Surgical Tattoos in Infrared: A Dataset for Quantifying Tissue Tracking and Mapping

Area of Science:

• Medical Imaging
• Surgical Technology
• Computer Vision

Background:

• Accurate tissue tracking and mapping are crucial for image-guided surgery and automation.
• Existing datasets for endoscopic environments have limitations, including rigid setups, visible markers, or costly, error-prone manual annotation.
• There is a need for robust datasets that facilitate the quantitative evaluation of tracking and mapping methods in realistic surgical scenarios.

Purpose of the Study:

• Introduce a novel dataset, Surgical Tattoos in Infrared (STIR), for evaluating endoscopic tissue tracking and mapping methods.
• Provide a dataset with persistent, yet invisible to visible spectrum algorithms, labels for improved accuracy and generalizability.
• Enable quantitative analysis and benchmarking of various tracking and mapping algorithms in diverse surgical contexts.

Main Methods:

• Developed a novel labeling methodology using indocyanine green (ICG), an IR-fluorescent dye, to create persistent labels on tissue.
• Collected hundreds of stereo video clips (in vivo and ex vivo) using visible light cameras, with labels marked in the IR spectrum.
• The STIR dataset comprises over 3,000 labeled points, offering a comprehensive resource for algorithm evaluation.

Main Results:

• The STIR dataset was successfully created, featuring labels invisible to visible light algorithms but detectable in the IR spectrum.
• Multiple frame-based tracking methods were analyzed using the STIR dataset.
• Performance was evaluated using both 3D and 2D endpoint error and accuracy metrics, providing quantitative insights into algorithm performance.

Conclusions:

• The STIR dataset provides a valuable resource for quantifying the performance of tissue tracking and mapping methods in endoscopic settings.
• The novel labeling approach overcomes limitations of previous datasets, offering improved robustness and generalizability.
• STIR facilitates advancements in image guidance and automation for medical interventions and surgery.