文献类型：期刊文献

英文题名：Accuracy of Artificial Intelligence in Detecting Tumor Bone Metastasis: A Systematic Review and Meta-Analysis

作者：Tao, Huimin[1,5]; Hui, Xu[2,3,4]; Zhang, Zhihong[1]; Fu, Zhenjiang[1]; Wang, Ping[5]; Zhou, Shen[5]; Yang, Kehu[2,3,4]

第一作者：Tao, Huimin

机构：[1] The First Clinical Medical College, Gansu University of Chinese Medicine, Gansu, Lanzhou, 730000, China; [2] Evidence-Based Medicine Centre, School of Basic Medical Science, Lanzhou University, Lanzhou, 730000, China; [3] Centre for Evidence-Based Social Science, Center for Health Technology Assessment, School of Public Health, Lanzhou University, Lanzhou, 730000, China; [4] Gansu Key Laboratory of Evidence-Based Medicine, Lanzhou University, Lanzhou, 730000, China; [5] Department of Radiology, Gansu Provincial Hospital, Gansu, Lanzhou, 730000, China

第一机构：甘肃中医药大学

年份：2024

外文期刊名：SSRN

收录：EI(收录号：20240183696)

语种：英文

外文关键词：Artificial intelligence - Computerized tomography - Database systems - Diagnosis - Grading - Hierarchical systems - Pathology - Quality control - Risk assessment - Tumors

摘要：Background: In recent years, artificial intelligence (AI) technology has emerged as a promising adjunctive tool for radiologists in detecting Bone metastasis (BM). Objective: To explore the diagnostic performance of AI in detecting BM. Methods and Analysis: Two reviewers conducted a comprehensive search in Ovid-Medline, Ovid-Embase, Web of Science, Cochrane Library, China National Knowledge Infrastructure (CNKI), WeiPu database (VIP), WanFang database, and China Biology Medicine (CBM) databases to identify eligible articles from inception to July 2023. The scope of the search encompassed studies focusing on the development and/or validation of AI techniques for detecting BM in computed tomography (CT) or magnetic resonance imaging (MRI) images. A meta-analysis employing a hierarchical model was performed to calculate pooled sensitivity (SE), specificity (SP), area under the curve (AUC), positive likelihood ratio (PLR), negative likelihood ratio (NLR) and diagnostic odds ratio (DOR). The risk of bias and applicability were assessed using the quality assessment of diagnostic accuracy studies (QUADAS-2) tool, while the quality of evidence was evaluated using Grading of Recommendations Assessment, Development and Evaluation (GRADE). Result: We included 17 articles and extracted 70 lists of columns from 13 articles with a pooled SE of 0.89 (0.82-0.94), a pooled SP of 0.89 (0.83-0.93), a pooled AUC of 0.95 (0.93-0.97), PLR of 8.2 (5.30-12.8), NLR of 0.12 (0.07-0.21) and DOR of 67 (32-144). In addition, our study was based on different imaging modalities. Basing on CT, a pooled SE of 0.92 (0.83-0.96), SP 0.93 (0.88-0.96), and AUC 0.97 (0.95-0.98). Based on MRI, a pooled SE of 0.89 (0.80-0.95), SP 0.79 (0.71-0.85), and AUC 0.88 (0.85, 0.91). GRADE results showed a low certainty of evidence for the included studies. Conclusion: The present meta-analysis demonstrated the substantial diagnostic value of AI in identifying BM, with CT exhibiting superior performance compared to MR. Consequently, we advocated for the integration of AI into CT image analysis for BM screening. However, further large-scale prospective studies are needed to validate the clinical utility of AI in managing BM. Funding: National Natural Science Foundation of China (NO: 82360358) and Gansu Provincial People's Hospital Intramural Research Fund Program (NO: 21GSSYB-20). Declaration of Interest: The authors declare that they have no competing interests. ? 2024, The Authors. All rights reserved.