Artificial Intelligence for the Intra-procedural Assessment of Uterine Artery Embolization

Background and Rationale.

Uterine fibroids affect up to 70-80% of women of reproductive age. Uterine artery embolization achieves technical success rates above 95% and symptom improvement in approximately 75-90% of patients; however, it is associated with a 20-30% cumulative risk of clinical failure or need for reintervention at 5 years. Current intra-procedural assessment of embolization adequacy is based on qualitative angiographic criteria (e.g., "5-10 heartbeats stasis," "pruned tree appearance"), which are subjective and operator-dependent. Emerging evidence suggests that achieving near-complete, rather than complete, flow stasis may reduce post-procedural pain, underscoring the need for quantitative and standardized assessment tools.

AQ-VERO is an internally developed software platform that performs quantitative time-to-density (TTD) analysis of angiographic images to objectively quantify uterine and fibroid perfusion in real time.

Objectives.

Primary Objective: To validate the accuracy and intra-/interobserver reproducibility of AQ-VERO TTD metrics in quantifying perfusion changes during uterine artery embolization.

Secondary Objectives: (a) To develop and internally validate an AI-based decision support model that uses AQ-VERO-derived metrics to identify predefined embolization endpoints; (b) To explore the correlation between intra-procedural TTD metrics and post-procedural clinical outcomes, including symptom improvement, early pain scores, and need for reintervention.

Study Design. This is an ambispective (includes retrospective and prospective follow-up), multicenter observational study including women undergoing uterine artery embolization for symptomatic uterine fibroids. Standardized angiograms will be acquired and analyzed with AQ-VERO to extract TTD perfusion parameters (e.g., time-to-peak, area under the curve, wash-in/wash-out characteristics). Operators will document conventional qualitative angiographic endpoints. Clinical and imaging follow-up will be collected according to institutional protocols.

Primary Objective:

• To evaluate whether the AI predictive model developed using AQ-VERO© metrics can predict the clinical outcome, defined as complete or significant resolution of fibroid-related symptoms.

Secondary Objectives:

* To correlate distinct TTD curve morphologies and AQ-VERO metrics with post-procedural pain. * To detect the presence of collateral or accessory arterial supply that may compromise embolization efficacy.

Significance. This study is expected to establish a quantitative and AI-augmented framework for intra-procedural embolization assessment during uterine artery embolization, potentially reducing variability and improving long-term clinical outcomes.