Scholarlink

<p>Pancreatic cancer continues to be one of oncology’s most formidable challenges, with a stubbornly low five-year survival rate, often hovering below 10%, due to late-stage diagnosis and limited therapeutic options. Recent advances in artificial intelligence (AI), particularly machine learning (ML), have opened new possibilities for improving early detection, risk stratification, and personalized treatment in oncology. This review aims to explore the clinical applications and challenges of ML in the management of pancreatic cancer. To provide a comprehensive overview of ML-based approaches for screening, diagnosis, and treatment of pancreatic cancer, with a focus on model performance, clinical integration, and ethical-legal considerations. We conducted a structured literature review of peer-reviewed studies published between 2019 and 2025, using PubMed, Scopus, and Web of Science. Inclusion criteria focused on studies with validated ML models or cohorts exceeding 100 participants. Applications across early detection, biomarker-based diagnosis, electronic health records (EHRs), imaging analysis, and personalized treatment strategies were included. ML models demonstrated high accuracy (AUROC 0.84–0.97) across various data types, including computer tomography (CT) imaging, serum biomarkers, and EHRs. Notably, integrated models combining molecular and clinical data outperformed traditional diagnostic approaches. However, real-world adoption remains limited due to data heterogeneity, lack of external validation, and ethical concerns such as bias, transparency, and patient consent. ML offers transformative potential for managing pancreatic cancer, yet its clinical implementation requires multidisciplinary collaboration, robust validation, and alignment with regulatory frameworks. Future efforts should focus on model generalizability, interpretability, explainability, and integration into clinical workflows to improve early detection and patient outcomes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Challenges in the Clinical Application of Machine Learning for Pancreatic Cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月5日

查看全文
  • Ivana Večurkovská,
  • Veronika Roškovičová,
  • Jana Kaťuchová

摘要

Pancreatic cancer continues to be one of oncology’s most formidable challenges, with a stubbornly low five-year survival rate, often hovering below 10%, due to late-stage diagnosis and limited therapeutic options. Recent advances in artificial intelligence (AI), particularly machine learning (ML), have opened new possibilities for improving early detection, risk stratification, and personalized treatment in oncology. This review aims to explore the clinical applications and challenges of ML in the management of pancreatic cancer. To provide a comprehensive overview of ML-based approaches for screening, diagnosis, and treatment of pancreatic cancer, with a focus on model performance, clinical integration, and ethical-legal considerations. We conducted a structured literature review of peer-reviewed studies published between 2019 and 2025, using PubMed, Scopus, and Web of Science. Inclusion criteria focused on studies with validated ML models or cohorts exceeding 100 participants. Applications across early detection, biomarker-based diagnosis, electronic health records (EHRs), imaging analysis, and personalized treatment strategies were included. ML models demonstrated high accuracy (AUROC 0.84–0.97) across various data types, including computer tomography (CT) imaging, serum biomarkers, and EHRs. Notably, integrated models combining molecular and clinical data outperformed traditional diagnostic approaches. However, real-world adoption remains limited due to data heterogeneity, lack of external validation, and ethical concerns such as bias, transparency, and patient consent. ML offers transformative potential for managing pancreatic cancer, yet its clinical implementation requires multidisciplinary collaboration, robust validation, and alignment with regulatory frameworks. Future efforts should focus on model generalizability, interpretability, explainability, and integration into clinical workflows to improve early detection and patient outcomes.