10.14489/vkit.2019.07.pp.003-009

DOI: 10.14489/vkit.2019.07.pp.003-009

Варфоломеев И. А., Якимчук И. В.
СЕГМЕНТАЦИЯ РЕНТГЕНОВСКИХ МИКРОТОМОГРАФИЧЕСКИХ ИЗОБРАЖЕНИЙ ГОРНЫХ ПОРОД С ОБУЧЕНИЕМ НА ДВУХМЕРНОЙ МИНЕРАЛЬНОЙ КАРТЕ
(c. 3-9)

Аннотация. Рассмотрена задача многоклассовой семантической сегментации трехмерного изображения – рентгеновской микротомограммы горной породы. При этом отдельные минералы или группы минералов считали классами, а истинная разметка известна для набора двухмерных сечений образца, которые предварительно пространственно совмещали с трехмерным микротомографическим изображением. Сегментацию изображения проводили как при помощи сверточной сети типа U-Net, так и ансамбля решающих деревьев, а также выбранного вручную набора признаков.

Ключевые слова: трехмерные изображения; обработка томографических изображений; сегментация изображений; сверточные сети; локальные бинарные шаблоны.

Varfolomeev I. A., Yakimchuk I. V.
3D MICRO-CT IMAGE SEGMENTATION OF ROCK SAMPLES, TRAINED USING 2D MINERAL MAPS
(pp. 3-9)

Abstract. We describe a method for multi-class segmentation of a 3D micro-CT image of a rock sample. We utilize a set of 2D mineral maps of the sample cross-sections as the ground truth labeling. Mineral maps were acquired with a scanning electron microscope, equip ped with energy dispersive X-Ray detectors and a special software suite – QEMSCAN. Each class represent a specific mineral or a group of minerals. Five ~4000x4000x4000 images of the Berea sandstone 8-mm miniplugs were acquired using laboratory micro-CT setup. Mineral maps and micro-CT images are spatially registered to each other using a method, in a rigid-body approximation. Next, we compensate minor smooth mutual geometrical distortions, using optical flow approach. We evaluate two perpixel classification methodologies: feature-based and convolutional neural network based. Feature-based approach utilizes LightGBM classifier and a set of features including rotation-invariant 3D local binary patterns, moving average, median, variance and etc. Classifier generates soft-segmentation, which is subsequently refined using graphcut technique. The training set for this method contains just two 40004000 slices. The second approach utilizes wellknown U-Net convolutional neural network architecture. For higher performance, we only use 3 parallel slices instead of full 3D receptive area. The training set contains 45 slices 4000x4000 each. As our first baseline method we use binary Otsu thresholding. Our second baseline is multiclass Naïve-Bayes classifier with a single feature – original image after bilateral filtering – this is roughly equivalent to the commonly used in practice “denoise and manually set thresholds” approach. Both LightGBM and U-Net demonstrate quality superior to the baseline approaches, but U-Net scores higher.

Keywords: 3D-images; Tomographic image processing; Image segmentation; Convolutional neural networks; Local Binary Patterns.

+ - Информация об авторах (About the Authors) Click to collapse

Рус

И. А. Варфоломеев (Московский научно-исследовательский центр «Шлюмберже», Московский физико-технический институт (национальный исследовательский университет), Москва, Россия) E-mail: Этот e-mail адрес защищен от спам-ботов, для его просмотра у Вас должен быть включен Javascript
И. В. Якимчук (Московский научно-исследовательский центр «Шлюмберже», Москва, Россия)

Eng

I. A. Varfolomeev (Schlumberger Moscow Research Center, Moscow Institute of Physics and Technology, Moscow, Russia) E-mail: Этот e-mail адрес защищен от спам-ботов, для его просмотра у Вас должен быть включен Javascript
I. V. Yakimchuk (Schlumberger Moscow Research Center, Moscow, Russia)

