SERV-CT: A disparity dataset from cone-beam CT for validation of endoscopic 3D reconstruction

This data set has been published in Medical Image Analysis under the title “SERV-CT: A disparity dataset from CT for validation of endoscopic 3D reconstruction.” The dataset contains 16 stereoendoscopic image pairs with reference anatomical segmentation derived from CT. Two different ex vivo porcine samples were imaged using the straight and 30° endoscopes from the original classic da Vinci™ Surgical System (Intuitive Inc., Sunnyvale, US)

A reference is provided by a CT scan from the O-arm™ system (Medtronic Inc, Dublin, Ireland). The scan contains both the anatomy and the endoscope, facilitating constrained manual alignment to provide pose of the viewed anatomical surface relative to the endoscope.

The reference is provided as rectified stereo pairs with corresponding disparity and depth as well as labels for overlap region and occlusions.

Two versions of this dataset have are incorporated. You can download a simplified version for validation (SERV-CT.zip) and a version including CT surfaces and software for manual alignment (SERV-CT-ALL.zip). You can also download the datasets evaluation code at: http://github.com/surgical-vision/servcttk