Scholarlink

<p>Among various methods available for fabricating nanofibrous scaffolds used in tissue engineering, the electrospinning process stands out due to its simplicity, versatility, and scalability. However, this process encounters considerable challenges due to the complex, non-linear interactions among various electrospinning experimental parameters. Recently, machine learning (ML) has shown substantial promise in predictive modeling across various fields, indicating its potential to streamline the electrospinning process by predicting fiber diameter of electrospun nanofibrous scaffolds. To further enhance control over fiber diameter, a genetic algorithm (GA) was integrated into the ML model, forming an ML-GA integration developed to identify optimal combinations of electrospinning experimental parameters necessary to achieve user-defined target fiber diameters. This study explores the capability of ML to expedite the electrospinning process by accurately predicting the fiber diameter of polyvinyl alcohol (PVA) nanofibrous scaffolds. To accomplish this, a dataset was compiled consisting of 397 data points extracted from 30 scientific publications, including various electrospinning experimental parameters and their corresponding scaffold fiber diameters. The performance of various ML models was evaluated using the coefficient of determination (R<sup>2</sup>) score and root mean square error (RMSE). Among the evaluated ML models, extreme gradient boosting (XGB) and light gradient boosting machine (LGBM) exhibited the highest predictive performance. Specifically, the XGB model achieved an R<sup>2</sup> score of 0.94 and an RMSE of 79.89&#xa0;nm on the testing dataset. Based on these results, an XGB-GA integration was developed, in which GA utilized the trained XGB model to identify optimal experimental parameter sets for target fiber diameters ranging from 100&#xa0;nm to 1,000&#xa0;nm, demonstrating robust optimization capability and reducing the necessity for extensive experimental trial-and-error. These findings highlight the potential of combining data-driven predictive modeling with evolutionary optimization through ML-GA integration, enabling intelligent fabrication of nanofibrous scaffolds with precisely tailored characteristics for tissue engineering applications.</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

Data-driven prediction and optimization of electrospun nanofibrous scaffold diameters for tissue engineering applications using machine learning and genetic algorithms

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月24日

查看全文
  • Balakrishnan Subeshan,
  • Eylem Asmatulu

摘要

Among various methods available for fabricating nanofibrous scaffolds used in tissue engineering, the electrospinning process stands out due to its simplicity, versatility, and scalability. However, this process encounters considerable challenges due to the complex, non-linear interactions among various electrospinning experimental parameters. Recently, machine learning (ML) has shown substantial promise in predictive modeling across various fields, indicating its potential to streamline the electrospinning process by predicting fiber diameter of electrospun nanofibrous scaffolds. To further enhance control over fiber diameter, a genetic algorithm (GA) was integrated into the ML model, forming an ML-GA integration developed to identify optimal combinations of electrospinning experimental parameters necessary to achieve user-defined target fiber diameters. This study explores the capability of ML to expedite the electrospinning process by accurately predicting the fiber diameter of polyvinyl alcohol (PVA) nanofibrous scaffolds. To accomplish this, a dataset was compiled consisting of 397 data points extracted from 30 scientific publications, including various electrospinning experimental parameters and their corresponding scaffold fiber diameters. The performance of various ML models was evaluated using the coefficient of determination (R2) score and root mean square error (RMSE). Among the evaluated ML models, extreme gradient boosting (XGB) and light gradient boosting machine (LGBM) exhibited the highest predictive performance. Specifically, the XGB model achieved an R2 score of 0.94 and an RMSE of 79.89 nm on the testing dataset. Based on these results, an XGB-GA integration was developed, in which GA utilized the trained XGB model to identify optimal experimental parameter sets for target fiber diameters ranging from 100 nm to 1,000 nm, demonstrating robust optimization capability and reducing the necessity for extensive experimental trial-and-error. These findings highlight the potential of combining data-driven predictive modeling with evolutionary optimization through ML-GA integration, enabling intelligent fabrication of nanofibrous scaffolds with precisely tailored characteristics for tissue engineering applications.