+ - Библиографический список (References) Click to collapse

Рус

1. Integrated Study of Thin Sections: Optical Petrography and Electron Microscopy / I. Varfolomeev et al. // SPE Russian Petroleum Technology Conf. and Exhibition. Moscow, 24 – 26 Oct., 2016. Moscow, 2016. P. 1398 – 1412.
2. High Resolution X-ray Fluorescence MicroTomography on Single Sediment Particles / L. Vincze et al. // Proc. of SPIE – The International Society for Optical Engineering. 2001. V. 4503. P. 240 – 248.
3. An Experimental Study of DualEnergy CT Imaging Using Synchrotron Radiation / J. Hao et al. // Nuclear Science and Techniques. 2013. V. 24, Is. 2. P. 7 – 11.
4. Burdet P., Croxall S. A., Midgley P. A. Enhanced Quantification for 3D SEM–EDS: Using the Full Set of Available X-ray Lines // Ultramicroscopy. 2015. V. 148. P. 158 – 167. doi: 10.1016/j.ultramic.2014.10.010
5. 3D Microporosity Mapping of Rock Samples by X-ray MicroCT with Contrast Agents / I. V. Yakimchuk et al. // Bruker micro-CT User Meeting. Luxembourg, 2016. P. 22 – 24.
6. Varfolomeev I., Yakimchuk I., Sharchilev B. Segmentation of 3D Image of a Rock Sample Supervised by 2D Mineralogical Image // 2015 3rd IAPR Asian Conf. on Pattern Recognition (ACPR). 3 – 6 Nov. 2015. Kuala Lumpur, Malaysia, 2015. P. 346 – 350. doi: 10.1109/ACPR.2015.7486523
7. Liu C. Beyond Pixels: Exploring New Representations and Applications for Motion Analysis: Thesis (PhD) / Massachusetts Institute of Technology. 2009. 164 p.
8. Mutina A., Koroteev D. Using X-Ray Microtomography for the Three Dimensional Mapping of Minerals // Microscopy and Analysis. V. 26, Is.2. 2012. P. 7 – 12.
9. Assessment of Bone Ingrowth into Porous Biomaterials Using MICRO-CT / A. C. Jones et al. // Biomaterials. 2007. V. 28, Is. 15. P. 2491 – 2504. doi: 10.1016/j.biomaterials.2007.01.046
10. Processing of Rock Core Microtomography Images: Using Seven Different Machine Learning Algorithms / Chauhan S. et al. // Computers & Geosciences. 2016. V. 86. P. 120 – 128. doi: 10.1016/j.cageo.2015.10.013
11. The Usage of Modern Data Science in Segmentation and Classification: Machine Learning and Microscopy / M. Andrew et al. // Proc. of Microscopy and Microanalysis. 2017. V. 23, Is. S1. P. 156–157. doi: 10.1017/S1431927617001465
12. Wang Y., Lin C. L., Miller J. D. Improved 3D Image Segmentation for X-ray Tomographic Analysis of Packed Particle Beds // Minerals Engineering. 2015. V. 83. P. 185 – 191. doi: 10.1016/j.mineng.2015.09.007
13. Trainable Weka Segmentation: a Machine Learning Tool for Microscopy Pixel Classification / I. Arganda-Carreras et al. // Bioinformatics. 2017. V. 33, Is. 15. P. 2424 – 2426.
14. Ilastik: Interactive Learning and Segmentation Toolkit / C. Sommer et al. // Proc. of IEEE Inter. Symposium on Biomedical Imaging: from Nano to Macro. 2011. P. 230 – 233. doi: 10.1109/ISBI.2011.5872394
15. LightGBM: A Highly Efficient Gradient Boosting Decision Tree / G. Ke et al. // Advances in Neural Information Processing Systems (NIPS’2017). 2017. P. 3149 – 3157.
16. 3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation / Ö. Çiçek et al. 2016. 8 p. URL: https://arxiv.org/pdf/1606.06650.pdf (дата обращения: 12.03.2019)
17. Local Binary Patterns for Still Images / M. Pietikäinen et al. // Computer Vision Using Local Binary Pattern. Springer, 2011. V. 40. P. 13 – 47.
18. Lai M. Deep Learning for Medical Image Segmentation. 2015. 23 p. URL: https://arxiv.org/pdf/1505.02000.pdf (дата обращения: 12.03.2019)

