Datasets


NBI-InfFrames

The NBI-InfFrames dataset aims to provide the surgical data science community with a labeled dataset for the identification of informative endoscopic video frames.

It is composed of 720 video frames. The frames are manually extracted and labeled from 18 narrow-band imaging (NBI) laryngoscopic videos of 18 different patients affected by laryngeal spinocellular carcinoma (diagnosed after histopathological examination).

The frames include 180 informative (I) video frames, 180 blurred (B) frames, 180 frames with saliva or specular reflections (S) and 180 underexposed (U) frames.

To download the dataset, please go to: https://zenodo.org/record/1162784#.WnFzLZOdX6Y.

If you use this dataset, please cite: S. Moccia, et al. “Learning-based classification of informative laryngoscopic frames.” COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE (accepted for publication) 

For further information, please contact sara.moccia@polimi.it.


Nephrec9

“Nephrec9” dataset contains frames of 14 steps of Robot-Assisted Partial Nephrectomy (RAPN) surgery. “Nephrec9”dataset is divided into testing, training and validation sets from 9 full video annotations of RAPN, annotated by an expert renal surgeon. The videos were divided into small videos of 30 seconds or 720 frames, processed at 24 FPS to capture the relevant step context for making predictions and to divide them manually into different datasets. We extracted a total of 1262 videos, out of which we used 769 (approx. 60%) for training, 372 (approx. 30%) for validation and 121 (approx. 10%) as a test dataset.

RAPN consists 14 steps:

  1. mobilization
  2. dissection
  3. identification
  4. ultrasound
  5. marking
  6. clamping
  7. resection
  8. midollar suturing
  9. cortical suturing
  10. unclamping
  11. inspection
  12. removal
  13. reconstruction
  14. drainage

If you use this dataset in your research, kindly please contact to Hirenkumar Nakawala to know more about this research and submitted work to a journal.

The dataset could be requested at: https://zenodo.org/record/1066831#.WmtRmIjOVPY


Laryngeal

The Laryngeal Dataset  aims to provide the computer assisted surgery community with a dataset for the validation of cancerous tissue classification algorithms.

It is composed of 1320 patches of healthy and early-stage cancerous laryngeal tissues extracted from 33 narrow-band laryngoscopic images of 33 different patients affected by laryngeal spinocellular carcinoma (diagnosed after histopathological examination).

To download the dataset, please go to: https://zenodo.org/record/1003200#.WdeQcnBx0nQ.

If you use this dataset, please cite:

Moccia, Sara, et al. “Confident texture-based laryngeal tissue classification for early stage diagnosis support.” JOURNAL OF MEDICAL IMAGING 4.03 (2017): 1-10

For further information, please contact sara.moccia@polimi.it.

Four patches, relative to the four analyzed laryngeal tissue classes, are extracted from the laryngoscopic image. Blue: tissue with intraepithelial papillary capillary loop-like vessels (IPCL); Red: tissue with Leukoplakia (Le); Green: healthy tissue (He); Yellow: tissue with hypertrophic vessels (Hbv). Figure modified from Moccia et al., 2017.

EndoAbS

The EndoAbS Dataset (Endoscopic Abdominal Stereo Images Dataset) aims to provide the computer assisted surgery community with a dataset for the validation of 3D reconstruction algorithms.

It is composed of:

  • 120 pairs of endoscopic stereo images of abdominal organs (liver, kidney, spleen);
  • corresponding ground truth in left­camera reference frame, generated using a laser scanner;
  • camera calibration parameters;

The images were captured under different conditions:

  • 3 different light levels;
  • presence of smoke;
  • two phantom­endoscope distances (~5cm or ~10cm);

To download the dataset, please goto: https://zenodo.org/record/60593

If you use this dataset, please cite:

Penza, V., Ciullo, A. S., Moccia, S., Mattos, L. S., & De Momi, E. (2018). EndoAbS dataset: Endoscopic abdominal stereo image dataset for benchmarking 3D stereo reconstruction algorithms. The International Journal of Medical Robotics and Computer Assisted Surgery, 14(5), e1926.

For further information, please contact veronica.penza@polimi.it.

VeronicaThesisImage1

TrackVes

The TrackVes dataset provides to the computer assisted surgery community a dataset for the validation of soft tissue tracking algorithms.

It is composed of:

– 3 video sequences of ex-vivo organs (kidney and liver);
– 6 video sequences of in-vivo organs (real abdominal surgical scenes);

The C++ code implementing the ground truth generation and the tracking algorithm evaluation is available on Bitbucket –
https://bitbucket.org/ververo/envisors_track/ as:app/gt_generatorapp/tracking_evaluation.

To download the dataset, please go to: https://zenodo.org/record/822053

If you use this dataset, please cite:

V. Penza, X. Du, D. Stoyanov, A. Forgione, L. Mattos and E. De Momi. “Long Term Safety Area Tracking (LT-SAT) with Online Failure Detection and Recovery for Robotic Minimally Invasive Surgery”, Medical Image Analysis, 2017.

For further information, please contact veronica.penza@iit.it


Find Us

NEARLab is located inside the Leonardo Robotics Labs space at Politecnico di Milano, piazza Leonardo da Vinci 32, Building 7, 20133, Milano, Italy
and at Campus Colombo in Via Giuseppe Colombo, 40, 20133 Milano MI

Hours

Monday to Friday: 8.00 A.M. – 20.00 P.M.


More


Website Maintainers
Alberto Rota, Mattia Magro, Alessandra Maria Trapani

Search


Get in touch

or visit the Research Areas and contact the corresponding team directly


Connections

Materials