iMED Challenge 2026

Multi-Endoscope Dataset for Surgical 3D Perception

The iMED Challenge is the first benchmark for 3D surgical scene understanding using synchronized multi-endoscope data, with official tasks in pose estimation and deformable novel view synthesis for robot-assisted minimally invasive surgery.

Read Participation Guide Explore Dataset Details
Dataset figure placeholder

Description

Vision-based 3D reconstruction is critical for real-time guidance systems, autonomous camera control, and digital twins in robot-assisted minimally invasive surgery. Surgical environments remain especially challenging because of constrained viewpoints through narrow trocar ports, tissue deformation from respiratory motion and tool interactions, non-Lambertian surfaces with strong specularities, and transient phenomena such as smoke and bleeding.

Objective

Benchmark robust surgical 3D perception using synchronized multi-endoscope data in deformable, clinically realistic scenes.

Motivation

Existing surgical vision datasets capture scenes from a single endoscope along a single trajectory, making it difficult to distinguish true geometric reconstruction from photometric interpolation along a narrow forward-facing path. The iMED Challenge addresses this limitation by evaluating methods on synchronized multi-endoscope data with held-out viewpoints and cross-trajectory validation.

Tasks

Task 1: iMED PE

Metric: ATE

Estimate accurate per-frame relative camera poses between two endoscopes in deformable surgical environments.

69 train 21 test

Task 2: iMED NVS

Metrics: PSNR, SSIM

Synthesize novel views of deformable surgical scenes from a held-out endoscope viewpoint using a train-on-one-endoscope, test-on-another protocol.

20 sequences

Key Dates

  • April 15, 2026 Registration opens and the dataset is released on Synapse.
  • October 2026 (TBA) Challenge Day at MICCAI 2026 / EndoVis.

Organizers

Name Institution
Sierra Bonilla University College London
John Han Vanderbilt University
Tianyi Song University College London
Adam Schmidt Intuitive Surgical, Inc.
Omid Mohareri Intuitive Surgical, Inc.
Francisco Vasconcelos University College London
Sophia Bano University College London

Sponsors

References