Background and Aims: While reports suggest that the Inner Cell Mass (ICM) ratio may impact pregnancy rates, the Gardner criteria for grading ICM does not specify the ICM size or ratio. This study investigates the potential impact of the ICM ratio on predicting ICM grades using artificial intelligence. Methods: The study analyzed 1,963 Day 5 blastocyst images obtained from seven IVF clinics between June 2011 and May 2022. The images were matched with metadata, including ICM grade and ICM ratio. ICM ratio was computed as the ratio of the ICM area to the embryo area. ICM was graded according to the Gardner criteria and only ICM A and C grade groups were used to avoid ambiguity. Two CNN (Convolutional Neural Network) models were built: one with embryo images including ICM ratio and one without. Results: The use of ICM ratio in conjunction with the original grading model resulted in an improved model performance. Logistic regression analysis revealed the ICM ratio as a significant variable impacting embryo grade, with a p-value of less than 0.001. The coefficient for ICM ratio demonstrated an 8-fold increase in the odds ratio for grade A embryos with every 1-unit increase. Conclusions: We demonstrated that the CNN model could predict ICM grades with fair accuracy and that the inclusion of the ICM ratio significantly improved its performance. While CNNs are widely used for image analysis by extracting features from patterns, they may not effectively learn the characteristics of an area. To capture the important area-related features, measured values should be considered along with the image. Furthermore, the ICM ratio appears to be a significant feature that embryologists consider when grading embryos in clinical settings, beyond the Gardner criteria. Additional studies are warranted to investigate other features for evaluating ICM.
Read full abstract