Computers in biology and medicine · 2025
From images to insights: Cell counting and uniformity grading of day 3 embryos.
СтатьяНазвание 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.
Tan Nguyen Duy, Huy Tran Phuong, Thuy Tran Thi Thanh, Tuyet Hoang Thi Diem, Son Dang Truong, Bao Pham The, Sang Vu Ngoc Thanh
Аннотация
BACKGROUND: Morphological evaluation of Day 3 embryos in vitro fertilization (IVF) is critical for selecting viable embryos for transfer but is hindered by subjectivity and variability in manual grading. Inter-observer and intra-observer inconsistencies, coupled with the limitations of two-dimensional imaging, compromise the accuracy of cell counting and uniformity assessment.
METHODS: This study proposes a hybrid approach integrating deep learning and image processing techniques to enhance and standardize evaluation processes, focusing on improving accuracy in cell counting and cell uniformity assessment. The extraction of blastomere boundaries is conducted using an active contour model to enhance the initial elliptical boundary coordinates surrounding the blastomere, which are identified and localized by the YOLOv8 object detection model. To enhance the precision of contour refinement in a noisy environment with intricate edge textures, Gradient Vector Flow (GVF) is utilized to standardize the gradient field within the embryo, leading to improved convergence of the active contour. Subsequently, the Normalized Uniformity Value (NUV) is calculated to evaluate variability in cell sizes, offering an objective metric for assessing developmental uniformity.
RESULTS: The hybrid model effectively handles complex imaging scenarios, including overlapping cells and low-contrast conditions. YOLOv8 delivers precise cell counting, while GVF-snake ensures accurate boundary delineation and spatial measurements. NUV offers a robust metric for consistent and reliable embryo grading, significantly enhancing decision-making in embryo selection.
CONCLUSION: In summary, our main contributions are: (1) We develop a hybrid model integrating YOLOv8 for precise blastomere detection with GVF-snake for accurate boundary refinement, enhancing embryo grading accuracy. (2) We reduce subjectivity and variability in embryo evaluation by combining deep learning with image processing, addressing key challenges in manual grading. (3) We introduce the NUV as a quantitative metric for assessing cell uniformity, providing a standardized approach to improve IVF decision-making.
METHODS: This study proposes a hybrid approach integrating deep learning and image processing techniques to enhance and standardize evaluation processes, focusing on improving accuracy in cell counting and cell uniformity assessment. The extraction of blastomere boundaries is conducted using an active contour model to enhance the initial elliptical boundary coordinates surrounding the blastomere, which are identified and localized by the YOLOv8 object detection model. To enhance the precision of contour refinement in a noisy environment with intricate edge textures, Gradient Vector Flow (GVF) is utilized to standardize the gradient field within the embryo, leading to improved convergence of the active contour. Subsequently, the Normalized Uniformity Value (NUV) is calculated to evaluate variability in cell sizes, offering an objective metric for assessing developmental uniformity.
RESULTS: The hybrid model effectively handles complex imaging scenarios, including overlapping cells and low-contrast conditions. YOLOv8 delivers precise cell counting, while GVF-snake ensures accurate boundary delineation and spatial measurements. NUV offers a robust metric for consistent and reliable embryo grading, significantly enhancing decision-making in embryo selection.
CONCLUSION: In summary, our main contributions are: (1) We develop a hybrid model integrating YOLOv8 for precise blastomere detection with GVF-snake for accurate boundary refinement, enhancing embryo grading accuracy. (2) We reduce subjectivity and variability in embryo evaluation by combining deep learning with image processing, addressing key challenges in manual grading. (3) We introduce the NUV as a quantitative metric for assessing cell uniformity, providing a standardized approach to improve IVF decision-making.