Scholarlink

<p>Prolactinomas are common pituitary adenomas that present significant diagnostic challenges due to their varied clinical manifestations. This systematic review assesses machine learning (ML) applications in prolactinoma detection by analyzing MRI scans and hormonal biomarkers across 25 studies published between 2000–2024. A comprehensive literature search was conducted across multiple databases, yielding 550 papers from which 25 relevant studies. This paper provides the entire matrix of literature reviewed for relational analysis. Conventional visual assessment of MRI scans by radiologists typically achieves 75–85% sensitivity for microadenomas, while ML models have demonstrated sensitivities exceeding 92% in multiple studies. Key findings indicate that convolutional neural networks (CNNs) show superior diagnostic accuracy compared to traditional methods, with an average accuracy exceeding 90% and significantly improved sensitivity in detecting microadenomas on MRI scans. ML models also demonstrate enhanced ability to distinguish prolactinomas from other Sellar lesions using novel metrics such as Prolactin-Volume-Ratio (PVR). However, challenges related to data quality and generalizability persist. The review identifies critical research gaps and proposes future directions, emphasizing the need for larger, diverse datasets and user-friendly ML tools to advance early detection and personalized treatment strategies for prolactinomas. </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

Machine learning in prolactinoma detection: a systematic review and meta analysis of diagnostic innovations

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月2日

查看全文
  • Kashif Raza Siddique,
  • Nivith P. Murali,
  • Urwa Khalid,
  • Naveed Jeelani Khan,
  • Asif Adil,
  • Kamil Reza Khondakar

摘要

Prolactinomas are common pituitary adenomas that present significant diagnostic challenges due to their varied clinical manifestations. This systematic review assesses machine learning (ML) applications in prolactinoma detection by analyzing MRI scans and hormonal biomarkers across 25 studies published between 2000–2024. A comprehensive literature search was conducted across multiple databases, yielding 550 papers from which 25 relevant studies. This paper provides the entire matrix of literature reviewed for relational analysis. Conventional visual assessment of MRI scans by radiologists typically achieves 75–85% sensitivity for microadenomas, while ML models have demonstrated sensitivities exceeding 92% in multiple studies. Key findings indicate that convolutional neural networks (CNNs) show superior diagnostic accuracy compared to traditional methods, with an average accuracy exceeding 90% and significantly improved sensitivity in detecting microadenomas on MRI scans. ML models also demonstrate enhanced ability to distinguish prolactinomas from other Sellar lesions using novel metrics such as Prolactin-Volume-Ratio (PVR). However, challenges related to data quality and generalizability persist. The review identifies critical research gaps and proposes future directions, emphasizing the need for larger, diverse datasets and user-friendly ML tools to advance early detection and personalized treatment strategies for prolactinomas.