The dataset contains multimedia recorded on cadaver and phantom (plastic model) tissue in relation to the PhD thesis "Robust and dense Visual SLAM for robot-assisted minimally invasive orthopedic procedures". Related publications can be consulted here.

The dataset contains timestamped recordings of two video sources (arthroscopic camera and external camera), robotic motion data and motion capture data. The data can be used to develop and evaluate assistive systems and algorithms (e.g. SLAM, SfM, visual servoing) for challenging minimally invasive orthopedic procedures.

Data file types include .png images and .txt files for calibration and ground truth data.

The related PhD thesis developed a vision-based robotic surgical assistant for minimally invasive orthopedic procedures. The system is composed of a robotic arm with an attached camera-arthroscope bundle for intra-articular navigation. The system is capable of a) localizing instruments robustly and reliably inside the human joints and b) generating dense and accurate 3D reconstructed models of the knee joint from intra-articular images. Thanks to these capabilities the system would allow for the semi-autonomous navigaton of the camera (via visual servoing) to follow the surgeons’ tools.

Data acquisition was approved by the Australian National Health and Medical Research Council (NHMRC) – Registered Committee Number EC00171 under Approval Number 1400000856.