Background/Objectives: Pancreatic ductal adenocarcinoma (PDAC) is an aggressive and lethal malignancy with increasing incidence and low survival rate, primarily due to the late detection of the disease. Radiomics has demonstrated its utility in recognizing patterns and anomalies not perceptible to the human eye. This systematic literature review aims to assess the application of radiomics in the analysis of pancreatic parenchyma images to identify early indicators predictive of PDAC. Methods: A systematic search of original research papers was performed on three databases: PubMed, Embase, and Scopus. Two reviewers applied the inclusion and exclusion criteria, and one expert solved conflicts for selecting the articles. After extraction and analysis of the data, there was a quality assessment of these articles using the Methodological Radiomics Score (METRICS) tool. The METRICS assessment was carried out by two raters, and conflicts were solved by a third reviewer. Results: Ten articles for analysis were retrieved. CT scan was the diagnostic imaging used in all the articles. All the studies were retrospective and published between 2019 and 2024. The main objective of the articles was to generate radiomics-based machine learning models able to differentiate pancreatic tumors from healthy tissue. The reported diagnostic performance of the model chosen yielded very high results, with a diagnostic accuracy between 86.5% and 99.2%. Texture and shape features were the most frequently implemented. The METRICS scoring assessment demonstrated that three articles obtained a moderate quality, five a good quality, and, finally, two articles yielded excellent quality. The lack of external validation and available model, code, and data were the major limitations according to the qualitative assessment. Conclusions: There is high heterogeneity in the research question regarding radiomics and pancreatic cancer. The principal limitations of the studies were mainly due to the nature of the trials and the considerable heterogeneity of the radiomic features reported. Nonetheless, the work in this field is promising, and further studies are still required to adopt radiomics in the early detection of PDAC.
背景/目的：胰腺导管腺癌（PDAC）是一种侵袭性强、致死率高的恶性肿瘤，其发病率不断上升且生存率较低，主要原因是疾病发现较晚。影像组学已证明其在识别人眼无法察觉的模式和异常方面的实用性。本系统性文献综述旨在评估影像组学在胰腺实质图像分析中的应用，以识别预测PDAC的早期指标。方法：在PubMed、Embase和Scopus三个数据库中对原始研究论文进行了系统性检索。两位评审员应用纳入和排除标准，并由一位专家解决文章选择中的分歧。数据提取和分析后，使用影像组学方法学质量评分（METRICS）工具对这些文章进行了质量评估。METRICS评估由两位评分者进行，分歧由第三位评审员解决。结果：共检索到十篇分析文章。所有文章均使用CT扫描作为诊断影像。所有研究均为回顾性研究，发表于2019年至2024年之间。文章的主要目标是建立基于影像组学的机器学习模型，以区分胰腺肿瘤与健康组织。所选模型报告的诊断性能结果非常高，诊断准确率在86.5%至99.2%之间。纹理和形状特征是最常使用的特征。METRICS评分评估显示，三篇文章质量中等，五篇质量良好，最后两篇质量优秀。根据定性评估，缺乏外部验证以及模型、代码和数据的可用性是主要局限性。结论：关于影像组学与胰腺癌的研究问题存在高度异质性。研究的主要局限性主要源于试验性质以及所报告的影像组学特征存在显著异质性。尽管如此，该领域的工作前景广阔，仍需进一步研究以将影像组学应用于PDAC的早期检测。
肿瘤(癌症)患者之家