.png){kind=icon}
.png){kind=icon}
Return
From static snapshots to longitudinal trajectories: artificial intelligence in women’s reproductive and ovarian health
.png){kind=icon}
OA
AI
Y
Y
Y
Y
X
DOI：10.3389/fendo.2026.1893963.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
AimTo distinguish ovarian longitudinal biology from direct ovarian AI evidence and transferred time-series methodology; and to evaluate how temporal information is used and linked to clinical action across ovarian malignancy; PCOS; ovarian aging/POI; and ART.MethodsThis structured narrative review mapped prior reviews and was supplemented by design verification of representative primary studies identified through searches covering 1 January 1989 to 1 January 2026. Two authors independently conducted the design-level verification and classification; disagreements were resolved by consensus. Evidence was classified as direct ovarian AI evidence; ovarian longitudinal biological evidence; or transferred methodological evidence. Temporal use and decision linkage were classified as T0 static prediction; T1 serial-feature summarization; T2 explicit time-aware modeling; or T3 decision-linked temporal modeling. T0–T3 describe how temporal information is used and linked to clinical decisions; they do not represent an automatic hierarchy of model quality. Recorded/calendar time; biological time; and decision time were distinguished because observation time is not necessarily equivalent to biological time.ResultsAvailable review-level evidence and the representative primary studies verified in detail indicated that AI applications in PCOS; ovarian-cancer imaging; and ovarian-reserve assessment were predominantly T0. CA-125 velocity; half-life; and nadir represented T1 serial-feature evidence rather than AI time-series modeling. Direct T2 evidence was sparse and included a longitudinal CA-125 joint model and a small number of within-cycle ART prescription models. T3 evidence linking outputs to prespecified action thresholds; calibration; false-positive burden; net benefit; clinician override; and prospective clinical outcomes was limited. Several studies described as longitudinal AI instead established longitudinal ovarian biology or transferred temporal methods from other diseases.ConclusionThe clinical value of temporal AI in ovarian health depends less on architectural complexity than on correct alignment with ovarian biology; direct within-patient temporal evidence; calibrated validation; and a beneficial clinical action.
Keywords:
PCOS
artificial intelligence
longitudinal studies
ovarian neoplasms
ovarian reserve
Journal
IF:
4.6
Papers:
1.9W
Citations:
6.4W
Organization
Cited Papers