EMBRACE A Multi-task Framework for Comprehensive Quality Assessment in Cleavage-stage Embryo

Jul 16, 2026

Upload by Oliver

Cleavage-stage embryo assessment in IVF requires integrated interpretation of cytoplasmic fragmentation, developmental stage, and blastomere symmetry, but conventional visual scoring suffers from observer variability. EMBRACE presents a multi-task deep learning framework that jointly segments cytoplasmic fragmentation, classifies t2/t4 developmental stages, and grades blastomere symmetry from static cleavage-stage microscopy images. The method uses a shared ResNet-50 backbone, C-MSFF feature fusion, a U-Net-style decoder, and task-specific classification heads. Trained on 9,137 annotated images, it reaches Dice 0.781 for segmentation, accuracy 0.995 for stage classification, and balanced accuracy 0.901 for symmetry grading.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages29

Reading Time1 h 13 min

Introduction

Clinical importance of cleavage-stage morphology assessment

Limitations of conventional visual assessment

Motivation for multi-task learning in IVF

EMBRACE: A Multi-task Framework for Comprehensive Quality Assessment in Cleavage-stage Embryo Anwar Hussain Sofia, Jung-Hua Wangb,c*, Ming-Jer Chend, Tsung-Hsien Leee, Yu-Chiao Yif, Ming-Kuan Linc, Yi-Chung Laib a International Master Program in Applied Artificial Intelligence, National Taiwan Ocean University, Keelung, 202301 , Taiwan b AI Research Center, National Taiwan Ocean University, Keelung 20224, Taiwan c Department of Electrical Engineering, National Taiwan Ocean University, Keelung 20224, Taiwan (e-mail: [j](jacksonlin092@gmail.com)[acksonlin092@gmail.com](jacksonlin092@gmail.com)) . d Department of Obstetrics and Gynecology, Lee Women's Hospital, Taichung 40652, Taiwan e Department of Obstetrics and Gynecology, Chung Shan Medical University Hospital, Taichung 40201, Taiwan f Department of Obstetrics and Gynecology, Taichung Veterans General Hospital, Taichung 407219, Taiwan * Corresponding author: [j](jhwang@email.ntou.edu.tw)[hwang@email.ntou.edu.tw](jhwang@email.ntou.edu.tw), ORCID ID:0000-0003-1769-7396, Keelung, 202301, Taiwan Abstract Cleavage-stage embryo assessment in in vitro fertilization requires the integrated interpretation of cytoplasmic fragmentation, developmental stage, and blastomere symmetry. However, conventional visual assessment is affected by observer variability, particularly when fragmented regions are small, irregular, or low contrast. This study presents EMBRACE, a multi-task deep learning framework for jointly performing cytoplasmic-fragmentation segmentation, t2/t4 developmental-stage classification, and blastomere-symmetry grading from static cleavage-stage embryo microscopy images. EMBRACE combines a shared ResNet-50 backbone, a concatenation-based multi-scale feature-fusion (C-MSFF) module, a U-Net-style segmentation decoder, and two task-specific classification heads. After predefined inclusion and exclusion criteria, 9,137 annotated embryo images were divided into 7,309 training, 914 validation, and 914 held-out test images. On the held-out test set, EMBRACE achieved a Dice coefficient of 0.781 and an intersection over union of 0.677 for fragmentation segmentation. Developmental-stage classification achieved an accuracy of 0.995, macro-F1 of 0.994, and AUC of 1.000. Blastomere-symmetry grading achieved a balanced accuracy of 0.901, macro-F1 of 0.907, and quadratic weighted kappa of 0.859. These findings support the feasibility of combining spatially inspectable fragmentation localization with embryo-level morphology assessment in a single framework. External and prospective validation is required before clinical deployment. Keywords: Cleavage-stage embryo, Multi-task learning, Fragmentation Segmentation, Blastomere symmetry, Cytoplasmic fragmentation, Embryo morphology, Medical image segmentation, in vitro fertilization. 1. Introduction Embryo assessment remains a central decision point in In Vitro Fertilization (IVF), where laboratory evaluation supports decisions related to embryo transfer, continued culture, and cryopreservation. In routine embryology practice, embryo prioritization commonly relies on non-invasive morphological features, including blastomere number, cytoplasmic appearance, fragmentation, blastomere regularity, and developmental progression [1,2] . Although embryo developmental competence is influenced by biological, laboratory, and patient-specific factors, morphology-based assessment remains clinically important because it provides immediate visual information without additional embryo manipulation. At the cleavage stage, embryo evaluation requires the combined interpretation of multiple morphology-related features rather than reliance on a single visual cue. Developmental cell stage reflects cleavage progression and provides temporal context for embryo assessment. Cytoplasmic fragmentation represents anucleate cytoplasmic material and has been associated with reduced implantation and pregnancy potential. Blastomere symmetry reflects the re

FAQs

What tasks does EMBRACE perform on cleavage-stage embryo images?

EMBRACE jointly performs cytoplasmic-fragmentation segmentation, t2/t4 developmental-stage classification, and blastomere-symmetry grading from static microscopy images.

Why does the framework rely on t2/t4 instead of tPNf?

tPNf occurs before visible two- or four-cell blastomere structures, so it is incompatible with symmetry grading and localized fragmentation segmentation that require visible morphology.

How well does EMBRACE perform on the held-out test set?

On the held-out set, it achieves Dice 0.781 and IoU 0.677 for fragmentation segmentation, accuracy 0.995 with macro-F1 0.994 and AUC 1.000 for developmental-stage classification, and balanced accuracy 0.901 with macro-F1 0.907 and quadratic weighted kappa 0.859 for symmetry grading.

Jul 16, 2026

Upload by Oliver