The Impact of Image Acquisition in Cervical Ultrasound on AI-Based Prediction of Preterm Birth in Clinical Practice

Rigshospitalet, Denmark2,000 enrolled

Overview

This study prospectively evaluates whether the performance of an already-developed artificial intelligence (AI) model for predicting spontaneous preterm birth changes when cervical ultrasound images are obtained using different ultrasound image settings. The primary research question is whether the AI model performs differently across images acquired with different imaging settings.

Study Type

OBSERVATIONAL

Enrollment

2,000

Conditions

Preterm Birth Artificial Intelligence (AI) in Diagnosis

Interventions

Cervical ultrasound image acquisitionOTHER

Acquisition of cervical ultrasound images with variation in image acquisition parameters.

Eligibility

Medical Language ↔ Plain English

Inclusion Criteria: * Pregnant women aged ≥18 years * Attending routine second-trimester scan (and scheduled transvaginal cervical assessment per local protocol/workflow) Exclusion Criteria: * Absence of transvaginal cervical assessment at the second-trimester scan * Missing follow-up data on pregnancy outcome (gestational age at delivery) * Inadequate image quality or missing required cervical ultrasound image

Outcomes

Primary Outcomes

Spontaneous preterm birth <37+0 weeks

Birth \<37+0 weeks after spontaneous onset of labor (with or without preterm prelabor rupture of membranes \[PPROM\]), regardless of mode of delivery, and excluding medically indicated (iatrogenic) preterm births without spontaneous onset.

Time frame: At delivery.

Secondary Outcomes

Spontaneous preterm birth <34+0 and <32+0 weeks