Scholarlink

Objective <p>This study aims to develop a practical predictive model for day 4 embryonic developmental potential, focusing on analyzing key influencing factors and visually presenting their relative importance, and evaluating the model's clinical applicability.</p> Methods <p>Patients undergoing fresh day 4 embryo transfer were included. Embryos cultured to day 5 or 6 were randomly allocated into training (70%) and validation (30%) datasets. Predictive factors for transferable blastocyst formation were identified using univariate and multivariable logistic regression analyses. The discriminative ability of three models—Model 1 (day 3 morphology), Model 2 (day 3 morphology + iDAScore), and Model 3 (day 3 morphology + D4 morphology + iDAScore)—was compared using receiver operating characteristic (ROC) analysis. A nomogram was developed using independent predictors, and its performance was assessed through calibration analyses. In addition, the predictive accuracy for high-quality blastocyst formation, implantation, and live birth rates was evaluated.</p> Results <p>The data from 9779 embryos across 2409 patients were analyzed. Among 5977 embryos cultured to the blastocyst stage, 4201 were assigned to the training cohort and 1776 to the validation cohort. Degree of compaction, iDAScore, and cell numbers on day 3 were identified as independent predictors of transferable blastyst formation. Model 3 outperformed the other models, achieving an area under the ROC curve (AUC) of 0.841 (95% CI 0.824–0.858) in the training cohort and 0.828 (95% CI 0.801–0.855) in the validation cohort. The nomogram based on Model 3 demonstrated strong calibration, with predicted and observed rates closely aligned (relative mean error: 0.019 in the training cohort, 0.360 in the validation cohort). The model also accurately predicted high-quality blastocyst formation (AUC: 0.808, 95% CI 0.789–0.826), implantation rates (AUC: 0.635, 95% CI 0.598–0.672), and live birth rates (AUC: 0.643, 95% CI 0.593–0.692).</p> Conclusions <p>Degree of compaction on day 4, iDAScore, and cell numbers on day 3 are independent predictors of transferable blastocyst formation. The proposed nomogram provides an effective tool for predicting embryonic development potential, with strong performance across multiple outcomes.</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

Practical prediction model for day 4 embryonic development potential: integration of morphology and AI scoring

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月21日

查看全文
  • Yasong Geng,
  • Fangfang Dai,
  • Haoyang Dai,
  • Linlin Tao,
  • Guozhen Li,
  • Jing Ma,
  • Lingyin Kong,
  • Bo Zheng,
  • Fang Wang

摘要

Objective

This study aims to develop a practical predictive model for day 4 embryonic developmental potential, focusing on analyzing key influencing factors and visually presenting their relative importance, and evaluating the model's clinical applicability.

Methods

Patients undergoing fresh day 4 embryo transfer were included. Embryos cultured to day 5 or 6 were randomly allocated into training (70%) and validation (30%) datasets. Predictive factors for transferable blastocyst formation were identified using univariate and multivariable logistic regression analyses. The discriminative ability of three models—Model 1 (day 3 morphology), Model 2 (day 3 morphology + iDAScore), and Model 3 (day 3 morphology + D4 morphology + iDAScore)—was compared using receiver operating characteristic (ROC) analysis. A nomogram was developed using independent predictors, and its performance was assessed through calibration analyses. In addition, the predictive accuracy for high-quality blastocyst formation, implantation, and live birth rates was evaluated.

Results

The data from 9779 embryos across 2409 patients were analyzed. Among 5977 embryos cultured to the blastocyst stage, 4201 were assigned to the training cohort and 1776 to the validation cohort. Degree of compaction, iDAScore, and cell numbers on day 3 were identified as independent predictors of transferable blastyst formation. Model 3 outperformed the other models, achieving an area under the ROC curve (AUC) of 0.841 (95% CI 0.824–0.858) in the training cohort and 0.828 (95% CI 0.801–0.855) in the validation cohort. The nomogram based on Model 3 demonstrated strong calibration, with predicted and observed rates closely aligned (relative mean error: 0.019 in the training cohort, 0.360 in the validation cohort). The model also accurately predicted high-quality blastocyst formation (AUC: 0.808, 95% CI 0.789–0.826), implantation rates (AUC: 0.635, 95% CI 0.598–0.672), and live birth rates (AUC: 0.643, 95% CI 0.593–0.692).

Conclusions

Degree of compaction on day 4, iDAScore, and cell numbers on day 3 are independent predictors of transferable blastocyst formation. The proposed nomogram provides an effective tool for predicting embryonic development potential, with strong performance across multiple outcomes.