Scholarlink

<p>The outcome of In Vitro Fertilisation (IVF) success depends heavily on accurate embryo grading, which we have performed manually for many years. The authors develop EmbryoNet-VGG16, a system that functions as an automated embryo quality evaluation tool, combining Otsu segmentation with a modified Visual Geometry Group-16 (VGG16) Convolutional Neural Network (CNN) architecture. Training on 84 synthesised embryo pictures from a balanced dataset allowed our model to learn better generalisation. The healthcare imaging process begins with Otsu thresholding segmentation of embryo pictures and continues with the application of our 16-layer CNN model for embryonic quality assessment. The network contains specialised convolutional layers that identify important quality indicators through the analysis of border characteristics and structural integrity. Our EmbryoNet-VGG16 achieves superior classification results compared to traditional machine learning models, as indicated by an accuracy of 88.1%, with a precision of 0.90 and a recall of 0.86. This outperforms Random Forest and Decision Trees, as well as Logistic Regression models, which yield 83.41%, 82.31%, and 77.42%, respectively. EmbryoNet-VGG16 shows reliable good and poor embryo segregation by extracting quality features from manual expert assessments. The implementation of an automated system in IVF clinics would create standardised embryo assessment protocols that enhance treatment success rates while alleviating the time-consuming requirements of manual assessment processes.</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

EmbryoNet-VGG16 framework for deep learning-based embryo classification with Otsu segmentation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月1日

查看全文
  • M. Saraniya,
  • J. Anitha Ruth

摘要

The outcome of In Vitro Fertilisation (IVF) success depends heavily on accurate embryo grading, which we have performed manually for many years. The authors develop EmbryoNet-VGG16, a system that functions as an automated embryo quality evaluation tool, combining Otsu segmentation with a modified Visual Geometry Group-16 (VGG16) Convolutional Neural Network (CNN) architecture. Training on 84 synthesised embryo pictures from a balanced dataset allowed our model to learn better generalisation. The healthcare imaging process begins with Otsu thresholding segmentation of embryo pictures and continues with the application of our 16-layer CNN model for embryonic quality assessment. The network contains specialised convolutional layers that identify important quality indicators through the analysis of border characteristics and structural integrity. Our EmbryoNet-VGG16 achieves superior classification results compared to traditional machine learning models, as indicated by an accuracy of 88.1%, with a precision of 0.90 and a recall of 0.86. This outperforms Random Forest and Decision Trees, as well as Logistic Regression models, which yield 83.41%, 82.31%, and 77.42%, respectively. EmbryoNet-VGG16 shows reliable good and poor embryo segregation by extracting quality features from manual expert assessments. The implementation of an automated system in IVF clinics would create standardised embryo assessment protocols that enhance treatment success rates while alleviating the time-consuming requirements of manual assessment processes.