Liu Xuewu, Lin Feng, Li Danni, Lei Nan
Radiology Department, The People's Hospital of Lezhi, Ziyang, Sichuan, China.
Front Oncol. 2025 Jan 8;14:1425665. doi: 10.3389/fonc.2024.1425665. eCollection 2024.
Radiomics has emerged as a promising approach for diagnosing, treating, and evaluating the prognosis of various diseases in recent years. Some investigators have utilized radiomics to create preoperative diagnostic models for tumor deposits (TDs) and perineural invasion (PNI) in rectal cancer (RC). However, there is currently a lack of comprehensive, evidence-based support for the diagnostic performance of these models. Thus, the accuracy of radiomic models was assessed in diagnosing preoperative RC TDs and PNI in this study.
PubMed, EMBASE, Web of Science, and Cochrane Library were searched for relevant articles from their establishment up to December 11, 2023. The radiomics quality score (RQS) was used to evaluate the risk of bias in the methodological quality and research level of the included studies.
This meta-analysis included 15 eligible studies, most of which employed logistic regression models (LRMs). For diagnosing TDs, the c-index, sensitivity, and specificity of models based on radiomic features (RFs) alone were 0.85 (95% CI: 0.79 - 0.90), 0.85 (95% CI: 0.75 - 0.91), and 0.82 (95% CI: 0.70 - 0.89); in the validation set, the c-index, sensitivity, and specificity of models based on both RFs and interpretable CFs were 0.87 (95% CI: 0.83 - 0.91), 0.91 (95% CI: 0.72 - 0.99), and 0.65 (95% CI: 0.53 - 0.76), respectively. For diagnosing PNI, the c-index, sensitivity, and specificity of models based on RFs alone were 0.80 (95% CI: 0.74 - 0.86), 0.64 (95% CI: 0.44 - 0.80), and 0.79 (95% CI: 0.68 - 0.87) in the validation set; in the validation set, the c-index, sensitivity, and specificity of models based on both RFs and interpretable CFs were 0.83 (95% CI: 0.77 - 0.89), 0.60 (95% CI: 0.48 - 0.71), and 0.90 (95% CI: 0.84 - 0.94), respectively.
Diagnostic models based on both RFs and CFs have proven effective in preoperatively diagnosing TDs and PNI in RC. This non-invasive method shows promise as a new approach.
https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=498660, identifier CRD42024498660.
近年来，放射组学已成为诊断、治疗和评估各种疾病预后的一种有前景的方法。一些研究人员利用放射组学为直肠癌（RC）中的肿瘤沉积（TDs）和神经周围侵犯（PNI）创建术前诊断模型。然而，目前这些模型的诊断性能缺乏全面的、基于证据的支持。因此，本研究评估了放射组学模型在术前诊断RC TDs和PNI中的准确性。
检索了PubMed、EMBASE、Web of Science和Cochrane图书馆从建库至2023年12月11日的相关文章。使用放射组学质量评分（RQS）评估纳入研究的方法学质量和研究水平中的偏倚风险。
这项荟萃分析纳入了15项符合条件的研究，其中大多数采用逻辑回归模型（LRMs）。对于诊断TDs，仅基于放射组学特征（RFs）的模型的c指数、敏感性和特异性分别为0.85（95%CI：0.79 - 0.90）、0.85（95%CI：0.75 - 0.91）和0.82（95%CI：0.70 - 0.89）；在验证集中，基于RFs和可解释临床特征（CFs）的模型的c指数、敏感性和特异性分别为0.87（95%CI：0.83 - 0.91）、0.91（95%CI：0.72 - 0.99）和0.65（95%CI：0.53 - 0.76）。对于诊断PNI，在验证集中仅基于RFs的模型的c指数、敏感性和特异性分别为0.80（95%CI：0.74 - 0.86）、0.64（95%CI：0.44 - 0.80）和0.79（95%CI：0.68 - 0.87）；在验证集中，基于RFs和可解释CFs的模型的c指数、敏感性和特异性分别为0.83（95%CI：0.77 - 0.89）、0.60（95%CI：0.48 - 0.71）和0.90（95%CI：0.84 - 0.94）。
基于RFs和CFs的诊断模型已被证明在术前诊断RC的TDs和PNI方面有效。这种非侵入性方法作为一种新方法显示出前景。
https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=498660，标识符CRD42024498660。
