Automatic microscopic image analysis by moving window local Fourier Transform and Machine Learning.

Area of Science:

• Materials Science
• Computational Biology
• Physics

Background:

• Manual analysis of microscopic images is time-consuming and prone to human bias.
• Automated methods are needed to improve efficiency and objectivity in image analysis.

Purpose of the Study:

• To develop an automated approach for analyzing microscopic images using local Fourier Transforms and machine learning.
• To eliminate human bias in image analysis and enable rapid feature discovery.

Main Methods:

• A local moving window scans the image, calculating 2D Fourier Transforms for each window.
• Machine learning techniques, including Principal Component Analysis (PCA) and Non-Negative Matrix Factorization (NMF), are applied to the local Fourier Transforms.
• The method decomposes data into spatial maps and Fourier Transform factors for feature identification.

Main Results:

• The automated approach successfully analyzed various microscopic images, including High Angle Annular Dark Field (HAADF) Scanning Transmission Electron Microscopy (STEM), Scanning Tunneling Microscopy (STM), Scanning Electron Microscopy (SEM), and fluorescence microscopy.
• Features were automatically discovered based on local Fourier Transform changes without human bias.
• Analysis of a single image takes approximately one minute on a standard computer.

Conclusions:

• The proposed method offers an efficient and unbiased solution for microscopic image analysis.
• It is versatile, applicable to images with and without local periodicity.
• The approach is freely available as a Python analysis notebook and program for batch processing.