Scholarlink

Purpose <p>Pancreatic cancer (PC) is one of the most lethal malignancies, often presenting with nonspecific symptoms and a dismal prognosis. Despite advancements in treatments, the 5-year survival rate remains low, highlighting the urgent need for effective early diagnostic approaches.</p> Methods <p>This study investigated the potential of an artificial intelligence (AI)-based approach to enhance diagnostic accuracy for PC by integrating diverse data, including clinical variables, laboratory results, and biomarkers such as two circulating microRNAs and two periodontal pathogens. A cohort of 123 PC patients and 120 matched non-cancer controls were used to train the machine learning (ML) models.</p> Results <p>Findings indicated that new-onset diabetes, levels of miR-21 and miR-155 in blood, and loads of <i>Porphyromonas gingivalis</i> and <i>Aggregatibacter actinomycetemcomitans</i> in the oral cavity were the most important predictors of PC. Among predictive models, the ensemble learning exhibited superior accuracy with an area under the curve (AUC) of 0.87, a sensitivity of 0.89, and a specificity of 0.86.</p> Conclusion <p>This research highlights the promise of integrating AI techniques to improve early detection of PC using non-invasive biomarkers. However, further external validation in diverse and multinational cohorts is needed to confirm generalizability before clinical implementation. If validated, this approach could pave the way for personalized screening strategies.</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

Harnessing artificial intelligence for detection of pancreatic cancer: a machine learning approach

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月1日

查看全文
  • Babak Khorsand,
  • Zeinab Hesami,
  • Samira Alipour,
  • Maryam Farmani,
  • Hamidreza Houri

摘要

Purpose

Pancreatic cancer (PC) is one of the most lethal malignancies, often presenting with nonspecific symptoms and a dismal prognosis. Despite advancements in treatments, the 5-year survival rate remains low, highlighting the urgent need for effective early diagnostic approaches.

Methods

This study investigated the potential of an artificial intelligence (AI)-based approach to enhance diagnostic accuracy for PC by integrating diverse data, including clinical variables, laboratory results, and biomarkers such as two circulating microRNAs and two periodontal pathogens. A cohort of 123 PC patients and 120 matched non-cancer controls were used to train the machine learning (ML) models.

Results

Findings indicated that new-onset diabetes, levels of miR-21 and miR-155 in blood, and loads of Porphyromonas gingivalis and Aggregatibacter actinomycetemcomitans in the oral cavity were the most important predictors of PC. Among predictive models, the ensemble learning exhibited superior accuracy with an area under the curve (AUC) of 0.87, a sensitivity of 0.89, and a specificity of 0.86.

Conclusion

This research highlights the promise of integrating AI techniques to improve early detection of PC using non-invasive biomarkers. However, further external validation in diverse and multinational cohorts is needed to confirm generalizability before clinical implementation. If validated, this approach could pave the way for personalized screening strategies.