Euploidy Prediction

Product Description

Euploidy Prediction is a technological product recorded in 2025 in the Lattes CV of Prof. George Marsicano Corrêa . The work investigates the use of Machine Learning techniques for the prediction of embryonic euploidy in in vitro fertilization treatments, based on morphokinetic data obtained from Time-Lapse systems, as a less invasive alternative to Preimplantation Genetic Testing for Aneuploidy (PGT-A).

INPI Registration

In 2026, the software received a computer program registration certificate granted by INPI under process BR512026003650-0. The registration secures legal protection for 50 years, as established by Law No. 9.609/1998, and reinforces the institutional recognition of the technological output associated with the Artificial intelligence and data analysis for digital transformation project. More details are available in the news item about the registration .

Main Function and Results

The program performs embryonic euploidy prediction from the automated analysis of morphokinetic and morphological data obtained by continuous embryonic development monitoring systems (Time-Lapse). It uses Artificial Intelligence and Machine Learning techniques to identify temporal and morphological patterns associated with the genetic viability of embryos, providing a non-invasive probabilistic estimate of euploidy.

The developed model, based on an Artificial Neural Network of the Multilayer Perceptron (MLP) type, achieved 88.2% accuracy and AUC of 0.944, demonstrating high discriminatory capability between euploid and aneuploid embryos. Application of the LIME method made the predictions interpretable, increasing transparency and reliability in the medical context. Correlation analysis identified the most relevant morphokinetic variables, notably maternal age and the tb-t2b variable as factors with the greatest negative influence.

Purpose and Differentiators

The system is intended for assisted reproduction clinics, embryologists, researchers, and fertility centers, acting as a clinical decision-support tool for embryo selection. Unlike existing market solutions — which mostly rely on proprietary scores or subjective morphological analysis — the program employs statistically trained and validated machine learning models combined with explainability methods, which are essential requirements in the biomedical context.

Future Directions

Incorporation of new predictive models (including regression for continuous euploidy percentage estimation), integration with clinical systems, expansion of the data base for greater statistical robustness, graphical interfaces for clinical use, and auditing and traceability modules — broadening applicability, strengthening clinical validation, and enhancing social impact by reducing costs and risks and increasing treatment success rates.

Academic Linkage

• Maria Eduarda dos Santos Abritta Ferreira (product development)
• Sabrina Caldas Berno (product development)
• Prof. George Marsicano Corrêa (academic supervision and guidance)
• Rights holder: Fundação Universidade de Brasília (UnB)