Identification of grain size in lanthanum-based materials using a computer program

Abstract

Lanthanum-based materials are widely used in advanced ceramics, oxides, metallic composites, and thin films due to their superior electrical, magnetic, mechanical, and thermal properties, which are strongly governed by microstructural characteristics, particularly grain size. Accurate and reliable grain size identification is therefore essential for understanding structure–property relationships and optimizing material performance. This review systematically examines computer-based approaches for grain size identification in lanthanum-based materials using scanning electron microscopy (SEM) images. Conventional image processing techniques implemented in MATLAB, including thresholding, morphological operations, and signal-processing methods based on the Radon transform, are discussed alongside modern Python-based machine learning and deep learning frameworks such as U-Net and the Segment Anything Model (SAM). The comparative analysis highlights the strengths and limitations of each approach in handling complex microstructures, low-contrast grain boundaries, and overlapping grains. The reviewed studies demonstrate that while MATLAB-based methods remain effective for relatively simple microstructures, deep learning–based segmentation provides superior robustness and accuracy for complex lanthanum-based systems. Overall, this review emphasizes the growing importance of computer-assisted and deep learning–based methodologies as reliable tools for quantitative grain size characterization in advanced lanthanum-containing materials.