2DeteCT - A large 2D expandable, trainable, experimental Computed Tomography dataset for machine learning

Maximilian B. Kiss; Sophia B. Coban; K. Joost Batenburg; Tristan van Leeuwen; Felix Lucka

doi:10.1038/s41597-023-02484-6

2DeteCT - A large 2D expandable, trainable, experimental Computed Tomography dataset for machine learning

Maximilian B. Kiss^*, Sophia B. Coban, K. Joost Batenburg, Tristan van Leeuwen, Felix Lucka^*

^*Corresponding author for this work

Research output: Contribution to journal › Article › Academic › peer-review

Abstract

Recent research in computational imaging largely focuses on developing machine learning (ML) techniques for image reconstruction, which requires large-scale training datasets consisting of measurement data and ground-truth images. However, suitable experimental datasets for X-ray Computed Tomography (CT) are scarce, and methods are often developed and evaluated only on simulated data. We fill this gap by providing the community with a versatile, open 2D fan-beam CT dataset suitable for developing ML techniques for a range of image reconstruction tasks. To acquire it, we designed a sophisticated, semi-automatic scan procedure that utilizes a highly-flexible laboratory X-ray CT setup. A diverse mix of samples with high natural variability in shape and density was scanned slice-by-slice (5,000 slices in total) with high angular and spatial resolution and three different beam characteristics: A high-fidelity, a low-dose and a beam-hardening-inflicted mode. In addition, 750 out-of-distribution slices were scanned with sample and beam variations to accommodate robustness and segmentation tasks. We provide raw projection data, reference reconstructions and segmentations based on an open-source data processing pipeline.

Original language	English
Article number	576
Pages (from-to)	1-12
Number of pages	12
Journal	Scientific data
Volume	10
Issue number	1
DOIs	https://doi.org/10.1038/s41597-023-02484-6
Publication status	Published - 4 Sept 2023

Bibliographical note

Funding Information:
We deeply appreciate the help of Johannes Krauß with the visualizations for this paper and the help of Adrian Müller (SpineSave AG) providing the titanium prostheses screws for the OOD scans. We would like to thank Dr. Willem Jan Palenstijn for assisting with the computational methods. We are grateful to TESCAN-XRE NV, for their collaboration regarding the FleX-ray Laboratory. This work was supported by the Dutch Research Council (NWO, project numbers OCENW.KLEIN.285, 613.009.106, 639.073.506). The sponsors were not involved in the research and writing process.

Funding Information:
We deeply appreciate the help of Johannes Krauß with the visualizations for this paper and the help of Adrian Müller (SpineSave AG) providing the titanium prostheses screws for the OOD scans. We would like to thank Dr. Willem Jan Palenstijn for assisting with the computational methods. We are grateful to TESCAN-XRE NV, for their collaboration regarding the FleX-ray Laboratory. This work was supported by the Dutch Research Council (NWO, project numbers OCENW.KLEIN.285, 613.009.106, 639.073.506). The sponsors were not involved in the research and writing process.

Publisher Copyright:
© 2023, Springer Nature Limited.

Funding

We deeply appreciate the help of Johannes Krauß with the visualizations for this paper and the help of Adrian Müller (SpineSave AG) providing the titanium prostheses screws for the OOD scans. We would like to thank Dr. Willem Jan Palenstijn for assisting with the computational methods. We are grateful to TESCAN-XRE NV, for their collaboration regarding the FleX-ray Laboratory. This work was supported by the Dutch Research Council (NWO, project numbers OCENW.KLEIN.285, 613.009.106, 639.073.506). The sponsors were not involved in the research and writing process. We deeply appreciate the help of Johannes Krauß with the visualizations for this paper and the help of Adrian Müller (SpineSave AG) providing the titanium prostheses screws for the OOD scans. We would like to thank Dr. Willem Jan Palenstijn for assisting with the computational methods. We are grateful to TESCAN-XRE NV, for their collaboration regarding the FleX-ray Laboratory. This work was supported by the Dutch Research Council (NWO, project numbers OCENW.KLEIN.285, 613.009.106, 639.073.506). The sponsors were not involved in the research and writing process.

