Blasto-Net: An Explainable Multi-Task Learning for Blastocyst Segmentation, Grading, and Implantation Prediction

2026-06-24

AI Summary

This study addresses the challenges in blastocyst assessment during in vitro fertilization, where the zona pellucida (ZP), trophectoderm (TE), and inner cell mass (ICM) exhibit substantial structural differences yet share similar textures, complicating segmentation, morphological grading, and implantation prediction. To overcome these issues, this work proposes a multi-task deep learning model that simultaneously performs three-region segmentation, morphological grading, and implantation outcome prediction within a single forward pass. Built upon an EfficientNet-B3-UNet architecture, the model incorporates an edge-aware attention module (EAAM) alongside CBAM and employs a novel region-boundary joint loss function. Anatomical consistency and interpretability are validated via Grad-CAM++. On the HMC dataset, the model achieves Dice coefficients of 94.93%, 91.60%, and 88.82% for ICM, ZP, and TE segmentation, respectively, and attains an F1-score of 80.0% for implantation prediction.

Abstract

This study introduces Blasto-Net, a multi-task deep learning model for comprehensive blastocyst analysis. The proposed model performs three tasks simultaneously in a single forward pass: segmentation of the ZP, TE, and ICM compartments, morphological grading, and implantation outcome prediction. Accurate blastocyst analysis in in vitro fertilization (IVF) is challenging. The compartments often have similar textures but very different structures. To address these challenges, Blasto-Net employs an EfficientNet-B3 encoder with a UNet-style decoder enhanced by the Convolutional Block Attention Module (CBAM) and a novel Edge-Aware Attention Module (EAAM) to effectively capture both semantic and boundary information. To handle distinct compartment topologies, the network employs specialized segmentation heads and a composite region- and boundary-based loss. Additionally, Grad-CAM++ visualizations are used to verify the anatomical consistency of the model's predictions. Evaluated on a public HMC blastocyst dataset, Blasto-Net achieves Dice scores of 94.93%, 91.60%, and 88.82% for ICM, ZP, and TE, respectively, alongside an implantation F1-score of 80.0%. These results demonstrate that Blasto-Net offers an accurate, interpretable, and efficient solution for automated blastocyst assessment, with strong potential to support clinical decision-making in IVF.

Problem

• blastocyst analysis
• in vitro fertilization
• multi-task learning
• segmentation
• implantation prediction

Innovation

• multi-task learning
• Edge-Aware Attention Module
• blastocyst segmentation
• implantation prediction
• explainable AI