FEMI: Next-Generation AI Platform for Comprehensive IVF Embryo Analysis and Selection

Background & Unmet Need

• Current IVF success heavily depends on accurate embryo assessment, requiring multiple complex tasks including ploidy prediction, quality scoring, and developmental monitoring
• Traditional diagnostic tools are costly, lack standardization, and often rely on invasive procedures like PGT-A testing
• Current AI models are constrained by insufficient training data diversity, reliance on manual input, and limited accessibility for clinical adoption
• Available AI platforms show reduced accuracy when analyzing specific developmental stages and lack standardization across different clinical settings
• High costs and varying regulations around genetic testing create barriers to access for many patients
• Unmet Need: Efficient non-invasive, and standardized approach for comprehensive embryo assessment

Technology Overview

• The Technology: FEMI is a foundation model trained on ~18 million time-lapse embryo images for comprehensive, non-invasive assessment of embryo viability and development
• Self-supervised learning approach enables comprehensive feature extraction without manual annotation requirements
• A single unified platform handles multiple critical assessment tasks through task-specific fine-tuning
• PoC Data: Achieves >0.75 AUROC for ploidy prediction using single images, improving to >0.80 AUROC with video sequences and >0.85 AUROC when including maternal age data
• Demonstrates >90% F1 score in embryo tracking across 96-112 hours post-insemination
• Predicts developmental timing with mean absolute errors under 6 hours across datasets

Technology Applications

• Clinical decision support tool for IVF clinics enabling standardized embryo selection
• Potential applications in broader reproductive medicine research and developmental biology for studying embryo morphology and development patterns

Technology Advantages

• State-of-the-art performance across multiple embryo assessment tasks
• Standardized, objective analysis with generalization across multiple clinical settings and datasets, enabling consistent embryo assessment protocols
• Provides cost-effective implementation through seamless integration with existing time-lapse imaging equipment in IVF clinics