Automatic optimization of flat-field corrections by evaluation and enhancement (EVEN) in multimodal optical...

Area of Science:

• Microscopy
• Image Processing
• Machine Learning

Background:

• Uneven illumination is a common artifact in optical microscopy, affecting image quality, especially in large, multicolour, and nonlinear measurements.
• Current methods for assessing and correcting uneven illumination lack standardized workflows, hindering reliable image quality evaluation.

Purpose of the Study:

• To introduce a machine learning-based method, EVEN (Evaluation and Enhancement), for assessing and optimizing image corrections in optical microscopy.
• To provide a standardized workflow for image quality assessment in the presence of uneven illumination.

Main Methods:

• Developed a machine learning model (EVEN) integrating quantitative image metrics.
• Utilized Linear Discriminant Analysis (LDA) for image quality detection and prediction.
• Implemented EVEN within the optical microscopy pipeline for automated correction optimization.

Main Results:

• EVEN successfully detects and predicts image quality, automatically optimizing corrections.
• Demonstrated EVEN's capability in multimodal nonlinear imaging of tissue slices and multichannel fluorescence microscopy of stained cells.
• Showcased automated image quality optimization through single-channel correction assessment.

Conclusions:

• EVEN offers a robust, automated solution for assessing and enhancing image quality in optical microscopy.
• The method simplifies downstream image processing and analysis by providing reliable quality control.
• EVEN's adaptability across different imaging modalities highlights its potential for broad application in microscopy.