An Efficient Deep Learning Model for Embryo Classification

This research addresses the challenge of embryo classification in the field of In Vitro Fertilization (IVF), using advanced Convolutional Neural Network (CNN) models. Traditional embryo evaluations, which depends on manual morphological analysis conducted by skilled embryologists, prompted the need for a more objective and efficient approach. Previous research efforts done using (VGG16, ResNet50, DenseNet201, InceptionV3, InceptionResNetV2) provided valuable insights, but a novel model is introduced in this study to overcome limitations and improve classification accuracy. The proposed model has a unique and efficient architecture. The experimental process involves following steps: the expansion of the dataset through data augmentation, the development of the novel CNN model, and comparative analysis against other established models. The proposed model emerged as the top performer in the experimental results, having an accuracy of 95.83%, precision of 98.25%, and a balanced F1-Score of 92.64%. The results validated the model's enhanced proficiency in predicting embryo quality, providing a more objective and efficient alternative to conventional evaluation methods. This model not only outperformed individual existing models but also introduced automation to the IVF embryo classification system, addressing the subjective and time-consuming nature of traditional evaluations.