Artificial intelligence-assisted endoscopists improve the detection rate of high-risk gastric lesions: A propensity score-matched retrospective study

doi:10.3748/wjg.v32.i21.117299

Advanced Search

BPG is committed to discovery and dissemination of knowledge

Home / Archive / Volume 32, Issue 21

This Article

(5)

(7)

(0)

(1)

(2)

(131)

Peer-Review Report of This Article

CrossCheck and Google Search of This Article

Academic Rules and Norms of This Article

Citation of this article

Corresponding Author of This Article

Research Domain of This Article

Article-Type of This Article

Open-Access Policy of This Article

Times Cited Counts in Google of This Article

Journal Information of This Article

Publication Name

World Journal of Gastroenterology

ISSN

1007-9327

Publisher of This Article

Baishideng Publishing Group Inc, 7041 Koll Center Parkway, Suite 160, Pleasanton, CA 94566, USA

Retrospective Study

Copyright: ©Author(s) 2026. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution-NonCommercial (CC BY-NC 4.0) license. No commercial re-use. See permissions. Published by Baishideng Publishing Group Inc.

World J Gastroenterol. Jun 7, 2026; 32(21): 117299
Published online Jun 7, 2026. doi: 10.3748/wjg.v32.i21.117299

Table 1 Baseline characteristics and diagnostic outcomes of patients before and after propensity score matching, n (%)

Variable	All patients			P value²	SMD	Propensity score-matched patients			P value²	SMD
Variable	Overall (n = 40474)¹	Non-AI group (n = 29942)	AI group (n = 10532)	P value²	SMD	Overall (n = 15528)¹	Non-AI group (n = 7764)	AI group (n = 7764)	P value²	SMD
Gender				0.046	0.023	-			0.987	< 0.001
Male	19519 (48.23)	14528 (48.52)	4991 (47.39)			7481 (48.18)	3741 (48.18)	3740 (48.17)
Female	20955 (51.77)	15414 (51.48)	5541 (52.61)			8047 (51.82)	4023 (51.82)	4024 (51.83)
Age	54 (44, 61)	54 (45, 61)	53(43, 60)	< 0.001	0.012	53 (43, 61)	53 (43, 61)	53 (43, 60)	0.006	0.054
Biopsy number	2 (1, 2)	2 (1, 2)	2 (1, 2)	< 0.001	0.028	2 (1, 2)	2 (1, 2)	2 (1, 2)	< 0.001	0.082
Anesthesia assistance				< 0.001	0.331	-			0.009	0.042
No	10874 (26.87)	9125 (30.48)	1749 (16.61)			2332 (15.02)	1224 (15.77)	1108 (14.27)
Yes	29600 (73.13)	20817 (69.52)	8783 (83.39)			13196 (84.98)	6540 (84.23)	6656 (85.73)
LGIN				0.040	0.031	-			0.014	0.040
No	40322 (99.62)	29845 (99.68)	10477 (99.48)			15455 (99.53)	7738 (99.67)	7717 (99.39)
Yes	152 (0.38)	97 (0.32)	55 (0.52)			73 (0.47)	26 (0.33)	47 (0.61)
HGIN				0.300	0.012	-			0.022	0.037
No	40408 (99.84)	29897 (99.85)	10511 (99.80)			15505 (99.85)	7758 (99.92)	7747 (99.78)
Yes	66 (0.16)	45 (0.15)	21 (0.20)			23 (0.15)	6 (0.08)	17 (0.22)
EGC				0.008	0.028	-			0.004	0.047
No	40370 (99.74)	29877 (99.78)	10493 (99.63)			15485 (99.72)	7752 (99.85)	7733 (99.60)
Yes	104 (0.26)	65 (0.22)	39 (0.37)			43 (0.28)	12 (0.15)	31 (0.40)
HrGLs				0.002	0.033	-			0.001	0.055
No	40219 (99.37)	29775 (99.44)	10444 (99.16)			15418 (99.29)	7727 (99.52)	7691 (99.06)
Yes	255 (0.63)	167 (0.56)	88 (0.84)			110 (0.71)	37 (0.48)	73 (0.94)
Endoscopist seniority				< 0.001	0.362	-			> 0.9	0
Junior (< 5 years)	6580 (16.26)	5704 (19.05)	876 (8.32)			1752 (11.28)	876 (11.28)	876 (11.28)
Medium (5-10 years)	13265 (32.78)	10135 (33.85)	3130 (29.72)			6260 (40.31)	3130 (40.31)	3130 (40.31)
Senior (> 10 years)	20629 (50.97)	14103 (47.10)	6526 (61.96)			7516 (48.40)	3758 (48.40)	3758 (48.40)

¹Median (interquartile range).

