Scholarlink

<p>Uterine adhesions, a common complication following repeated cesarean sections, can lead to significant issues such as chronic pain, infertility, and complications in subsequent pregnancies. Early prediction of adhesion severity is crucial for timely intervention and improved patient outcomes. This study aims to develop predictive models to assess the severity of uterine adhesions using machine-learning techniques. Utilizing a dataset of 300 clinical cases containing demographic, obstetric, intraoperative, and early postoperative indicators, all of which are available prior to the patient’s subsequent cesarean delivery, two models, Artificial Neural Networks (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS), were implemented. The dataset was split into training (70%), validation (15%), and testing (15%) subsets. The ANN model, featuring three hidden layers and using early stopping after 11 epochs, achieved an overall correlation coefficient (R-value) of 2242, with a test set R-value of 0.84802, indicating strong generalization. In contrast, the ANFIS model, designed with a hybrid approach combining neural networks with fuzzy logic, achieved a smooth error convergence, reaching a minimal error rate of approximately 9 × 10<sup>–6</sup>. The ANFIS model also demonstrated enhanced interpretability by utilizing fuzzy rules to capture nonlinear relationships between variables. The comparative analysis showed that while ANN offers efficient and accurate predictions, ANFIS provides deeper interpretability, making it a valuable tool for clinical decision-making. The findings suggest that integrating these models into clinical practice can support early diagnosis and personalized treatment, ultimately improving patient care. Moreover, the ANN and ANFIS architecture is intrinsically suited for high-performance and parallel computing settings, facilitating real-time analysis of extensive clinical datasets. This scalability emphasizes the study’s significance for supercomputing applications in medical decision assistance.</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

Early predictive modeling of uterine adhesions severity: a comparative study of ANN and ANFIS techniques

  • 期刊论文

  • 发表日期：

    2026年1月4日

查看全文
  • Abdelrahman T. Elgohr,
  • Mohamed S. Elhadidy,
  • Rehab Abdelhamid Aboshama,
  • M. A. Elazab

摘要

Uterine adhesions, a common complication following repeated cesarean sections, can lead to significant issues such as chronic pain, infertility, and complications in subsequent pregnancies. Early prediction of adhesion severity is crucial for timely intervention and improved patient outcomes. This study aims to develop predictive models to assess the severity of uterine adhesions using machine-learning techniques. Utilizing a dataset of 300 clinical cases containing demographic, obstetric, intraoperative, and early postoperative indicators, all of which are available prior to the patient’s subsequent cesarean delivery, two models, Artificial Neural Networks (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS), were implemented. The dataset was split into training (70%), validation (15%), and testing (15%) subsets. The ANN model, featuring three hidden layers and using early stopping after 11 epochs, achieved an overall correlation coefficient (R-value) of 2242, with a test set R-value of 0.84802, indicating strong generalization. In contrast, the ANFIS model, designed with a hybrid approach combining neural networks with fuzzy logic, achieved a smooth error convergence, reaching a minimal error rate of approximately 9 × 10–6. The ANFIS model also demonstrated enhanced interpretability by utilizing fuzzy rules to capture nonlinear relationships between variables. The comparative analysis showed that while ANN offers efficient and accurate predictions, ANFIS provides deeper interpretability, making it a valuable tool for clinical decision-making. The findings suggest that integrating these models into clinical practice can support early diagnosis and personalized treatment, ultimately improving patient care. Moreover, the ANN and ANFIS architecture is intrinsically suited for high-performance and parallel computing settings, facilitating real-time analysis of extensive clinical datasets. This scalability emphasizes the study’s significance for supercomputing applications in medical decision assistance.