Study question

Can AI-driven radiomics analysis identify distinct morphological changes in aging oocytes?

Summary answer

Radiomics-based AI models successfully classified oocytes by age and revealed oocytes from older mice exhibited larger and smoother cytoplasmic structures.

What is known already

Oocyte aging is one of the most critical yet challenging issues in fertility treatment. While its functional decline is well established, the specific morphological changes in aging oocytes remain unclear due to the difficulty of direct evaluation. Genetic testing is not always feasible, and manually labeling every detail is impractical and expensive. AI models for oocytes to predict blastulation exist, but the underlying morphological features that drive these predictions remain poorly understood due to the inherent black-box nature. By leveraging AI to detect and quantify distinct morphological patterns in aging oocytes, we might identify biologically meaningful explanations for oocyte aging.

Study design, size, duration

We analyzed 695 oocyte images from B6D2F1 (BDF1) mice collected at a single research center between July 2022 and January 2023. The samples were separated into two groups based on maternal age, 396 from young mice and 299 from old mice. Radiomics is a method that extracts a large number of features from medical images using data-characterisation algorithms. Radiomics features were extracted from the cytoplasm areas through expert-labelled annotations, while removing the polar body region.

Participants/materials, setting, methods

The mice were divided into two groups: young mice (7-9 weeks old) and old mice (62-68 weeks old). Mature MII oocytes were collected following superovulation and imaged using a 200x inverted microscope. The dataset was partitioned into 80% training and validation sets, and 20% test set. Radiomics analysis focused on nine distinct shape features to quantify differences between oocytes from young and aged mice. Five Machine learning models were optimized to maximize classification accuracy.

Main results and the role of chance

Among five machine learning models tested—Random Forest, Logistic Regression, Support Vector Machine, K-Nearest Neighbor, and Gradient Boosting—Logistic Regression achieved the highest classification performance, with an AUROC of 0.792 and an accuracy of 75.6%. Gradient Boosting demonstrated comparable results, with an AUROC of 0.785 and an accuracy of 74.6%. Morphological analysis revealed significant differences in cytoplasmic shape between young and old mice. Old mice exhibited greater elongation, major axis length, maximum diameter, and surface area. Parenthetical values represent the mean measurements for old and young mice, respectively, along with their p-values. Specifically, cytoplasms from old mice had longer major axis lengths (671.51 vs. 662.04, p < 0.001), minor axis lengths (640.76 vs. 632.99, p < 0.001), maximum diameters (686.34 vs. 675.93, p = 0.016), and larger surface areas (337,420.18 vs. 328,669.35, p < 0.001). In contrast, cytoplasms from young mice were characterized by a higher perimeter-to-surface (P/S) ratio, indicating more complex and rougher surfaces (0.00709 vs. 0.00729, p < 0.001). These findings suggest that while old mice display smoother, simpler cytoplasmic surfaces, the cytoplasmic regions in young mice exhibit greater surface complexity and irregularity.

Limitations, reasons for caution

This study is constrained by the use of a limited dataset derived from mice oocytes, which may restrict the generalizability of the findings to human applications. Additional investigations are warranted to explore subgroups within the aged oocyte population that exhibit morphological characteristics akin to younger oocytes, and vice versa.

Wider implications of the findings

This study provides quantitative evidence that oocyte aging alters cytoplasmic morphology. AI-driven radiomics offers a non-invasive method for assessing oocyte quality, bridging predictive modeling with biological interpretation. Integrating AI in reproductive medicine could enhance fertility assessments, but validation in human oocytes is needed to confirm clinical relevance.

Trial registration number

No