Comparison of Methods to Segment Variable-Contrast XCT Images of Methane-Bearing Sand Using U-Nets Trained on Single Dataset Sub-Volumes
Methane (CH<sub>4</sub>) hydrate dissociation and CH<sub>4</sub> release are potential geohazards currently investigated using X-ray computed tomography (XCT). Image segmentation is an important data processing step for this type of research. However, it is often time consumi...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-12-01
|
| Series: | Methane |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2674-0389/2/1/1 |
| _version_ | 1797858375771357184 |
|---|---|
| author | Fernando J. Alvarez-Borges Oliver N. F. King Bangalore N. Madhusudhan Thomas Connolley Mark Basham Sharif I. Ahmed |
| author_facet | Fernando J. Alvarez-Borges Oliver N. F. King Bangalore N. Madhusudhan Thomas Connolley Mark Basham Sharif I. Ahmed |
| author_sort | Fernando J. Alvarez-Borges |
| collection | DOAJ |
| description | Methane (CH<sub>4</sub>) hydrate dissociation and CH<sub>4</sub> release are potential geohazards currently investigated using X-ray computed tomography (XCT). Image segmentation is an important data processing step for this type of research. However, it is often time consuming, computing resource-intensive, operator-dependent, and tailored for each XCT dataset due to differences in greyscale contrast. In this paper, an investigation is carried out using U-Nets, a class of Convolutional Neural Network, to segment synchrotron XCT images of CH<sub>4</sub>-bearing sand during hydrate formation, and extract porosity and CH<sub>4</sub> gas saturation. Three U-Net deployments previously untried for this task are assessed: (1) a bespoke 3D hierarchical method, (2) a 2D multi-label, multi-axis method and (3) RootPainter, a 2D U-Net application with interactive corrections. U-Nets are trained using small, targeted hand-annotated datasets to reduce operator time. It was found that the segmentation accuracy of all three methods surpass mainstream watershed and thresholding techniques. Accuracy slightly reduces in low-contrast data, which affects volume fraction measurements, but errors are small compared with gravimetric methods. Moreover, U-Net models trained on low-contrast images can be used to segment higher-contrast datasets, without further training. This demonstrates model portability, which can expedite the segmentation of large datasets over short timespans. |
| first_indexed | 2024-04-09T21:13:23Z |
| format | Article |
| id | doaj.art-c75915565af0487484e385a729a5c004 |
| institution | Directory Open Access Journal |
| issn | 2674-0389 |
| language | English |
| last_indexed | 2024-04-09T21:13:23Z |
| publishDate | 2022-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Methane |
| spelling | doaj.art-c75915565af0487484e385a729a5c0042023-03-28T14:14:10ZengMDPI AGMethane2674-03892022-12-012112310.3390/methane2010001Comparison of Methods to Segment Variable-Contrast XCT Images of Methane-Bearing Sand Using U-Nets Trained on Single Dataset Sub-VolumesFernando J. Alvarez-Borges0Oliver N. F. King1Bangalore N. Madhusudhan2Thomas Connolley3Mark Basham4Sharif I. Ahmed5Faculty of Engineering and Physical Sciences, University of Southampton, Southampton SO17 1BJ, UKDiamond Light Source Ltd., Harwell Science & Innovation Campus, Didcot OX11 0DE, UKFaculty of Engineering and Physical Sciences, University of Southampton, Southampton SO17 1BJ, UKDiamond Light Source Ltd., Harwell Science & Innovation Campus, Didcot OX11 0DE, UKThe Rosalind Franklin Institute, Harwell Science & Innovation Campus, Didcot OX11 0QS, UKDiamond Light Source Ltd., Harwell Science & Innovation Campus, Didcot OX11 0DE, UKMethane (CH<sub>4</sub>) hydrate dissociation and CH<sub>4</sub> release are potential geohazards currently investigated using X-ray computed tomography (XCT). Image segmentation is an important data processing step for this type of research. However, it is often time consuming, computing resource-intensive, operator-dependent, and tailored for each XCT dataset due to differences in greyscale contrast. In this paper, an investigation is carried out using U-Nets, a class of Convolutional Neural Network, to segment synchrotron XCT images of CH<sub>4</sub>-bearing sand during hydrate formation, and extract porosity and CH<sub>4</sub> gas saturation. Three U-Net deployments previously untried for this task are assessed: (1) a bespoke 3D hierarchical method, (2) a 2D multi-label, multi-axis method and (3) RootPainter, a 2D U-Net application with interactive corrections. U-Nets are trained using small, targeted hand-annotated datasets to reduce operator time. It was found that the segmentation accuracy of all three methods surpass mainstream watershed and thresholding techniques. Accuracy slightly reduces in low-contrast data, which affects volume fraction measurements, but errors are small compared with gravimetric methods. Moreover, U-Net models trained on low-contrast images can be used to segment higher-contrast datasets, without further training. This demonstrates model portability, which can expedite the segmentation of large datasets over short timespans.https://www.mdpi.com/2674-0389/2/1/1U-Netmethane hydratesmicrotomographysediment microstructuresemantic segmentation |
| spellingShingle | Fernando J. Alvarez-Borges Oliver N. F. King Bangalore N. Madhusudhan Thomas Connolley Mark Basham Sharif I. Ahmed Comparison of Methods to Segment Variable-Contrast XCT Images of Methane-Bearing Sand Using U-Nets Trained on Single Dataset Sub-Volumes Methane U-Net methane hydrates microtomography sediment microstructure semantic segmentation |
| title | Comparison of Methods to Segment Variable-Contrast XCT Images of Methane-Bearing Sand Using U-Nets Trained on Single Dataset Sub-Volumes |
| title_full | Comparison of Methods to Segment Variable-Contrast XCT Images of Methane-Bearing Sand Using U-Nets Trained on Single Dataset Sub-Volumes |
| title_fullStr | Comparison of Methods to Segment Variable-Contrast XCT Images of Methane-Bearing Sand Using U-Nets Trained on Single Dataset Sub-Volumes |
| title_full_unstemmed | Comparison of Methods to Segment Variable-Contrast XCT Images of Methane-Bearing Sand Using U-Nets Trained on Single Dataset Sub-Volumes |
| title_short | Comparison of Methods to Segment Variable-Contrast XCT Images of Methane-Bearing Sand Using U-Nets Trained on Single Dataset Sub-Volumes |
| title_sort | comparison of methods to segment variable contrast xct images of methane bearing sand using u nets trained on single dataset sub volumes |
| topic | U-Net methane hydrates microtomography sediment microstructure semantic segmentation |
| url | https://www.mdpi.com/2674-0389/2/1/1 |
| work_keys_str_mv | AT fernandojalvarezborges comparisonofmethodstosegmentvariablecontrastxctimagesofmethanebearingsandusingunetstrainedonsingledatasetsubvolumes AT olivernfking comparisonofmethodstosegmentvariablecontrastxctimagesofmethanebearingsandusingunetstrainedonsingledatasetsubvolumes AT bangalorenmadhusudhan comparisonofmethodstosegmentvariablecontrastxctimagesofmethanebearingsandusingunetstrainedonsingledatasetsubvolumes AT thomasconnolley comparisonofmethodstosegmentvariablecontrastxctimagesofmethanebearingsandusingunetstrainedonsingledatasetsubvolumes AT markbasham comparisonofmethodstosegmentvariablecontrastxctimagesofmethanebearingsandusingunetstrainedonsingledatasetsubvolumes AT sharifiahmed comparisonofmethodstosegmentvariablecontrastxctimagesofmethanebearingsandusingunetstrainedonsingledatasetsubvolumes |