²Pearson’s χ² d test; Wilcoxon rank sum test; Fisher’s exact test.

SMD: Standardized mean difference; AI: Artificial intelligence; LGIN: Low-grade intraepithelial neoplasia; HGIN: High-grade intraepithelial neoplasia; EGC: Early gastric cancer; HrGLs: High-risk gastric lesions.

Full Size Table

Table 2 The impact of artificial intelligence assistance on the detection rate of high-risk gastric lesions in propensity score matching groups, n (%)

Lesion types	Non-AI group (n = 7764)	AI group (n = 7764)	OR (95%CI)	P value
LGIN	26 (0.33)	47 (0.61)	1.81 (1.12-2.93)	0.015
HGIN	6 (0.08)	17 (0.22)	2.84 (1.12-7.20)	0.028
EGC	12 (0.15)	31 (0.40)	2.59 (1.33-5.05)	0.005
HrGLs	37 (0.48)	73 (0.94)	1.98 (1.33-2.95)	0.001

OR: Odds ratio; CI: Confidence interval; LGIN: Low-grade intraepithelial neoplasia; HGIN: High-grade intraepithelial neoplasia; EGC: Early gastric cancer; HrGLs: High-risk gastric lesions.

Full Size Table

Table 3 The impact of artificial intelligence assistance on the detection of high-risk gastric lesions by endoscopists of different levels of experience in propensity score matching groups

Endoscopist seniority	OR (95%CI)	P value
Senior (> 10 years)
LGIN	1.96 (1.12-3.41)	0.018
HGIN	3.67 (1.02-13.18)	0.046
EGC	3.35 (1.34-8.34)	0.010
HrGLs	2.23 (1.37-3.61)	0.001
Medium (5-10 years)
LGIN	1.17 (0.39-3.48)	0.782
HGIN	3.00 (0.61-14.89)	0.178
EGC	2.20 (0.77-6.35)	0.143
HrGLs	1.55 (0.72-3.31)	0.260
Junior (< 5 years)
LGIN	3.01 (0.31-28.96)	0.341
HGIN	2.00 (0.18-22.1)	0.992
EGC	1.33 (0.30-5.97)	0.992
HrGLs	1.50 (0.25-9.01)	0.656

OR: Odds ratio; CI: Confidence interval; LGIN: Low-grade intraepithelial neoplasia; HGIN: High-grade intraepithelial neoplasia; EGC: Early gastric cancer; HrGLs: High-risk gastric lesions.

Full Size Table

Table 4 The impact of artificial intelligence assistance on the detection rate of high-risk gastric lesions in anesthetized patients of propensity score matching groups, n (%)

Lesion types	Non-AI group (n = 6540)	AI group (n = 6656)	OR (95%CI)	P value
LGIN	19 (0.29)	42 (0.63)	2.18 (1.27-3.75)	0.005
HGIN	5 (0.08)	14 (0.21)	2.76 (0.99-7.65)	0.052
EGC	10 (0.15)	22 (0.33)	2.17 (1.03-4.58)	0.043
HrGLs	29 (0.44)	60 (0.90)	2.04 (1.31-3.19)	0.002

OR: Odds ratio; CI: Confidence interval; AI: Artificial intelligence; LGIN: Low-grade intraepithelial neoplasia; HGIN: High-grade intraepithelial neoplasia; EGC: Early gastric cancer; HrGLs: High-risk gastric lesions.

Full Size Table

Table 5 Influence of artificial intelligence assistance on detection rate of high-risk gastric lesions in different parts of stomach in propensity score matching groups, n (%)

Location of lesion	AI group (n = 7764)	Non-AI group (n = 7764)	OR (95%CI)	P value
Cardia	11 (0.14)	6 (0.08)	1.84 (0.68-4.96)	0.232
Gastric fundus	3 (0.04)	1 (0.01)	3.00 (0.31-28.85)	0.341
Gastric body	11 (0.14)	5 (0.06)	2.20 (0.77-6.34)	0.144
Gastric angle	11 (0.14)	3 (0.04)	3.67 (1.02-13.16)	0.046
Gastric antrum	35 (0.45)	23 (0.30)	1.52 (0.90-2.58)	0.117
Gastric pylorus	3 (0.04)	3 (0.04)	1.00 (0.20-4.96)	1.000

OR: Odds ratio; CI: Confidence interval; AI: Artificial intelligence.

