Feasibility and effectiveness of automatic deep learning network and radiomics models for differentiating tumor stroma ratio in pancreatic ductal adenocarcinoma

Liao, Hongfan; Yuan, Jiang; Liu, Chunhua; Zhang, Jiao; Yang, Yaying; Liang, Hongwei; Jiang, Song; Chen, Shanxiong; Li, Yongmei; Liu, Yanbing

doi:10.1186/s13244-023-01553-z

Insights into Imaging

Table 1 Baseline characteristics in the training and test cohorts

From: Feasibility and effectiveness of automatic deep learning network and radiomics models for differentiating tumor stroma ratio in pancreatic ductal adenocarcinoma

Characteristics	Training cohort (n = 160)			External test cohort (n = 47)			P (Inter)
Characteristics	TSR-low (n = 72)	TSR-high (n = 88)	P (Intra)	TSR-low (n = 20)	TSR-high (n = 27)	P (Intra)	P (Inter)
Clinical characteristics
Age (years), mean ± SD	60.40 ± 9.45	60.55 ± 9.66	0.925	59.48 ± 9.04	59.85 ± 9.25	0.892	0.784
Gender			0.720			0.726	0.895
Female	24 (33.3)	27 (30.7)		5 (25.0)	8 (29.6)
Male	48 (66.7)	61 (69.3)		15 (75.0)	19 (70.4)
Abdominal pain			0.106			0.770	0.592
Yes	37 (51.4)	34 (38.6)		9 (45.0)	11 (40.7)
No	35 (48.6)	54 (61.4)		11 (55.0)	16 (59.3)
Pancreatitis history			0.302			0.640	0.611
No	55 (76.4)	73 (83.0)		16 (80.0)	23 (85.2)
Yes	17 (23.6)	15 (17.0)		4 (20.0)	4 (14.8)
Jaundice			0.806			0.380	0.457
No	56 (77.8)	67 (76.1)		15 ((75.0)	23 (85.2)
Yes	16 (22.2)	21 (23.9)		5 (25.0)	4 (14.8)
Pathological characteristics
T stage			0.048*			0.095	0.301
T1–2	61 (84.7)	63 (71.6)		17 (85.0)	17 (63.0)
T3–4	11 (15.3)	25 (28.4)		3 (15.0)	10 (37.0)
Histological grade			0.829			0.013*	0.052
Low-grade	47 (65.3)	56 (63.6)		14 (70.0)	9 (33.3)
High-grade	25 (34.7)	32 (36.4)		6 (30.0)	18 (66.7)
Lymph node metastasis			0.423			0.638	0.413
Negative	43 (59.7)	47 (53.4)		12 (60.0)	18 (66.7)
Positive	29 (40.3)	41 (46.6)		8 (40.0)	9 (33.3)
Duodenum invasion			0.112			0.333	0.599
Negative	41 (56.9)	39 (44.3)		9 (45.0)	16 (59.3)
Positive	31 (43.1)	49 (55.7)		11 (55.0)	11 (40.7)
Imaging characteristics
CT-reported tumor size	26.65 ± 11.31	28.57 ± 10.43	0.270	24.78 ± 17.31	29.95 ± 8.78	0.228	0.772
Location			0.173			0.905	0.850
Head and neck	50 (69.4)	73 (83.0)		13 (65.0)	18 (66.7)
Body and tail	22 (30.6)	15 (17.0)		7 (35.0)	9 (33.3)
Parenchymal atrophy			0.946			0.642	0.368
No	27 (38.0)	33 (37.5)		9 (45.0)	14 (51.9)
Yes	44 (62.0)	55 (62.5)		11 (55.0)	13 (48.1)
PD dilatation			0.493			0.635	0.859
No	18 (25.0)	18 (20.5)		4 (20.0)	7 (25.9)
Yes	54 (75.0)	70 (79.5)		16 (80.0)	20 (74.1)
CBD dilatation			0.607			0.381	0.914
No	24 (33.3)	26 (29.5)		5 (25.0)	10 (37.0)
Yes	48 (66.7)	62 (70.5)		15 (75.0)	17 (63.0)
Biochemical characteristics
CA-199 level			0.167			0.260	0.774
Normal	27 (18.1)	24 (35.2)		2 (10.0)	11 (40.7)
Abnormal	45 (81.9)	64 (64.8)		18 (90.0)	16 (59.3)
CEA level			0.626			0.404	0.378
Normal	61 (84.7)	72 (81.8)		17 (85.0)	25 (92.6)
Abnormal	11 (15.3)	26 (18.2)		3 (15.0)	2 (7.4)
TBIL level			0.274			0.689	0.699
Normal	34 (47.2)	34 (38.6)		7 (35.0)	11 (40.7)
Abnormal	38 (52.8)	54 (61.4)		13 (65.0)	16 (59.3)

p(Intra) represents the result of univariable analyses between TSR-low and TSR-high groups, p(Inter) represents the significant difference between training and test groups
PD pancreatic duct, CBD common bile duct, CA-199 carbohydrate antigen 199, CEA carcino-embryonic antigen, TBIL total bilirubin
^*Represents p < 0.05

Back to article page