Microscopic Video-Based Grouped Embryo Segmentation: A Deep Learning Approach.

PURPOSE: The primary aim of this research is to enhance the utilization of advanced deep learning (DL) techniques in the domain of in vitro fertilization (IVF) by presenting a more refined approach to the segmentation and organization of microscopic embryos. This study also seeks to establish a comprehensive embryo database that can be employed for future research and educational purposes. METHODS: This study introduces an advanced methodology for embryo segmentation and organization using DL. The approach comprises three primary steps: Embryo Segmentation Model, Segmented Embryo Image Organization, and Clear and Blur Image Classification. The proposed approach was rigorously evaluated on a sample of 5182 embryos extracted from 362 microscopic embryo videos. RESULTS: The study's results show that the proposed method is highly effective in accurately segmenting and organizing embryo images. This is evidenced by the high mean average precision values of 1.0 at an intersection over union threshold of 0.5 and across the range of 0.5 to 0.95, indicating a robust object detection capability that is vital in the IVF process. Segmentation of images based on various factors such as the day of development, patient, growth medium, and embryo facilitates easy comparison and identification of potential issues. Finally, appropriate threshold values for clear and blur image classification are proposed. CONCLUSION: The suggested technique represents an indispensable stage of data preparation for IVF training and education. Furthermore, this study provides a solid foundation for future research and adoption of DL in IVF, which is expected to have a significant positive impact on IVF outcomes.

Leaf Recognition Based on Joint Learning Multiloss of Multimodel Convolutional Neural Networks: A Testing for Vietnamese Herb.

A new modification of multi-CNN ensemble training is investigated by combining multiloss functions from state-of-the-art deep CNN architectures for leaf image recognition. We first apply the U-Net model to segment leaf images from the background to improve the performance of the recognition system. Then, we introduce a multimodel approach based on a combination of loss functions from the EfficientNet and MobileNet (called as multimodel CNN (MMCNN)) to generalize a multiloss function. The joint learning multiloss model designed for leaf recognition allows each network to perform its task and cooperate with the others simultaneously, where knowledge from various trained deep networks is shared. This cooperation-proposed multimodel is forced to deal with more complicated problems rather than a simple classification. Therefore, the network can learn much rich information and improve its generalization capability. Furthermore, a multiloss trade-off strategy between two deep learning models can reduce the effect of redundancy problems in ensemble classifiers. The performance of our approach is evaluated by our custom Vietnamese herbal leaf species dataset, and public datasets such as Flavia, Leafsnap, and Folio are used to build test cases. The results confirm that our approach enhances the leaf recognition performance and outperforms the current standard single networks while having less low computation cost.