Improving cellular nuclei detection in histological images by combining super-resolution techniques and deep...

Area of Science:

• Digital pathology
• Computational imaging
• Biomedical image analysis

Background:

• Accurate cellular nuclei detection is crucial for digital pathology but challenged by image resolution variability.
• Existing object detection methods often fail on low-resolution images, and solutions typically address enhancement or multi-scale detection separately.

Purpose of the Study:

• To develop a robust hybrid framework for cellular nuclei detection in histopathological images.
• To improve detection accuracy, especially in low-resolution scenarios, by integrating image enhancement and multi-scale detection.

Main Methods:

• A hybrid framework combining super-resolution with a dual-branch detection strategy (full-image and patch-based inference).
• Fusion of outputs from both branches using a confidence-weighted mechanism.
• Post-processing refinement including non-maximum suppression and clustering.

Main Results:

• The proposed method demonstrated consistent improvements over baseline approaches on the NuCLS dataset.
• Achieved up to a 20% increase in mAP@0.5-0.95 at higher confidence thresholds.
• Maintained competitive performance with state-of-the-art methods using a lightweight architecture.

Conclusions:

• Integrating super-resolution and multi-scale detection strategies effectively enhances nuclei detection in histopathological images.
• The hybrid framework offers a robust and efficient solution for challenges posed by low-resolution digital pathology data.
• This approach advances the capabilities of automated analysis in digital pathology.