文献速递肿瘤 PET/CT

人工智能在淋巴瘤中的作用:最新进展

The Role of AI in Lymphoma: An Update.
PubMed 文献 · Semin Nucl Med · 2025-03-11

本文综述人工智能在淋巴瘤PET/CT中的应用进展,包括预约管理减少失约、卷积神经网络重建降低放射性药物注射活度和患者剂量、3D U-Net自动分割改善总代谢肿瘤体积和总病灶糖酵解量化,并讨论分割性能差异、Deauville评分解释及标准化等挑战,指出需进一步研究解决AI误差和临床工作流整合问题。

要点

  • 人工智能用于预约规划可降低失约率并提高科室效率。
  • 卷积神经网络重建可在保持诊断准确性的同时降低注射活度和患者剂量。
  • 3D U-Net自动分割改善总代谢肿瘤体积和总病灶糖酵解量化,辅助预后分层。

核心结论

  • 人工智能在淋巴瘤PET/CT的预约规划、图像重建、自动分割和报告生成等方面具有转化潜力,但需进一步研究解决AI误差、生理性摄取区分和临床工作流整合等局限。
  • 经稳健验证和协调后,人工智能整合可能显著改善淋巴瘤诊疗,提高诊断精度、工作流效率和患者结局。

研究设计与样本

  • 本文为综述,检索范围未在原文摘要中报告。

主要发现

  • 人工智能在预约规划中已显示降低失约率和提高科室效率的效用。
  • 利用卷积神经网络(CNN)的先进重建技术可降低放射性药物注射活度和患者剂量,同时保持诊断准确性,尤其使需要多次扫描的年轻患者受益。
  • 主要使用3D U-Net架构的自动分割工具改善了总代谢肿瘤体积(TMTV)和总病灶糖酵解(TLG)等指标的量化,有助于预后判断和治疗分层。
  • 新兴大语言模型(LLM)在改善PET/CT报告、将自由文本报告转换为结构化格式及改善患者沟通方面显示出前景。

关键数值与推荐

  • PET/CT使用[18F]氟脱氧葡萄糖(FDG)是淋巴瘤分期、治疗反应评估和预后判断的金标准。

临床怎么用

  • 原文未据此给出推荐。

局限性

  • 分割性能存在变异性。
  • 对Deauville评分解释存在影响。
  • TMTV/TLG测量标准化问题。
  • 需进一步研究解决AI诱导的错误、生理性摄取区分及AI模型整合到临床工作流等局限。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

Malignant lymphomas encompass a range of malignancies with incidence rising globally, particularly with age. In younger populations, Hodgkin and Burkitt lymphomas predominate, while older populations more commonly experience subtypes such as diffuse large B-cell, follicular, marginal zone, and mantle cell lymphomas. Positron emission tomography/computed tomography (PET/CT) using [18F] fluorodeoxyglucose (FDG) is the gold standard for staging, treatment response assessment, and prognostication in lymphoma. However, interpretation of PET/CT is complex, time-consuming, and reliant on expert imaging specialists, exacerbating challenges associated with workforce shortages worldwide. Artificial intelligence (AI) offers transformative potential across multiple aspects of PET/CT imaging in this setting. AI applications in appointment planning have demonstrated utility in reducing nonattendance rates and improving departmental efficiency. Advanced reconstruction techniques leveraging convolutional neural networks (CNNs) enable reduced injected activities of radiopharmaceutical and patient dose whilst maintaining diagnostic accuracy, particularly benefiting younger patients requiring multiple scans. Automated segmentation tools, predominantly using 3D U-Net architectures, have improved quantification of metrics such as total metabolic tumour volume (TMTV) and total lesion glycolysis (TLG), facilitating prognostication and treatment stratification. Despite these advancements, challenges remain, including variability in segmentation performance, impact on Deauville Score interpretation, and standardization of TMTV/TLG measurements. Emerging large language models (LLMs) also show promise in enhancing PET/CT reporting, converting free-text reports into structured formats, and improving patient communication. Further research is required to address limitations such as AI-induced errors, physiological uptake differentiation, and the integration of AI models into clinical workflows. With robust validation and harmonization, AI integration could significantly enhance lymphoma care, improving diagnostic precision, workflow efficiency, and patient outcomes.

来源:Semin Nucl Med,查看原文。版权归原作者所有。

引用

  1. Cairns J, Frood R, Patel C, et al. The Role of AI in Lymphoma: An Update. Semin Nucl Med. 2025. PMID: 40069036. doi:10.1053/j.semnuclmed.2025.02.007

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。