Surgical Proficiency AI Video Analysis of Surgical Skill

Video-based assessment of surgical proficiency

Record a suturing or knot-tying task on an ordinary phone and this platform turns it into objective measures of technical skill. No sensors, no instrumented gloves, no special hardware. Upload the video and a full kinematic report is ready in minutes.

Log in to get started Create an account

What this project does

Technical skill in surgery is usually judged by an expert watching the trainee, which is slow, subjective, and hard to standardise. Objective motion analysis has typically required wearable sensors or instrumented equipment.

This platform instead tracks 21 landmarks on each hand directly in the video. Wrist motion is calibrated from pixels to meters using finger anatomy and distilled into 15 validated kinematic metrics per hand across three domains of technical skill:

Economy of Motion Spatial Organization Flow of Motion

Every result is compared with an attending surgeon reference, so trainees can see where their technique differs and follow their improvement over time.

Demo: AI hand tracking

The overlay shows both hands detected and labelled (R / L) in every frame of a knot-tying task. The wrist trajectory is the signal all metrics are computed from.

Results

Hand working area plot with 3-sigma confidence ellipses

The working-area plot summarises each hand's trajectory with a 3σ confidence ellipse; tighter, well-separated ellipses point to compact, organised technique. The full report also covers path length, movement count, speed, motion consistency, smoothness (LDLJ), and steadiness (SPARC), each shown next to the instructor reference.

100%
hand detection
2.0 m
total path length
11
distinct movements
101 mm/s
mean hand speed

Real output from this platform for the 10-second demo clip above.

How it works

  1. 1Film the task on any phone (1080p or 720p at 30 fps is plenty) and drag the file into the portal. Large uploads are optimised automatically.
  2. 2The AI model finds both hands in each frame and follows the wrists through the whole task.
  3. 3Path length, working area, movement count, motion consistency, smoothness, and steadiness are computed for each hand.
  4. 4An interactive dashboard shows every metric next to the reference surgeon, with one-click PDF and Excel export.

Download recording guide (PDF) Read the research paper Peer-reviewed · SAGE