Endoscopic ultrasound · Upper-GI subepithelial lesions

Find the lesion. Differentiate with clarity.

AI-assisted identification and differentiation of upper gastrointestinal subepithelial lesions, distinguishing leiomyoma from gastrointestinal stromal tumors in live and previously acquired EUS images.

UPPER GI · ENDOSCOPIC ULTRASOUND LIVE AI
EUS LESION CAPTURE
SEL DETECTED
BOUNDARY LOCKED
DIFFERENTIATION
LEIOMYOMA31%
GIST78%
CLINICIAN VALIDATION
AI ANALYSIS
Diagnostic accuracy>95%
Analysis modeReal-time · Post-procedural
Clinical supportDetection · Differentiation · Sampling

01 · Intended use

AI assistance for upper‑GI subepithelial lesions.

Sonomind® is a deep-learning technology developed to identify and differentiate subepithelial lesions of the upper gastrointestinal tract during endoscopic ultrasound.

It distinguishes leiomyoma from gastrointestinal stromal tumors in live and acquired EUS images, supports targeted sampling and generates standardized AI-driven reports.

AI
01Endoscopic ultrasoundLive or acquired EUS images
02Lesion intelligenceLeiomyoma–GIST differentiation
03Guided outputSampling target and structured report

The clinical challenge

Different lesions can look remarkably similar on EUS.

Leiomyomas and gastrointestinal stromal tumors may share overlapping sonographic characteristics. Clear differentiation can guide the need, location and urgency of tissue sampling.

“The critical step is converting a subtle subepithelial image into a focused diagnostic pathway.”
02

How it works

One continuous path from ultrasound to targeted sampling.

Sonomind® supports the EUS workflow while keeping sampling strategy and final interpretation in expert hands.

01

EUS acquisition

Live or previously acquired endoscopic ultrasound images enter the analysis workflow.

02

Lesion localization

The AI identifies an upper-GI subepithelial lesion and tracks it across EUS frames.

03

Layer and texture analysis

The model assesses lesion boundaries, echogenic texture and layer of origin.

04

Lesion differentiation

The finding is differentiated between leiomyoma and gastrointestinal stromal tumor.

05

Report and sampling

Bounding-box guidance and a standardized AI report support targeted tissue sampling.

AI capabilities

Designed for focused subepithelial‑lesion assessment.

>95%

Diagnostic accuracy presented for Sonomind®

01

Subepithelial-lesion identification

Automatically identifies upper-GI subepithelial lesions in endoscopic ultrasound images.

02

Leiomyoma–GIST differentiation

Analyzes sonographic patterns to assist distinction between leiomyoma and gastrointestinal stromal tumors.

03

Real-time and retrospective analysis

Works during live EUS and with previously acquired examinations.

04

Sampling-target guidance

Places bounding boxes around relevant tissue to support precise sampling.

05

Standardized AI reporting

Generates a structured AI-driven diagnostic report for expert validation.

03

Scientific validation

A multicentric transatlantic EUS study.

Scientific presentation · ACG 2025

Deep Learning for Detection and Differentiation of Subepithelial Lesions: A Multicentric Transatlantic Endoscopic Ultrasound Study

American College of Gastroenterology Annual Scientific Meeting · 2025

View study presentation ↗

Clinical benefits

Earlier clarity. Better‑targeted sampling.

01

Earlier and more accurate diagnosis of subepithelial lesions

02

Fewer unnecessary procedures through more focused decision-making

03

Target selection for sampling with bounding-box guidance

04

Real-time and post-procedural EUS analysis in one workflow

05

Standardized AI-driven diagnostic reporting

Sonomind®

See beneath the surface. Differentiate with clarity.

Talk to our team

This technology is part of DigestAID's research and development pipeline. Availability and regulatory status may vary by market.