Combined flat-field and frequency filter approach to correcting artifacts of multichannel two-photon microscopy

Multiphoton microscopy (MPM) is a useful biomedical imaging tool for its ability to probe labeled and unlabeled depth-resolved tissue biomarkers at high resolution. Automated MPM tile scanning allows for whole-slide image acquisition but can suffer from tile-stitching artifacts that prevent accurate quantitative data analysis.

We have investigated postprocessing artifact correction methods using ImageJ macros and custom Python code. Quantitative and qualitative comparisons of these methods were made using whole-slide MPM autofluorescence and second-harmonic generation images of human duodenal tissue.

Image quality after artifact removal is assessed by evaluating the processed image and its unprocessed counterpart using the root mean square error, structural similarity index, and image histogram measurements.

Consideration of both quantitative and qualitative results suggest that a combination of a custom flat-field-based correction and frequency filtering processing step provide improved artifact correction when compared with each method used independently to correct for tiling artifacts of tile-scan MPM images.

While some image artifacts remain with these methods, further optimization of these processing steps may result in computational-efficient methods for removing these artifacts that are ubiquitous in large-scale MPM imaging. Removal of these artifacts with retention of the original image information would facilitate the use of this imaging modality in both research and clinical settings, where it is highly useful in collecting detailed morphologic and optical properties of tissue.