The Artificial Laparoscopic Instrument Dataset (ALID) is a fine-grained labelled dataset acquired on the BellySim lumpectomy simulation training system.
- In order to simulate multi-station clinical training for thoracoscopic and laparoscopic surgery in the form of an artificial pneumoperitoneum, two different stations have been designed.
- The training system consists of trocars, a 30-degree focusable endoscope, a pneumoperitoneum morphology experimental platform, scissors, grasping forceps, and detachment forceps.
The ALID dataset was collaboratively overseen by two Medical consultants(one with knowledge of and involvement in laparoscopic surgery and one with a research interest in medical image processing). Annotation was facilitated by six other researchers using Efficient Interactive Segmentation (EISeg) toolbox, a high-precision interactive annotation tool. In this case, a frame would be annotated by two annotators and a reviewer would do the initial review and revision. Finally, it was presented to two medical consultants (experts) for final review and revision.
Each frame is annotated with three categories: background, shaft, and manipulator.
The ALID dataset contains 1 250 images with the size of 1080 × 1920, involving six scenarios of occlusion, illumination imbalance, category imbalance, smoke, motion artefacts, and blood stains, also including their pair-wise combinations.
Here are the specific settings and collection procedures for the three scenarios:
We use a humidifier to generate a smoke effect, strategically positioning it at various locations. The humidifier operates continuously for one minute, allowing us to capture the smoke's pro-gression from its onset, through even dispersion, to gradual intensification.
Artificial blood plasma is used to simulate blood stains. We apply different concentrations of plasma to various parts of the artificial tissue and then move the surgical instruments in con-tact with the tissue to emulate the effect of staining.
Motion artifacts are simulated by rapidly moving the surgical instruments within the camera's field of view. The operator performs a variety of movements at different speeds and directions to capture a range of artifact effects, without the necessity for precise numerical consistency.
The corresponding number of categories is as follows:
| Category | Normal | Occlusion | Smoke | Blood Stains | Category Imbalance | Illumination Imbalance | Motion Artifacts |
|---|---|---|---|---|---|---|---|
| Normal | 150 | 120 | 60 | 60 | 130 | 120 | 60 |
| Occlusion | - | - | 20 | 20 | 60 | 60 | 60 |
| Smoke | - | - | - | 30 | 20 | 20 | 20 |
| Blood Stains | - | - | - | - | 20 | 20 | 20 |
| Category Imbalance | - | - | - | - | - | 60 | 60 |
| Illumination Imbalance | - | - | - | - | - | - | 60 |
For more information and access to datasets please contact us ([email protected])