Scholarlink

<p>Accurate grading of buffalo oocytes is essential for improving the efficiency of in vitro embryo production (IVEP); however, conventional morphological assessment by embryologists is subjective and affected by inter-observer variability, resulting in inconsistent grading. To address this limitation, we developed a deep learning-based automated grading framework for buffalo oocytes. Preliminary experiments were conducted using multiple convolutional neural network architectures, including ResNet50, DenseNet121, VGG19, Xception, and EfficientNetV2S, combined with four different optimizers. Based on comparative performance, the ResNet50 model optimized with the Adam optimizer was selected for final evaluation and used to classify oocytes into Grades A, B, and C. Model predictions were compared with independent assessments provided by five expert embryologists on unseen images. The proposed model showed the highest agreement with Embryologist 1 (79.7% macro accuracy) and Embryologist 5 (65.4%), achieving complete agreement in specific cases and more than 80% accuracy across several test images. Grade-wise similarity analysis revealed strong concordance for Grade A (81.53%) and Grade B (83.34%), while moderate agreement was observed for Grade C (45.46%). This study represents the first application of deep learning for automated classification of buffalo oocytes and demonstrates that the proposed framework can provide a consistent and reproducible grading strategy. The model is intended as a decision-support tool rather than a replacement for embryologists, offering improved objectivity and reliability in oocyte quality assessment and contributing to the advancement of buffalo reproductive biotechnologies.</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

Automatic grading of buffalo oocyte using deep learning for enhancing precision in in vitro embryo production

  • 期刊论文

  • 发表日期：

    2026年3月14日

查看全文
  • Bharati Pandey,
  • Rutuja N. Shelke,
  • Manbir K. Singh,
  • Gaurav Tripathi,
  • Manoj K. Singh,
  • Naresh L. Selokar

摘要

Accurate grading of buffalo oocytes is essential for improving the efficiency of in vitro embryo production (IVEP); however, conventional morphological assessment by embryologists is subjective and affected by inter-observer variability, resulting in inconsistent grading. To address this limitation, we developed a deep learning-based automated grading framework for buffalo oocytes. Preliminary experiments were conducted using multiple convolutional neural network architectures, including ResNet50, DenseNet121, VGG19, Xception, and EfficientNetV2S, combined with four different optimizers. Based on comparative performance, the ResNet50 model optimized with the Adam optimizer was selected for final evaluation and used to classify oocytes into Grades A, B, and C. Model predictions were compared with independent assessments provided by five expert embryologists on unseen images. The proposed model showed the highest agreement with Embryologist 1 (79.7% macro accuracy) and Embryologist 5 (65.4%), achieving complete agreement in specific cases and more than 80% accuracy across several test images. Grade-wise similarity analysis revealed strong concordance for Grade A (81.53%) and Grade B (83.34%), while moderate agreement was observed for Grade C (45.46%). This study represents the first application of deep learning for automated classification of buffalo oocytes and demonstrates that the proposed framework can provide a consistent and reproducible grading strategy. The model is intended as a decision-support tool rather than a replacement for embryologists, offering improved objectivity and reliability in oocyte quality assessment and contributing to the advancement of buffalo reproductive biotechnologies.