Contrast-enhanced endoscopic ultrasound (CE-EUS) provides high-resolution assessment of hepatic microvascularity but requires contrast agents and remains largely qualitative (1,2). Detective flow imaging (DFI) is a novel EUS modality capable of visualizing low-velocity microvascular flow without contrast (3); however, its application in liver evaluation is limited. We aimed to evaluate a real-time AI-based vessel-detection algorithm for objective quantification of microvascular density to distinguish between healthy and pathological liver tissue.

A prospective study including consecutive adult patients undergoing EUS was performed between June 2024 and October 2025. The liver was assessed using DFI. Real-time CADe software analyzed live DFI images, detecting and counting fine vessels. Average fine vessel count (AVC) was defined as the mean vessel count detected by the CADe during the procedure. Ten external physicians, divided into senior (40%) and fellow (60%) endoscopists, evaluated 10 randomly selected EUS-DFI videos to provide vessel counts and categorical interpretation. Intra-class correlation (ICC) and interobserver agreement (IOA) were calculated.

Sixty participants were included (Pathological: 48; Healthy: 12). Mean age was 59.7 ± 13.6 and 52.0 ± 16.0 (p>.05), and 58.3% were female (p=.512). The most common pathology was alcoholic liver disease, and 68.8% of cases had mild liver disease (Child-Pugh A) (Table 1). CADe provided an AVC of 13.6 ± 5.6 in pathologic and 6.2 ± 7.2 for healthy liver (p<0.05). The maximum fine vessel count (MVC) for pathological liver was 49.0 ± 13.3 and 25.1 ± 16.8 for healthy liver (p<.05). Sensitivity, specificity, positive and negative predictive values, and observed agreement for distinguishing between pathological from healthy liver were 94%, 83%, 96%, 77%, and 92%, respectively.

Visual assessment showed poor ICC in both senior (-0.02, p=.545) and fellow (-0.01, p=.513). Additionally, categorical interpretation (healthy or pathological liver) achieved poor agreement (senior: k=0.05, p=.697; fellows: -0.01, p=.846).

AI-assisted DFI enables objective quantification of hepatic microvascularity and distinguishes pathological from healthy liver tissue. The AI-derived average vessel count demonstrated strong diagnostic performance. Human evaluation, regardless of expertise level, showed poor interobserver agreement and minimal reliability. Real-time AI-enhanced DFI vessel count is a quantitative alternative for liver microvascular assessment, with potential applicability when contrast agents are unavailable or contraindicated. Further validation against CE-EUS across broader liver disease severities are needed.