Segmentation of Multiple Nuclei from Non-overlapping Immuno-histochemically Stained Histological Hepatic Images

Area of Science:

• Digital pathology
• Computational biology
• Medical image analysis

Background:

• Accurate segmentation of multiple nuclei in histological images is crucial for diagnosing liver diseases like hepatocellular carcinoma.
• Challenges include stain variability, poor contrast, and interfering cells, complicating automated analysis.

Purpose of the Study:

• To develop an automated algorithm for precise segmentation of multiple nuclei in immuno-histochemically stained liver images.
• To address difficulties posed by staining, low contrast, and non-target cells for improved diagnostic accuracy.

Main Methods:

• A two-step algorithm employing the Quickhull algorithm to define cell convex hulls.
• Morphological operations identify candidate nuclei regions, followed by feature extraction (local minima, shape-dependent features).

Main Results:

• The algorithm achieved an accuracy of 89.76% in detecting single or multiple nuclei per cell.
• Demonstrated significant reduction in false positives and false negatives compared to existing methods.

Conclusions:

• The proposed method offers an effective solution for automated nucleus segmentation in challenging liver histology images.
• This contributes to the development of more reliable automated diagnostic systems for hepatocellular carcinoma.