Scholarlink

<p>Tissue-engineered scaffolds incorporating oxygen-releasing biomaterials (ORBs) are widely investigated for alleviating hypoxia in chronic diabetic wounds. However, experimental determination of oxygen release kinetics is time-consuming and resource-intensive. In this study, machine-learning-based regression models were developed to predict the release of dissolved oxygen (%) from hydrogel scaffolds containing calcium peroxide (CPO). Multiple linear regression (MLR) was implemented using scaffold parameters—CPO concentration (mg/mL), swelling ratio (%), mass remaining (%), and compressive modulus (kPa)—as predictor variables. Experimental oxygen release data, consisting of 150 observations collected over 35&#xa0;days, were used, with a 70:30 train–test split. The linear regression model achieved mean absolute errors of 1.58% and 1.76% for the training and testing datasets, respectively. Higher-order polynomial regression models did not yield satisfactory improvements in the coefficient of determination (R<sup>2</sup>) and exhibited signs of overfitting; therefore, linear and low-order polynomial models were retained for analysis. These findings demonstrate that linear regression provides a more robust and generalisable predictive framework for limited biomedical datasets. The proposed approach enables predictive optimisation of oxygen-releasing scaffolds and supports patient-specific scaffold design before in vitro and in vivo evaluation.</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-Based Predictive Modelling of Oxygen Release from Oxygen-Releasing Biomaterials

  • 期刊论文

  • 发表日期：

    2026年5月17日

查看全文
  • Pooja Thiyagarajan,
  • AramValartha Nayaki,
  • S. Hariharan,
  • Hemalatha Karnan

摘要

Tissue-engineered scaffolds incorporating oxygen-releasing biomaterials (ORBs) are widely investigated for alleviating hypoxia in chronic diabetic wounds. However, experimental determination of oxygen release kinetics is time-consuming and resource-intensive. In this study, machine-learning-based regression models were developed to predict the release of dissolved oxygen (%) from hydrogel scaffolds containing calcium peroxide (CPO). Multiple linear regression (MLR) was implemented using scaffold parameters—CPO concentration (mg/mL), swelling ratio (%), mass remaining (%), and compressive modulus (kPa)—as predictor variables. Experimental oxygen release data, consisting of 150 observations collected over 35 days, were used, with a 70:30 train–test split. The linear regression model achieved mean absolute errors of 1.58% and 1.76% for the training and testing datasets, respectively. Higher-order polynomial regression models did not yield satisfactory improvements in the coefficient of determination (R2) and exhibited signs of overfitting; therefore, linear and low-order polynomial models were retained for analysis. These findings demonstrate that linear regression provides a more robust and generalisable predictive framework for limited biomedical datasets. The proposed approach enables predictive optimisation of oxygen-releasing scaffolds and supports patient-specific scaffold design before in vitro and in vivo evaluation.