Reproductive biomedicine online · 2025
Improving embryo ploidy prediction: a machine learning approach using morphokinetic meta-variables and clinical data.
СтатьяМногоцентровое исследованиеНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Güell-Penas Enric, Ferrer-Buitrago Minerva, Robles Empar Ferrer I, Lliso Carmen Calatayud, Rugero Fernando Marina, Olmedo Pilar Nieto, Rodié Ester Vilamajó I, Rubira María José López, Arcos Iria Martínez, Álvarez Elsi Suárez, Soler Laura Marquès, Riuró Eugènia Francisco-Busquets, Miguel Ignacio Santiago Álvarez, Méndez Eva Palacios, López Alba Mauri, Pérez José Ignacio Mazzanti
Аннотация
RESEARCH QUESTION: Can a machine learning model integrating time-lapse morphokinetic meta-variables with clinical data predict embryo aneuploidy accurately for improved non-invasive embryo selection?
DESIGN: A retrospective multicentre cohort study using time-lapse data from 1190 blastocysts from nine Spanish fertility clinics (2017-2024), with 70% (n = 833) and 30% (n = 357) used for model training/testing and external validation, respectively. The primary dataset included well-defined outcomes (live birth or aneuploidy diagnosis via preimplantation genetic testing). The LIFE Predict v1.1 model integrated clinical data and novel morphokinetic meta-variables (range and mean absolute error) representing deviations from expected embryo development. Model performance was assessed using area under the receiver operating characteristic curve (AUC-ROC) and confusion matrix metrics. Logistic regression was used to calculate OR for risk of aneuploidy. Morphological assessments using ASEBIR grading were combined with algorithmic scoring.
RESULTS: The final ensemble model achieved an AUC of 0.824 (95% CI 0.796-0.857) in cross-validation and 0.818 (95% CI 0.794-0.851) in external validation. The LIFE Predict v1.1 score showed a significant inverse relationship with risk of aneuploidy, with each one-point decrease increasing the odds of aneuploidy by 76% (OR = 1.76, 95% CI 1.52-2.05). The aneuploidy rate decreased across ascending score quartiles: 76.4% (lowest), 64.0%, 25.8% and 13.3% (highest). Combining morphological grading with the LIFE Predict v1.1 model revealed substantial risk stratification within identical morphological categories, with A-grade (Day 5) embryos showing aneuploidy rates from 11-14% (highest score quartiles) to 60-86% (lowest quartiles).
CONCLUSIONS: The LIFE Predict v1.1 model predicts embryo outcomes accurately using morphokinetic meta-variables and clinical data, providing actionable risk stratification that complements conventional morphological assessment for enhanced non-invasive embryo selection. Prospective clinical validation is required to confirm its real-world utility.
DESIGN: A retrospective multicentre cohort study using time-lapse data from 1190 blastocysts from nine Spanish fertility clinics (2017-2024), with 70% (n = 833) and 30% (n = 357) used for model training/testing and external validation, respectively. The primary dataset included well-defined outcomes (live birth or aneuploidy diagnosis via preimplantation genetic testing). The LIFE Predict v1.1 model integrated clinical data and novel morphokinetic meta-variables (range and mean absolute error) representing deviations from expected embryo development. Model performance was assessed using area under the receiver operating characteristic curve (AUC-ROC) and confusion matrix metrics. Logistic regression was used to calculate OR for risk of aneuploidy. Morphological assessments using ASEBIR grading were combined with algorithmic scoring.
RESULTS: The final ensemble model achieved an AUC of 0.824 (95% CI 0.796-0.857) in cross-validation and 0.818 (95% CI 0.794-0.851) in external validation. The LIFE Predict v1.1 score showed a significant inverse relationship with risk of aneuploidy, with each one-point decrease increasing the odds of aneuploidy by 76% (OR = 1.76, 95% CI 1.52-2.05). The aneuploidy rate decreased across ascending score quartiles: 76.4% (lowest), 64.0%, 25.8% and 13.3% (highest). Combining morphological grading with the LIFE Predict v1.1 model revealed substantial risk stratification within identical morphological categories, with A-grade (Day 5) embryos showing aneuploidy rates from 11-14% (highest score quartiles) to 60-86% (lowest quartiles).
CONCLUSIONS: The LIFE Predict v1.1 model predicts embryo outcomes accurately using morphokinetic meta-variables and clinical data, providing actionable risk stratification that complements conventional morphological assessment for enhanced non-invasive embryo selection. Prospective clinical validation is required to confirm its real-world utility.