Deep-learning Predictors of Abdominal Aortic Aneurysm Enlargement in Type II Endoleaks After Endovascular Aortic Repair: The RADAR Study

RADAR is a retrospective multicenter observational study designed to develop and independently evaluate a three-dimensional deep-learning model for preoperative prediction of clinically relevant type II endoleak (T2EL) after endovascular aneurysm repair (EVAR).

The model will be developed using preoperative CTA data from designated development centers. Model fitting, hyperparameter tuning, and selection of operating thresholds will be performed exclusively within these centers using center-aware cross-validation. After completion of development, the model architecture, parameters, and operating thresholds will be frozen before evaluation on independent held-out test centers. Data from these centers will not be used for model development, tuning, or threshold selection.

The primary analysis will assess the ability of the frozen model to distinguish three predefined clinical outcome classes based on the occurrence of T2EL and subsequent aneurysm-sac behavior. This center-level separation between development and testing is intended to provide an estimate of model performance and generalizability across institutions and heterogeneous imaging protocols.

Secondary analyses will assess inter-center variability in performance, model calibration, and the incremental predictive value of clinical and anatomical variables. The newly developed model will also be compared with the original NornirNet model on a separate common dataset reserved specifically for this comparison.

Eligibility is defined by the inclusion and exclusion criteria below. Conditions that depend on the study outcome (occurrence of type II endoleak, competing endoleak types, and duration of imaging follow-up) are deliberately not treated as eligibility criteria; they are specified separately as analysis-set, outcome-classification and censoring rules, in order to avoid outcome-dependent selection bias. The minimum duration of imaging follow-up specified in the section "Outcome classification and follow-up requirement" is a requirement for valid assignment of the reference outcome class and applies only to the event-free classes, for which the absence of the outcome can be established only by a sufficient period of observation. It does not apply to patients in whom the outcome has already been documented.

Inclusion criteria

• Age ≥ 18 years at the time of the index procedure.
• Elective repair of an intact, fusiform, infrarenal abdominal aortic aneurysm (AAA), symptomatic or asymptomatic, with a maximum aneurysm diameter ≥50 mm in women and 55 mm in men, or repaired for symptoms or documented rapid growth (≥5 mm in 6 months or 10 mm in 12 months).
• Treatment by standard EVAR, defined as implantation of a commercially available bifurcated infrarenal stent-graft with proximal sealing in the infrarenal neck, below the lowermost renal artery, and without any of the following:

• fenestrated, branched or scallop-modified devices (F/BEVAR);
• parallel-graft techniques;
• adjunctive sac or side-branch embolization at or before the index procedure;
• endostapling endoanchor devices at the index procedure;
• endovascular aneurysm sealing systems (e.g. EVAS);
• proximal or distal extension cuffs placed during or after the index procedure;
• aorto-uni-iliac configurations and iliac branch devices;
• Index procedure performed between 1 January 2015 and 31 December 2025 at one of the participating centers, with follow-up data censored at the administrative cut-off date of 31 December 2026.
• Preoperative arterial-phase CTA available in digital (DICOM) format, acquired ≤ 6 months before the index procedure.
• Baseline postoperative CTA acquired between 1 and 3 months after the index procedure.
• All qualifying CTA examinations (preoperative, baseline and follow-up) acquired with a reconstructed slice thickness ≤ 2.5 mm, with anatomical coverage extending at minimum from the celiac trunk to the external iliac arteries, and including an arterial phase and a delayed (venous) phase.
• Clinical and imaging documentation sufficient to determine the occurrence and clinical course of type II endoleak (primary outcome), including at least one CTA subsequent to the baseline postoperative CTA. Incompleteness of other clinical variables does not preclude inclusion; such variables are recorded as missing and handled according to the prespecified analytical procedures.

Exclusion criteria

• Repair for a ruptured AAA, or emergent repair for contained rupture.
• Aneurysm morphology other than fusiform infrarenal AAA, including saccular aneurysm, penetrating aortic ulcer, pseudoaneurysm, infected aneurysm, aortic dissection
• Any previous open or endovascular aortic intervention, thoracic or abdominal, including previous hypogastric or inferior mesenteric artery embolization.
• Non-diagnostic imaging, defined as a CTA with incomplete anatomical coverage, or with beam-hardening, motion or other artifact of sufficient severity to preclude assessment of the aneurysm sac, the stent-graft sealing zones, the patency of sac-feeding branches, or the presence of endoleak, as judged by the reviewing investigator.
• Documented refusal of consent or documented objection to the research use of health-related data, where applicable. Where the competent ethics committee has granted a waiver of consent (including for deceased or untraceable patients, in Switzerland under Article 34 HRA), such patients are included in accordance with that authorization.

Handling of competing endoleak types (analysis-set and censoring rule) The occurrence of a type I, III, IV or V endoleak is not an eligibility criterion but a censoring or competing event, adjudicated as follows.

• Patients in whom a type I, III, IV or V endoleak is identified on the baseline (1-3 month) CTA are excluded from the primary analysis.
• Patients who develop a type I, III, IV or V endoleak during follow-up are censored at the date of that diagnosis, with the exception of a type Ia or Ib endoleak judged to be secondary to sac enlargement of ≥ 5 mm attributable to a pre-existing and documented type II endoleak.
• This attribution is made by 2 independent specialists (1 vascular surgeon and 1 radiologist) at the treating institution with disagreements resolved by the central core imaging laboratory at EOC.