Funders	Funder number
Dutch Research Council (NWO)	OCENW.KLEIN.285, 613.009.106, 639.073.506

Keywords

Image Processing, Computer-Assisted
Laboratories
Machine Learning
Tomography, X-Ray Computed

Access to Document

10.1038/s41597-023-02484-6Licence: CC BY

s41597-023-02484-6Final published version, 1.95 MBLicence: CC BY

Cite this

@article{63dcbf0a52c04aff901a32c74d59359d,

title = "2DeteCT - A large 2D expandable, trainable, experimental Computed Tomography dataset for machine learning",

abstract = "Recent research in computational imaging largely focuses on developing machine learning (ML) techniques for image reconstruction, which requires large-scale training datasets consisting of measurement data and ground-truth images. However, suitable experimental datasets for X-ray Computed Tomography (CT) are scarce, and methods are often developed and evaluated only on simulated data. We fill this gap by providing the community with a versatile, open 2D fan-beam CT dataset suitable for developing ML techniques for a range of image reconstruction tasks. To acquire it, we designed a sophisticated, semi-automatic scan procedure that utilizes a highly-flexible laboratory X-ray CT setup. A diverse mix of samples with high natural variability in shape and density was scanned slice-by-slice (5,000 slices in total) with high angular and spatial resolution and three different beam characteristics: A high-fidelity, a low-dose and a beam-hardening-inflicted mode. In addition, 750 out-of-distribution slices were scanned with sample and beam variations to accommodate robustness and segmentation tasks. We provide raw projection data, reference reconstructions and segmentations based on an open-source data processing pipeline.",

keywords = "Image Processing, Computer-Assisted, Laboratories, Machine Learning, Tomography, X-Ray Computed",

author = "Kiss, {Maximilian B.} and Coban, {Sophia B.} and Batenburg, {K. Joost} and {van Leeuwen}, Tristan and Felix Lucka",

note = "Funding Information: We deeply appreciate the help of Johannes Krau{\ss} with the visualizations for this paper and the help of Adrian M{\"u}ller (SpineSave AG) providing the titanium prostheses screws for the OOD scans. We would like to thank Dr. Willem Jan Palenstijn for assisting with the computational methods. We are grateful to TESCAN-XRE NV, for their collaboration regarding the FleX-ray Laboratory. This work was supported by the Dutch Research Council (NWO, project numbers OCENW.KLEIN.285, 613.009.106, 639.073.506). The sponsors were not involved in the research and writing process. Funding Information: We deeply appreciate the help of Johannes Krau{\ss} with the visualizations for this paper and the help of Adrian M{\"u}ller (SpineSave AG) providing the titanium prostheses screws for the OOD scans. We would like to thank Dr. Willem Jan Palenstijn for assisting with the computational methods. We are grateful to TESCAN-XRE NV, for their collaboration regarding the FleX-ray Laboratory. This work was supported by the Dutch Research Council (NWO, project numbers OCENW.KLEIN.285, 613.009.106, 639.073.506). The sponsors were not involved in the research and writing process. Publisher Copyright: {\textcopyright} 2023, Springer Nature Limited.",

year = "2023",

month = sep,

day = "4",

doi = "10.1038/s41597-023-02484-6",

language = "English",

volume = "10",

pages = "1--12",

journal = " Scientific data",

issn = "2052-4463",

publisher = "Nature Research",

number = "1",

}

TY - JOUR

T1 - 2DeteCT - A large 2D expandable, trainable, experimental Computed Tomography dataset for machine learning

AU - Kiss, Maximilian B.

AU - Coban, Sophia B.

