014
Sinus tract, abscess (%)	9 (9.7)	3 (4.5)	0.218
Intestinal obstruction (%)	2 (2.2)	6 (9.0)	0.069
Stricture (%)	13 (14)	8 (11.9)	0.706
Lymphadenopathy (%)	30 (32.3)	27 (40.3)	0.295
Focal ileocecal lesion (%)	9 (9.7)	29 (43.3)	0.001
Small intestine segmental lesion (%)	63 (67.7)	21 (31.3)	0.001

Abbreviations: CRP, C-reactive protein; ESR, erythrocyte sedimentation rate.

Figure 2 The flow diagram of study.

Table 2 Radiomic Features Selection from the CTE in the Training Cohort

Feature Selected	Group	Coefficient
Wavelet-HHH_glszm_ZoneEntropy	Texture	0.8253
Wavelet-LLL_firstorder_Median	Intensity	1.0191
Log-sigma-1–0-mm-3D_glcm_Cluster Shade	Texture	0.8034
Wavelet-HHH_firstorder_Median	Intensity	0.6706
Log-sigma-4–0-mm-3D_glszm_Small Area Low Gray Level Emphasis	Texture	0.8357
Wavelet-LHL_firstorder_Median	Intensity	-1.0563
Wavelet-HHH_glszm_Low Gray Level Zone Emphasis	Texture	-0.2481
Original_shape_Elongation	Shape	-0.9146
Wavelet-LLL_glcm_Imc2	Texture	-0.8495

Table 3 Performance of the Clinical Model, Radiomics Model, and Clinical Model in the Training Cohort and Test Cohort

Different Models	Training Cohort (n=111)				Test Cohort (n=49)
	AUC (95% CI)	SEN	SPE	ACC	AUC (95% CI)	SEN	SPE	ACC
Clinical model	0.91(0.86–0.96)	0.462	1	0.685	0.90(0.82–0.99)	0.571	0.952	0.735
Radiomics model	0.90(0.84–0.95)	0.846	0.804	0.829	0.86(0.76–0.97)	0.786	0.762	0.776
Clinical radiomics model	0.96(0.86–0.96)	0.892	0.957	0.919	0.93(0.86–1.00)	0.893	0.905	0.898

Abbreviations: AUC, area under the curve; SEN, sensitivity; SPE, specificity; ACC, accuracy; 95% CI, 95% confidence interval.

Figure 3 The ROC curves of the three models. (A) Three models ROC curves in the training cohort. (B) Three models ROC curves in the validation cohort.

Figure 4 The decision curve analysis for three models. (A) Three models DCA in the training cohort. (B) Three models DCA curves in the validation cohort.

Figure 5 The clinical radiomics nomogram.