Eng

1. Varfolomeev I. et al. (2016). Integrated Study of Thin Sections: Optical Petrography and Electron Microscopy, pp. 1398 – 1412. SPE Russian Petroleum Technology Conference and Exhibition 2016. Society of Petroleum Engineers.
2. Vincze L. et al. (2001). High resolution X-ray fluorescence microtomography on single sediment particles, Vol. 4503, pp. 240 – 248.Proceedings of SPIE – The International Society for Optical Engineering.
3. Hao J. et al. (2013). An experimental study of dualenergy CT imaging using synchrotron radiation. Nuclear Science and Techniques, (2), pp. 7 – 11.
4. Burdet P., Croxall S. A., Midgley P. A. (2015). Enhanced quantification for 3D SEM–EDS: Using the full set of available X-ray lines. Ultramicroscopy, Vol. 148, pp. 158 – 167. doi:10.1016/j.ultramic.2014.10.010
5. Yakimchuk I. V. et al. (2016). 3D Microporosity Mapping of Rock Samples by X-ray MicroCT with Contrast Agents, pp. 22 – 24. Bruker microCT User Meeting. Luxembourg.
6. Varfolomeev I., Yakimchuk I., Sharchilev B. (2015). Segmentation of 3D image of a rock sample supervised by 2D mineralogical image, pp. 346 – 350. 2015 3rd IAPR Asian Conference on Pattern Recognition (ACPR). Kuala Lumpur, Malaysia. doi: 10.1109/ACPR. 2015.7486523
7. Liu C. (2009). Beyond pixels: exploring new representations and applications for motion analysis. Massachusetts Institute of Technology.
8. Mutina A., Koroteev D. (2012). Using X-Ray Microtomography for the Three Dimensional Mapping of Minerals. Microscopy and Analysis.
9. Jones A. C. et al. (2007). Assessment of bone ingrowth into porous biomaterials using MICRO-CT. Biomaterials, Vol. 28(15), pp. 2491 – 2504. doi: 10.1016/j.biomaterials.2007.01.046
10. Chauhan S. et al. (2016). Processing of rock core microtomography images: Using seven different machine learning algorithms. Computers & Geosciences, Vol. 86, pp. 120 – 128. doi:10.1016/j.cageo. 2015.10.013
11. Andrew M., Bhattiprolu S., Butnaru D., Correa J. (2017). The Usage of Modern Data Science in Segmentation and Classification: Machine Learning and Microscopy. Microscopy and Microanalysis, Vol. 23(S1), pp. 156–157.
12. Wang Y., Lin C. L., Miller J. D. (2015). Improved 3D image segmentation for X-ray tomographic analysis of packed particle beds. Minerals Engineering, Vol. 83, pp. 185 – 191.
13. Arganda-Carreras I. et al. (2017). Trainable Weka Segmentation: a machine learning tool for microscopy pixel classification. Bioinformatics, Vol. 33(15), pp. 2424 – 2426.
14. Sommer C. et al. (2011). Ilastik: Interactive learning and segmentation toolkit, pp. 230 – 233. 2011 IEEE International Symposium on Biomedical Imaging: from Nano to Macro.
15. Ke G. et al. (2017). LightGBM: A highly efficient gradient boosting decision tree, pp. 3149 – 3157. Advances in Neural Information Processing Systems.
16. Çiçek Ö. et al. (2016). 3D U-Net: learning dense volumetric segmentation from sparse annotation. arXiv preprint arXiv:1606.06650.
17. Pietikäinen M. et al. (2011). Local Binary Patterns for Still Images. SpringerLink, pp. 13 – 47.
18. Lai M. (2015). Deep Learning for Medical Image Segmentation. arXiv:1505.02000 [cs].

+ - Заказать электронную версию статьи (Purchase digital version of a single article) Click to collapse

Рус

Статью можно приобрести в электронном виде (PDF формат).

Стоимость статьи 350 руб. (в том числе НДС 18%). После оформления заказа, в течение нескольких дней, на указанный вами e-mail придут счет и квитанция для оплаты в банке.

После поступления денег на счет издательства, вам будет выслан электронный вариант статьи.

Для заказа скопируйте doi статьи:

10.14489/vkit.2019.07.pp.003-009

и заполните форму

Отправляя форму вы даете согласие на обработку персональных данных.

Eng

This article is available in electronic format (PDF).

The cost of a single article is 350 rubles. (including VAT 18%). After you place an order within a few days, you will receive following documents to your specified e-mail: account on payment and receipt to pay in the bank.

After depositing your payment on our bank account we send you file of the article by e-mail.

To order articles please copy the article doi:

10.14489/vkit.2019.07.pp.003-009

and fill out the form