Multifractal-based nuclei segmentation in fish images.

Area of Science:

• Medical Imaging
• Computational Biology
• Pathology

Background:

• Accurate nuclei segmentation is crucial for quantitative analysis in fluorescence in-situ hybridization (FISH) imaging.
• Existing segmentation methods may have limitations in precision and automation.
• Reliable segmentation directly impacts diagnostic accuracy, particularly for cancer biomarkers like HER2.

Purpose of the Study:

• To propose and evaluate a novel semi-automatic method for nuclei segmentation in FISH images using inverse multifractal analysis (IMFA).
• To assess the performance of the IMFA method in facilitating accurate human epidermal growth factor receptor 2 (HER2) scoring.
• To compare the IMFA method against existing techniques for nuclei segmentation in clinical settings.

Main Methods:

• The proposed method utilizes inverse multifractal analysis (IMFA) applied to the blue channel of RGB FISH images.
• A matrix of Holder exponents is generated, corresponding to image pixels.
• Semi-automatic segmentation involves initial automatic thresholding followed by user-guided refinement.

Main Results:

• The IMFA method successfully segmented nuclei across 100 clinical FISH cases.
• The segmentation accuracy was evaluated by a skilled pathologist.
• The IMFA approach demonstrated advantages over previously reported nuclei segmentation methods.

Conclusions:

• Inverse multifractal analysis provides an effective tool for nuclei segmentation in FISH images.
• The IMFA method supports accurate HER2 scoring through precise nuclei identification.
• This technique offers a promising advancement for automated and semi-automated image analysis in pathology.