Pore REconstruction and Segmentation (PORES) method for improved porosity quantification of nanoporous materials

G. Van Eyndhoven, Mert Kurttepeli, Cynthia J. Van Oers, Pegie Cool, Sara Bals, Kees Joost Batenburg, Jan SijbersUltramicroscopy 2014 · DOI

Introduction

Nanoporous materials are vital components in modern chemical engineering, serving critical roles in catalysis, gas separation, and drug delivery. Optimizing these materials requires precise knowledge of their internal 3D structure, specifically the size and shape of their pores. While traditional nitrogen sorption experiments provide bulk statistics, they often rely on simplified geometric models that fail to capture the complexity of amorphous or irregular structures.

Electron tomography (ET) offers a direct 3D imaging alternative, but it faces significant technical challenges. The "missing wedge" problem—caused by the inability to rotate samples a full 180 degrees—introduces blurring and elongation artifacts. Furthermore, converting raw tomographic data into quantifiable pore statistics typically involves subjective manual thresholding, leading to inconsistent results. The PORES (POre REconstruction and Segmentation) method addresses these issues by integrating reconstruction and segmentation into a single, automated workflow.

The PORES Framework: SUPPRESS and Segmentation

The PORES method consists of two primary stages designed to minimize artifacts and automate quantification.

1. SUPPRESS Algorithm

The core of the PORES method is the SUPPRESS (Simultaneous Update of Pore Pixels by iterative REconstruction and Simple Segmentation) algorithm. Unlike standard reconstruction techniques like SIRT (Simultaneous Iterative Reconstruction Technique), which treat the entire volume uniformly, SUPPRESS incorporates prior physical knowledge: pores are empty spaces with zero intensity.

During reconstruction, the algorithm iteratively identifies voxels likely to be pores, sets them to zero, and focuses the mathematical reconstruction on the remaining material voxels. This "pinning" of pore interiors prevents intensity leakage and significantly reduces the blurring caused by the missing wedge.

2. Automated Segmentation and Quantification

Following reconstruction, the PORES pipeline applies an anisotropic diffusion filter to suppress noise while maintaining sharp edges. A modified watershed algorithm is then used to segment individual pores, even when they are interconnected or irregularly shaped. This allows for the calculation of the "equivalent spherical diameter" for each pore, providing a comprehensive statistical distribution without the need for manual intervention.

The PORES workflow, highlighting the iterative SUPPRESS reconstruction loop and the subsequent segmentation and quantification steps.

Validation and Key Results

The researchers validated the PORES method using synthetic phantoms (2D Perlin noise and 3D ellipsoids) and real-world experimental data from an amorphous mesoporous aluminosilicate sample.

Improved Accuracy and Robustness

In simulations, PORES achieved a Number of Misclassified Pixels (NMP) of approximately 4.1%, a significant improvement over SIRT (6.4%) and WBP (10.8%). Furthermore, the Relative Root Mean Square Error (RRMSE) remained consistently lower for PORES across various missing wedge sizes, demonstrating its robustness to limited data.

Real-World Application

When applied to experimental HAADF-STEM data of an aluminosilicate sample, PORES provided a much clearer distinction between the material and the carbon support grid compared to traditional methods. The resulting pore size distributions were stable across different volumes of interest (VOIs), proving the reliability of the automated pipeline.

Experimental cross-section of a nanoporous aluminosilicate sample, showing the Volume of Interest (VOI) used for PORES quantification.

Conclusion

The PORES method represents a significant advancement in the characterization of nanoporous materials. By treating reconstruction and segmentation as a joint problem, it effectively mitigates the missing wedge artifact and removes the subjectivity inherent in manual thresholding. While some initial parameters currently require manual selection, the workflow offers a more accurate and consistent path from raw microscopy data to reliable pore statistics. Future work may focus on fully automating parameter selection and integrating tomographic data with nitrogen sorption models for a more holistic characterization standard.

Summary updated

BibTeX
@inproceedings{eyndhoven2014pore,
  author = {G. Van Eyndhoven and Mert Kurttepeli and Cynthia J. Van Oers and Pegie Cool and Sara Bals and Kees Joost Batenburg and Jan Sijbers},
  title = {{Pore REconstruction and Segmentation (PORES) method for improved porosity quantification of nanoporous materials}},
  booktitle = {Ultramicroscopy},
  year = {2014},
  doi = {10.1016/j.ultramic.2014.08.008},
}
Figure