Quantitative Assessment of Free Steroid Hormones to Characterize Endometrial Receptivity Towards Improving Pregnancy Outcomes in Frozen Embryo Transfer (FET) Cycles - PROJECT SUMMARY Frozen embryo transfer (FET) failure remains a major barrier to successful in vitro fertilization, contributing to substantial financial and emotional burden on patients. Today, nearly half of competent embryos still fail to implant. An important contributor to high implantation failure rate is the lack of accurate tools to assess endometrial readiness status. To address this unmet clinical need, we propose to develop TruFERT™, a quantitative platform that offers a fundamentally new way to assess endometrial receptivity and optimize FET timing and outcomes through three key innovations: incorporating free steroid hormone interactome during optimized preimplantation windows into ART decision-making, applying AI-driven multimodal modeling to define individualized receptivity windows, and translating the AI-generated integrated endometrial receptivity assessment into a scalable, noninvasive diagnostic framework integrated into clinical practice. The TruFERT™ integrates dynamic trajectory of free hormone profiles during optimized preimplantation time windows, endometrial thickness and morphology, cycle parameters, and other clinical characteristics into a clinical decision support. We propose a two-phase development program. In the R61 phase, we will measure total and free estradiol, progesterone, testosterone, and cortisol concentrations in archived, de-identified samples collected from 350 prior FET cycles, representing different cycle types and preimplantation time points. We will develop an explainable AI model for endometrial preparedness by relating integrated free hormone profiles, endometrial thickness and morphology, cycle parameters, and other clinical characteristics to clinical outcomes (implantation, clinical pregnancy, miscarriage). Multiple model classes will be evaluated, including gradient boosting methods (XGBoost, LightGBM), neural architectures (TabNet), and generalized additive models, as appropriate. Model performance will be assessed using discrimination, calibration, and decision-analytic metrics. We will develop a lightweight prototype interface capable of ingesting structured hormone and clinical inputs and returning (i) a predicted receptivity probability, (ii) the key physiological drivers of the prediction, and (iii) recommended free-hormone “target zones”. Transition to R33 will require meeting pre-specified go-no-go criteria based on safety and proof-of-concept. The R33 phase will prospectively validate the integrated endometrial receptivity index in natural and programmed cycles in 692 women undergoing FET and pilot a free-hormone–guided clinical decision workflow to optimize hormone dosing or transfer timing and compare it with the 'standard-of-care . We will also deploy a HIPAA-compliant, FHIR/HL7- integrated XAI decision-support system embedded in the Boston IVF electronic medical record. The successful completion of the proposed R61/R33 project will yield the first validated, scalable, and noninvasive, IP-protected, diagnostic platform that provides a rigorous quantitative framework predictive of endometrial receptivity, thereby reducing the rates of failed FETs.
