# Particle-Scale Mineral Dataset – Imaging and Derived Data ## General Remark This dataset is directly associated with the following publication and contains the imaging data and derived datasets used for the presented analysis: **CNN-based 3D characterization and liberation analysis of lithium-bearing slag particles using correlative CT and SEM imaging** Tom Kirstein et al. The repository provides the underlying X-ray CT data and processed image stacks used for particle-scale characterization and evaluation of the proposed workflow. --- ## Abstract Efficient recovery of critical raw materials such as lithium from metallurgical slags requires optimized liberation of target phases during comminution. To determine effective mechanical process parameters for target phase recovery, an in-depth understanding of the characteristics of slag particles is crucial. For this purpose, modern tomography techniques, such as computed tomography (CT), can provide high-resolution 3D images of micrometer-sized slag particles. However, analysis of such CT images poses challenges, such as insufficient grayscale contrast between mineral phases and partial-volume effects. This work presents a scalable workflow for accurate phase- and particle-wise 3D characterization of particle systems by correlating 3D CT images with 2D mineral maps. High-resolution SEM slices are registered in 3D CT images and used as ground truth to train 3D convolutional neural networks (CNNs) for segmentation of individual particles and mineral phases. The approach enables particle-wise 3D characterization of complex slag systems with minimal manual labeling effort. The trained models are applied to particle systems with size fractions of 63–100 µm and 100–250 µm of a lithium-bearing slag with LiAlO₂ as the target phase. While virtual cross-sections show strong agreement with 2D validation data, the resulting 3D mineral liberation statistics differ significantly from conventional 2D estimates, which tend to overestimate liberation. By addressing this stereological bias, the workflow provides improved insights for optimizing pyrometallurgical and mechanical processing parameters for raw material recovery. --- This dataset contains particle-resolved imaging data and derived information for mineral systems across multiple particle size fractions. It is intended to support image-based analysis workflows, particle characterization, and method development in mineral processing and materials science. The repository includes raw and processed image data, mineral classification outputs, and supporting datasets used during analysis. --- ## Dataset Structure ``` /MLA │ ├── Datasources/ │ ├── SPP2315_GXMAP/ │ ├── Images/ │ ├── Mineral Lists/ │ └── 2022_10_04_LiSlag_Cindy/ │ └── SlagCharacterization/ ├── tiff_stitch_63-100µm/ └── tiff_stitch_100-250µm/ ``` --- ## Folder Description ### 1. Datasources This directory contains external or intermediate datasets used within the workflow. * **SPP2315_GXMAP/** Contains mapping or coordinate-based datasets used for spatial referencing or integration with imaging data. --- ### 2. Images Central storage for image data used throughout the analysis workflow. Typical contents may include: * Raw image acquisitions * Preprocessed image data * Segmented or labeled images The exact content depends on the processing stage. --- ### 3. Mineral Lists Contains mineralogical classification outputs and related metadata. * **2022_10_04_LiSlag_Cindy/** Dataset snapshot containing mineral classification results at a defined processing stage. Typical contents may include: * Phase lists * Lookup tables * Classification metadata --- ### 4. SlagCharacterization Contains processed image stacks used for particle-scale characterization. #### Size Fraction: 63–100 µm * **tiff_stitch_63-100µm/** Stitched TIFF image stacks representing reconstructed particle systems. #### Size Fraction: 100–250 µm * **tiff_stitch_100-250µm/** Equivalent datasets for a coarser particle size fraction. Each folder typically contains: * 2D TIFF slices or stitched volumes * Reconstructed or processed image stacks * Intermediate results from stitching or preprocessing --- ## X-ray CT Equipment and Acquisition Parameters All X-ray tomography scans were performed using a **Zeiss Xradia 510 VERSA** system. ### General Acquisition Settings * Voltage: 80 kV * Current: 87 µA * Exposure Time: 1.5 s * Optical Magnification: ~4× * Pixel Size: ~2.296 µm * Data Type: 16-bit (unsigned short) ### Dataset-Specific Parameters #### tiff_stitch_63-100µm * Particle size range: 63–100 µm * Image dimensions: 984 × 1011 px * Number of projections: 3566 * Detector-to-rotation-axis distance: 0.029002 * Source-to-rotation-axis distance: -0.015004 #### tiff_stitch_100-250µm * Particle size range: 100–250 µm * Image dimensions: 984 × 1010 px * Number of projections: 3275 * Detector-to-rotation-axis distance: 0.029004 * Source-to-rotation-axis distance: -0.015004 --- ## Data Characteristics * **Data Type:** Image-based particle datasets (TIFF) * **Scale:** Micrometer range * **Structure:** Slice-based or stitched volumetric data * **Content:** Mineral particles with varying morphology and composition --- ## Processing Context The dataset represents different stages of an image-based characterization workflow, which may include: * Image acquisition * Image stitching and reconstruction * Segmentation and particle separation * Feature extraction and classification Not all intermediate steps are necessarily included. --- ## Intended Use This dataset is suitable for: * Particle-scale image analysis * Mineral processing research * Development of segmentation and classification algorithms * Comparative studies across particle size fractions * Validation of image-based analysis workflows --- ## Notes * Image data may contain reconstruction or stitching artifacts. * Resolution and quality may vary between datasets. * Mineral classification outputs depend on the applied methodology. * Users should verify consistency between image data and metadata before quantitative analysis. --- ## Data prepared for Publication Further methodological details and analysis context are provided in the associated publication (if applicable). --- ## Contact For questions regarding the dataset, processing workflow, or reuse, contact the data provider.