AI Demos · Educational concepts

See what Surgical AI tries to detect, and where its answers can break down.

These static concepts explain common tasks using published research patterns. They do not run live inference and are not clinical tools.

Demo controls

TaskSensitivity50%

Output console

Choose a task and run the simulation.
No video is uploaded or processed.

Educational demo based on published Surgical AI tasks. Not for clinical use.

Example output layer · 01

Concept 01Static walkthrough

Critical View of Safety Assessment

Clinical question: Have the required anatomical criteria been achieved?
Example output: Criteria-level assessment with supporting frames.
Why it matters: Could support structured review of a safety-critical step.
Limitations: Image quality, anatomy, and labels vary; an output is not a clinical decision.

Discuss this concept

Example output layer · 02

Concept 02Static walkthrough

Tool Detection

Clinical question: Which instruments are visible and when?
Example output: Instrument classes, locations, and confidence values.
Why it matters: Enables workflow analysis and can support downstream task recognition.
Limitations: Occlusion, smoke, blood, and unfamiliar tools can reduce reliability.

Discuss this concept

Example output layer · 03

Concept 03Static walkthrough

Phase Recognition

Clinical question: Where is the procedure within its workflow?
Example output: A timeline of predicted procedural phases.
Why it matters: Can organize long videos and support retrospective analysis.
Limitations: Procedures do not always follow a standard sequence.

Discuss this concept

Example output layer · 04

Concept 04Static walkthrough

Anatomical Segmentation

Clinical question: Which pixels correspond to relevant anatomy?
Example output: Color-coded masks over selected structures.
Why it matters: Makes spatial model outputs easier to inspect.
Limitations: Boundaries can be ambiguous and errors may appear visually plausible.

Discuss this concept

Example output layer · 05

Concept 05Static walkthrough

Workflow Analysis

Clinical question: How do events, tools, and phases relate over time?
Example output: A structured event timeline and transition map.
Why it matters: Helps researchers quantify variation and define study endpoints.
Limitations: Observed workflow does not explain clinical intent on its own.

Discuss this concept

Example output layer · 06

Concept 06Static walkthrough

Surgical Report Generation

Clinical question: Can recorded events support a draft structured summary?
Example output: A reviewable outline linked to observable events.
Why it matters: May reduce documentation friction when paired with human review.
Limitations: Video cannot capture every clinically relevant fact or decision.

Discuss this concept

Example output layer · 07

Concept 07Static walkthrough

Skill and Quality Assessment

Clinical question: Which observable features may relate to performance?
Example output: Task-specific metrics and review prompts.
Why it matters: Can make feedback more structured and reproducible.
Limitations: Skill is multidimensional; proxy metrics can be misleading or unfair.

Discuss this concept

Example output layer · 08

Concept 08Static walkthrough

Dataset Annotation Demo

Clinical question: How does a clinical definition become a machine-readable label?
Example output: Example boxes, masks, phases, and event labels.
Why it matters: Shows why annotation design is a clinical and technical task.
Limitations: A label set always simplifies reality and must be documented.

Discuss this concept

The Weekly Signal

Get the weekly Surgical AI signal.

A concise email with practical explainers, useful papers, dataset updates, tools, and ways to participate.

Read and subscribe