Full Size Table

Table 6 Logistic regression analysis for detection rate of high-risk gastric lesions in propensity score matching groups

Variable	Univariate regression analysis			Multivariate regression analysis
Variable	OR	95%CI	P value	OR	95%CI	P value
Gender
Male	-	-	-	-	-	-
Female	0.35	0.23-0.53	< 0.001	0.44	0.29-0.67	< 0.001
Age	1.08	1.06-1.09	< 0.001	1.06	1.04-1.08	< 0.001
Biopsy number	1.89	1.71-2.08	< 0.001	1.67	1.51-1.85	< 0.001
AI assistance
No	-	-	-	-	-	-
Yes	1.98	1.33-2.95	0.001	1.96	1.30-2.94	0.001
Endoscopist seniority
Junior	-	-	-	-	-	-
Medium	1.57	0.61-4.07	0.354	1.51	0.58-3.93	0.402
Senior	3.62	1.46-8.95	0.005	3.02	1.21-7.54	0.018
Anesthesia assistance
No	-	-	-	-	-	-
Yes	0.75	0.46-1.21	0.232	-	-	-

OR: Odds ratio; CI: Confidence interval.

Full Size Table

Table 7 Stratified analysis by biopsy number for detection rate of high-risk gastric lesions in propensity score matching groups

Biopsy number	AI group (n = 7764)	Non-AI group (n = 7764)	OR (95%CI)	P value
1	0.36%	0.07%	5.53 (1.46-35.99)	0.027
2	0.45%	0.22%	2.08 (0.82-5.93)	0.138
≥ 3	2.39%	1.51%	1.60 (1.02-2.57)	0.145

OR: Odds ratio; CI: Confidence interval; AI: Artificial intelligence.

Full Size Table

Table 8 Mediation analysis evaluating biopsy number as a potential mediator between artificial intelligence assistance and high-risk gastric lesions detection

Parameter	Unadjusted model	Adjusted model¹
Sample size	15528	15528
a-path: AI assistance → biopsy number	0.093 (SE = 0.018, P < 0.001)	0.107 (SE = 0.017, P < 0.001)
b-path: Biopsy number → HrGLs	0.605 (SE = 0.047, P < 0.001)	0.480 (SE = 0.050, P < 0.001)
Total effect²	0.0039 (0.0015-0.0063)	0.0044 (0.0020-0.0069)
Direct effect²	0.0034 (0.0012-0.0058)	0.0039 (0.0017-0.0064)
Indirect effect²	0.0002 (0.0001-0.0004)	0.0002 (0.0001-0.0003)
Proportion mediated	6.3%	4.9%

¹The adjusted model controlled for age, sex, anesthesia use, and endoscopist experience.

²Total, direct, and indirect effects are reported as risk differences (with 95% confidence intervals in parentheses) on the probability scale (0-1).

a-path coefficients (reported as mean difference ± standard error) represent the average increase in biopsy count in artificial intelligence-assisted vs non-artificial intelligence-assisted procedures. b-path coefficients (reported as log-odds ± standard error) represent the increase in log-odds of high-risk gastric lesions detection per additional biopsy. AI: Artificial intelligence; HrGLs: High-risk gastric lesions; SE: Standard error.

Full Size Table

Citation: Ying JX, Yan SY, Fu XY, Zhou YJ, Zhou JJ, Yang Y, Zhou XB, Wang ZZ, Li SW, Fang LN, Mao XL. Artificial intelligence-assisted endoscopists improve the detection rate of high-risk gastric lesions: A propensity score-matched retrospective study. World J Gastroenterol 2026; 32(21): 117299
URL: https://www.wjgnet.com/1007-9327/full/v32/i21/117299.htm
DOI: https://dx.doi.org/10.3748/wjg.v32.i21.117299

Ying JX, Yan SY, Fu XY, Zhou YJ, Zhou JJ, Yang Y, Zhou XB, Wang ZZ, Li SW, Fang LN, Mao XL. Artificial intelligence-assisted endoscopists improve the detection rate of high-risk gastric lesions: A propensity score-matched retrospective study. World J Gastroenterol 2026; 32(21): 117299 [DOI: 10.3748/wjg.v32.i21.117299]

All content on this site: Copyright © 1993-2026 Baishideng Publishing Group Inc, its licensors, and contributors. All rights are reserved, including those for text and data mining, AI training, and similar technologies. For all open access content, the relevant licensing terms apply.