---
# For reference on dataset card metadata, see the spec: https://github.com/huggingface/hub-docs/blob/main/datasetcard.md?plain=1
# Doc / guide: https://huggingface.co/docs/hub/datasets-cards
{}
---

# Dataset Card for Dataset Name

This dataset is a collection of colour pointcloud sequences, captured from 
## Dataset Details

### Dataset Description

<!-- Provide a longer summary of what this dataset is. -->



- **Curated by:** Connor Daly, Mechatronics in Medicine, Imperial College London

- **License:** CC BY 4.0 (Creative Commons Attribution)

### Dataset Sources [optional]

<!-- Provide the basic links for the dataset. -->

- **Repository:** [SpineAlign](https://github.com/condog101/SpineAlign)
- **Paper [optional]:** [Coming soon]



## Dataset Structure

The test sequences as used in our MICCAI paper were: [3wdFkKgv_20241015_164758, 9hY9tyUS_20241016_140027, 4e5qD5n5_20241023_090919, 4rCfpioZ_20241129_093024]

[More Information Needed]

## Dataset Creation

### Curation Rationale

This dataset was created with the motivation of developing markerless tracking systems for spine surgery.


[More Information Needed]

### Source Data

The source of the data is lumbar spine surgeries taking place at Humanitas Research Hospital.
#### Data Collection and Processing

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->

[More Information Needed]

#### Who are the source data producers?

The dataset was produced from the caseload of several neuro-surgeons working at Humanitas Research Hopsital (Rozzano). The cases are lumbar surgeries.
The .pcd files were captured by an Azure Kinect camera.

### Annotations [optional]

Further annotations are contained within the CT_DATA folder of the accompanying code repository.
#### Annotation process

The annotations were created in a multi-step process, first coarsely by the operating surgeon. 
This was used to roughly align a preoperative CT/MRI mesh with a reference frame for a particular sequence.
We then use the SpineAlign process, described in the paper, to iteratively deform and transform the mesh to better align with 
the reference frame. Then points in the sequence are labelled positive (1) if within 50 mm of the registered mesh, otherwise negative (0).
Following the reference frame alignment, this is repeated for other frames, using ICP to keep the deformed mesh aligned.


#### Who are the annotators?

The annotators are surgeons of Humanitas Research hospital in Milan, alongside researchers at the Hamlyn Centre, of Imperial College London.

#### Personal and Sensitive Information

No personal identifying information is contained in this dataset


## Citation [optional]

If using this dataset, please cite the original paper ("Towards Markerless Intraoperative Tracking of Deformable Spine Tissue")
**BibTeX:**

## Dataset Card Contact
cd1723@ic.ac.uk