TY - JOUR
T1 - Artificial intelligence–enabled histology exhibits comparable accuracy to pathologists in assessing histological remission in ulcerative colitis
T2 - a systematic review, meta-analysis, and meta-regression
AU - Puga-Tejada, Miguel
AU - Majumder, Snehali
AU - Maeda, Yasuharu
AU - Zammarchi, Irene
AU - Ditonno, Ilaria
AU - Santacroce, Giovanni
AU - Capobianco, Ivan
AU - Robles-Medranda, Carlos
AU - Ghosh, Subrata
AU - Iacucci, Marietta
N1 - Publisher Copyright:
© The Author(s) 2025. Published by Oxford University Press on behalf of European Crohn’s and Colitis Organisation.
PY - 2025/1/1
Y1 - 2025/1/1
N2 - Background and Aims: Achieving histological remission is a desirable emerging treatment target in ulcerative colitis (UC), yet its assessment is challenging due to high inter- and intraobserver variability, reliance on experts, and lack of standardization. Artificial intelligence (AI) holds promise in addressing these issues. This systematic review, meta-analysis, and meta-regression evaluated the AI’s performance in assessing histological remission and compared it with that of pathologists. Methods: We searched Medline/PubMed and Scopus databases from inception to September 2024. We included studies on AI models assessing histological activity in UC, with or without comparison to pathologists. Pooled performance metrics were calculated: sensitivity, specificity, positive and negative predictive value (PPV and NPV), observed agreement, and F1 score. A pairwise meta-analysis compared AI and pathologists, while sub-meta-analysis and meta-regression evaluated heterogeneity and factors influencing AI performance. Results: Twelve studies met the inclusion criteria. AI models exhibited strong performance with a pooled sensitivity of 0.84 (95% CI, 0.80–0.88), specificity 0.87 (0.84–0.91), PPV 0.90 (0.87–0.92), NPV 0.80 (0.71–0.88), observed agreement 0.85 (0.82–0.89), and F1 score 0.85 (0.82–0.89). AI models demonstrated no significant differences with pathologists for specificity, observed agreement, and F1 score, while they were outperformed by pathologists for sensitivity and NPV. AI models for the adult population were linked to reduced heterogeneity and enhanced AI performance at meta-regression. Conclusions: AI shows significant potential for assessing histological remission in UC and performs comparably to pathologists. Future research should focus on standardized, large-scale studies to minimize heterogeneity and support widespread AI implementation in clinical practice.
AB - Background and Aims: Achieving histological remission is a desirable emerging treatment target in ulcerative colitis (UC), yet its assessment is challenging due to high inter- and intraobserver variability, reliance on experts, and lack of standardization. Artificial intelligence (AI) holds promise in addressing these issues. This systematic review, meta-analysis, and meta-regression evaluated the AI’s performance in assessing histological remission and compared it with that of pathologists. Methods: We searched Medline/PubMed and Scopus databases from inception to September 2024. We included studies on AI models assessing histological activity in UC, with or without comparison to pathologists. Pooled performance metrics were calculated: sensitivity, specificity, positive and negative predictive value (PPV and NPV), observed agreement, and F1 score. A pairwise meta-analysis compared AI and pathologists, while sub-meta-analysis and meta-regression evaluated heterogeneity and factors influencing AI performance. Results: Twelve studies met the inclusion criteria. AI models exhibited strong performance with a pooled sensitivity of 0.84 (95% CI, 0.80–0.88), specificity 0.87 (0.84–0.91), PPV 0.90 (0.87–0.92), NPV 0.80 (0.71–0.88), observed agreement 0.85 (0.82–0.89), and F1 score 0.85 (0.82–0.89). AI models demonstrated no significant differences with pathologists for specificity, observed agreement, and F1 score, while they were outperformed by pathologists for sensitivity and NPV. AI models for the adult population were linked to reduced heterogeneity and enhanced AI performance at meta-regression. Conclusions: AI shows significant potential for assessing histological remission in UC and performs comparably to pathologists. Future research should focus on standardized, large-scale studies to minimize heterogeneity and support widespread AI implementation in clinical practice.
KW - artificial intelligence
KW - histological assessment
KW - ulcerative colitis
UR - https://www.scopus.com/pages/publications/85215106479
U2 - 10.1093/ecco-jcc/jjae198
DO - 10.1093/ecco-jcc/jjae198
M3 - Article
C2 - 39742395
AN - SCOPUS:85215106479
SN - 1873-9946
VL - 19
JO - Journal of Crohn's and Colitis
JF - Journal of Crohn's and Colitis
IS - 1
M1 - jjae198
ER -