AU - Batenburg, K. Joost

AU - van Leeuwen, Tristan

AU - Lucka, Felix

N1 - Funding Information: We deeply appreciate the help of Johannes Krauß with the visualizations for this paper and the help of Adrian Müller (SpineSave AG) providing the titanium prostheses screws for the OOD scans. We would like to thank Dr. Willem Jan Palenstijn for assisting with the computational methods. We are grateful to TESCAN-XRE NV, for their collaboration regarding the FleX-ray Laboratory. This work was supported by the Dutch Research Council (NWO, project numbers OCENW.KLEIN.285, 613.009.106, 639.073.506). The sponsors were not involved in the research and writing process. Funding Information: We deeply appreciate the help of Johannes Krauß with the visualizations for this paper and the help of Adrian Müller (SpineSave AG) providing the titanium prostheses screws for the OOD scans. We would like to thank Dr. Willem Jan Palenstijn for assisting with the computational methods. We are grateful to TESCAN-XRE NV, for their collaboration regarding the FleX-ray Laboratory. This work was supported by the Dutch Research Council (NWO, project numbers OCENW.KLEIN.285, 613.009.106, 639.073.506). The sponsors were not involved in the research and writing process. Publisher Copyright: © 2023, Springer Nature Limited.

PY - 2023/9/4

Y1 - 2023/9/4

N2 - Recent research in computational imaging largely focuses on developing machine learning (ML) techniques for image reconstruction, which requires large-scale training datasets consisting of measurement data and ground-truth images. However, suitable experimental datasets for X-ray Computed Tomography (CT) are scarce, and methods are often developed and evaluated only on simulated data. We fill this gap by providing the community with a versatile, open 2D fan-beam CT dataset suitable for developing ML techniques for a range of image reconstruction tasks. To acquire it, we designed a sophisticated, semi-automatic scan procedure that utilizes a highly-flexible laboratory X-ray CT setup. A diverse mix of samples with high natural variability in shape and density was scanned slice-by-slice (5,000 slices in total) with high angular and spatial resolution and three different beam characteristics: A high-fidelity, a low-dose and a beam-hardening-inflicted mode. In addition, 750 out-of-distribution slices were scanned with sample and beam variations to accommodate robustness and segmentation tasks. We provide raw projection data, reference reconstructions and segmentations based on an open-source data processing pipeline.

AB - Recent research in computational imaging largely focuses on developing machine learning (ML) techniques for image reconstruction, which requires large-scale training datasets consisting of measurement data and ground-truth images. However, suitable experimental datasets for X-ray Computed Tomography (CT) are scarce, and methods are often developed and evaluated only on simulated data. We fill this gap by providing the community with a versatile, open 2D fan-beam CT dataset suitable for developing ML techniques for a range of image reconstruction tasks. To acquire it, we designed a sophisticated, semi-automatic scan procedure that utilizes a highly-flexible laboratory X-ray CT setup. A diverse mix of samples with high natural variability in shape and density was scanned slice-by-slice (5,000 slices in total) with high angular and spatial resolution and three different beam characteristics: A high-fidelity, a low-dose and a beam-hardening-inflicted mode. In addition, 750 out-of-distribution slices were scanned with sample and beam variations to accommodate robustness and segmentation tasks. We provide raw projection data, reference reconstructions and segmentations based on an open-source data processing pipeline.

KW - Image Processing, Computer-Assisted

KW - Laboratories

KW - Machine Learning

KW - Tomography, X-Ray Computed

UR - http://www.scopus.com/inward/record.url?scp=85169683931&partnerID=8YFLogxK

U2 - 10.1038/s41597-023-02484-6

DO - 10.1038/s41597-023-02484-6

M3 - Article

C2 - 37666897

AN - SCOPUS:85169683931

SN - 2052-4463

VL - 10

SP - 1

EP - 12

JO - Scientific data

JF - Scientific data

IS - 1

M1 - 576

ER -

2DeteCT - A large 2D expandable, trainable, experimental Computed Tomography dataset for machine learning

Abstract

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this