Outcome classification and follow-up requirement (analysis-set rule)

The primary outcome is a three-class variable adjudicated over the entire available imaging follow-up. Each patient is assigned to exactly one of the following classes:

• Class 0 - no type II endoleak: no type II endoleak identified on the baseline (1-3 month) CTA or on any subsequent CTA.
• Class 1 - type II endoleak without significant sac growth (benign): a type II endoleak documented on the baseline CTA and/or on any follow-up CTA, with a maximum increase in aneurysm diameter of < 5 mm relative to the baseline postoperative CTA throughout the observation period, and without T2EL-related reintervention. A type II endoleak present on the baseline CTA that subsequently resolves spontaneously, without sac growth ≥ 5 mm and without reintervention, is classified in this class and not in Class 0.
• Class 2 - type II endoleak with significant sac growth (malignant): a type II endoleak associated with an increase in maximum aneurysm diameter ≥ 5 mm relative to the baseline postoperative CTA, or with a T2EL-related reintervention.

Because Class 2 requires two sequential events (appearance of the endoleak and subsequent sac growth), whereas Classes 0 and 1 are defined by the sustained absence of an event, the minimum duration of imaging follow-up required for a valid class assignment differs by class. This is a label-ascertainment (outcome-adjudication) requirement, applied to ensure the validity of the reference standard against which the model is evaluated. It is not an eligibility criterion. The minimum follow-up requirement applies only to event-free patients (Classes 0 and 1), whereas patients in whom a Class 2 event has already been documented are assigned to Class 2 irrespective of the total duration of follow-up. It is not applied in an outcome-dependent manner that would preferentially remove patients at higher risk of the outcome. Class 2 (malignant): no minimum duration of follow-up is required. Once sac growth ≥ 5 mm attributable to a type II endoleak, or a T2EL-related reintervention, has been documented, the class assignment is definitive irrespective of the total duration of follow-up. Excluding such patients on the basis of short follow-up would remove confirmed events and would introduce outcome-dependent selection.

• Class 1 (benign): a minimum of 24 months of imaging follow-up from the baseline postoperative CTA is required, with the last available CTA showing an increase in maximum aneurysm diameter of < 5 mm relative to baseline. Where the endoleak is first detected after the baseline CTA, at least 24 months of imaging follow-up from the date of first detection of the endoleak is additionally required.
• Class 0 (no endoleak): a minimum of 24 months of imaging follow-up from the baseline postoperative CTA is required, with no type II endoleak identified on any examination and an increase in maximum aneurysm diameter of < 5 mm relative to baseline.
• Indeterminate: patients who are event-free but who do not meet the minimum follow-up requirement for Class 0 or Class 1 are classified as indeterminate. Indeterminate patients are not assigned a reference class and are excluded from the primary validation analysis. They are retained in the study database, reported in the participant flow diagram, and described in the analysis of the source population.

All eligible patients enter the cohort at the baseline (1-3 month) postoperative CTA and are followed until the earliest of: the last available CTA, a T2EL-related secondary intervention, a censoring event as defined in the preceding section, death, or administrative censoring at 31 December 2026. Administrative censoring at 31 December 2026 is a common closing date applied to the entire cohort and not a deadline by which follow-up must already have ended: a patient still under observation on that date contributes person-time up to that date, whereas a patient whose follow-up ends earlier for any of the reasons listed above is censored at the earlier date. The number of patients censored for each reason will be reported.

• The primary analysis includes all eligible patients with a definitive class assignment (Class 0, 1 or 2) as defined above.
• Death, rupture or secondary intervention for a cause other than type II endoleak occurring before the minimum follow-up requirement is met are treated as competing events. Such patients are classified as indeterminate unless a Class 2 event has already been documented, in which case the Class 2 assignment stands.
• The distribution of follow-up duration (median, interquartile range, and proportion below 24 months) will be reported overall and by class.
• Among Class 2 patients, the cumulative distribution of the interval from the baseline CTA to first detection of the type II endoleak, and from first detection to documented sac growth ≥ 5 mm, will be reported. These distributions provide an empirical justification of the 24-month threshold and an estimate of the residual proportion of late events that the threshold may fail to capture.
• Prespecified sensitivity analyses will be reported alongside the primary analysis: (i) the minimum follow-up requirement for Classes 0 and 1 varied to 12, 18 and 36 months, with the resulting class counts and model performance reported at each threshold; (ii) an analysis in which indeterminate patients are assigned to Class 0 or Class 1 according to their status at the last available CTA, to quantify the effect of their exclusion; and (iii) an analysis restricted to patients with ≥ 36 months of follow-up.
• Sac diameter change (in mm, and as annualized change in mm/year) relative to the baseline postoperative CTA will additionally be recorded as a continuous secondary outcome. This permits a threshold-independent assessment of model output and limits the extent to which the dichotomy at 5 mm determines the reported results.

Reproducibility of the diameter measurement on which the 5 mm threshold depends will be quantified. A random sample of at least 10% of examinations will be measured independently by two readers, and the intraclass correlation coefficient together with the Bland-Altman limits of agreement for maximum aneurysm diameter will be reported. Cases in which the measured diameter change lies within the limits of agreement of the 5 mm threshold will be identified and examined in a sensitivity analysis.

Variables recorded rather than applied as eligibility criteria The following are documented as covariates and are not used to exclude patients, since retrospective adjudication would be unreliable and would preferentially remove patients with hostile anatomy, in whom type II endoleak behavior may differ.

• Adherence to the device-specific instructions for use (IFU) in force at the time of implantation, assessed retrospectively per device and reported as adherent, non-adherent or indeterminate, together with the specific IFU parameter(s) not met.
• Antiplatelet and anticoagulant therapy at baseline and during follow-up.
• Patency, number and diameter of sac-feeding branches (inferior mesenteric artery, lumbar arteries, accessory renal arteries, median sacral artery) on the preoperative CTA.
• Device manufacturer and model, iliac limb configuration.
• Sac thrombus volume and morphology, infrarenal neck anatomy, and maximum aneurysm diameter at baseline.