How Hybrid Bright-Dark Field Imaging Could Enhance Digital Pathology Resolution

Digital pathology is changing how pathologists view, share, and analyze tissue samples.

At the center of this transformation is whole-slide imaging (WSI).

WSI converts glass slides into high-resolution digital images. These images can then support diagnosis, research, remote consultation, and AI-assisted analysis.

However, digital pathology still faces an important challenge.

How can imaging systems achieve higher resolution while maintaining scanning speed and efficiency?

A recent study published in Opto-Electronic Advances explores a new approach to this challenge. The research introduces a Hybrid Bright-Dark Field Resolution Enhancement framework, or HBDF-RE.

KFBIO’s GUI KUN Kehui Wang is a co-author of this research.

The study shows how hybrid imaging and physics-guided deep learning can work together to enhance digital pathology images.

The Resolution Challenge in Digital Pathology

High-resolution imaging is essential for digital pathology.

Fine cellular structures can contain important diagnostic information. Therefore, digital pathology systems need to capture these details accurately.

However, higher optical resolution can also increase system complexity.

High-numerical-aperture (NA) imaging systems can provide more detailed images. At the same time, they may introduce higher hardware requirements and slower scanning speeds.

This creates a long-standing balance between three factors:

• Image resolution
• Scanning efficiency
• System complexity

The challenge becomes even more important for large-scale applications.

For example, high-throughput WSI systems may need to scan hundreds or thousands of slides. In these workflows, both image quality and scanning efficiency matter.

AI creates another requirement.

Pathology AI models rely on digital images as their input. Therefore, image quality can directly affect downstream analysis.

This raises an important question:

Can computational imaging help improve resolution without relying only on more complex optical hardware?

A New Approach: Hybrid Bright-Dark Field Imaging

The HBDF-RE framework explores this possibility.

Instead of relying on a single low-resolution image, the method combines two types of optical information.

The first is a conventional bright-field image.

The second is an additional dark-field image.

Dark-field imaging provides complementary information related to light scattering and fine edges. This information can help reveal details that may not be clear in a conventional bright-field image.

The system then combines these complementary signals with physics-guided deep learning.

In this way, the additional dark-field information provides physical guidance for image reconstruction.

The researchers used a programmable LED illumination system to switch between bright-field and dark-field modes.

A paired bright-dark field image can be acquired in approximately 1/15 second.

This approach aims to address one of the key limitations of conventional single-image super-resolution.

When only one low-resolution image is available, the reconstruction process has limited physical information.

As a result, some methods may produce over-smoothed structures or artificial textures.

For pathology, this is particularly important.

Artificial image features could affect downstream image analysis. Therefore, preserving reliable tissue structures is critical.

What Did the Study Achieve?

The research team evaluated HBDF-RE through several experiments.

The results showed significant improvements in image resolution and reconstruction quality.

2.1× Spatial Resolution Enhancement

With one additional dark-field acquisition, the framework achieved approximately 2.1× spatial resolution enhancement.

The reconstructed images showed cellular structures approaching the performance of high-NA imaging.

84% Reduction in Reconstruction Artifacts

Compared with representative single-image resolution enhancement methods, HBDF-RE reduced reconstruction artifacts by approximately 84%.

This result is important for digital pathology.

Pathology images need to preserve meaningful tissue structures rather than simply appear sharper.

3.2 dB Improvement in PSNR

The method also improved peak signal-to-noise ratio (PSNR) by approximately 3.2 dB in the reported whole-slide imaging experiments.

This indicates improved reconstruction fidelity compared with the evaluated single-image methods.

11.14% Higher AI Screening Sensitivity

The researchers also evaluated HBDF-RE in an AI-assisted cervical cancer screening task.

The AI model using reconstructed images achieved an 11.14% improvement in diagnostic sensitivity compared with the original low-resolution images in the reported experiments.

The improvement was particularly noticeable in clinically ambiguous lesion categories.

These findings suggest that resolution enhancement may affect more than image appearance.

It may also influence downstream AI analysis.

From Better Images to Smarter Digital Pathology

The potential value of this research extends beyond image reconstruction.

Digital pathology is becoming increasingly connected with artificial intelligence.

A typical workflow can include:

Slide Preparation → Whole-Slide Imaging → Digital Slide Management → AI Analysis → Pathology Review

Each step depends on reliable digital image data.

Therefore, improvements at the imaging stage may create opportunities for the broader digital pathology workflow.

The study also demonstrated large-field super-resolution reconstruction on human thymus tissue whole-slide images.

The researchers reported improved visualization of fine tissue structures that were less clear in the original low-NA scans.

These findings point toward potential applications in large-scale digital pathology.

However, further research and validation will be needed before such approaches become part of routine clinical workflows.

Connecting Imaging Innovation with Digital Pathology

This research also highlights an important direction for the industry.

Digital pathology is no longer only about digitizing glass slides.

It is increasingly about creating a complete digital ecosystem.

High-quality imaging provides the foundation.

Digital slide management connects images with clinical workflows.

AI then adds automated analysis and decision-support capabilities.

At KFBIO, we focus on this broader digital pathology ecosystem.

KFBIO provides an end-to-end digital pathology and AI solution, covering slide scanners, histology solutions, information systems, and pathology AI.

Our digital pathology scanners are designed to provide fast, precise, and stable whole-slide imaging.

For example, the KF-PRO Series covers different throughput requirements, from 5 to 400 slides. The series combines automated scanning with high-resolution imaging and digital pathology software.

The KF-PRO-400 can scan a 15 mm × 15 mm area in 15 seconds at 20X and 25 seconds at 40X. It provides 0.25 μm/pixel resolution at 20X and 0.125 μm/pixel at 40X.

These capabilities support applications such as clinical diagnosis, research, education, and remote consultation.

From High-Quality Imaging to AI-Powered Analysis

Imaging is only one part of the digital pathology workflow.

Once slides are digitized, they can enter a broader ecosystem of digital management and AI analysis.

KFBIO provides digital slide management and pathology AI solutions for different applications. These include cervical cancer, gastric cancer, thyroid cancer, and pathology quality control.

This creates a connected workflow:

High-Quality Imaging

↓

Digital Slide Management

↓

AI-Powered Analysis

↓

Pathology Workflow

The HBDF-RE study explores how advanced computational imaging may further improve the first step of this process.

Meanwhile, digital pathology platforms provide the infrastructure needed to capture, manage, analyze, and share pathology images.

Together, these technologies point toward a more connected approach to digital pathology.

Looking Toward the Next Generation of Digital Pathology

The future of digital pathology will depend on more than higher image resolution.

It will also require speed, reliability, scalability, and intelligent analysis.

Hybrid imaging provides one possible path toward this goal.

By combining complementary optical information with physics-guided deep learning, HBDF-RE demonstrates a new approach to resolution enhancement.

The reported results are promising.

At the same time, the technology remains an area of ongoing research and validation.

The researchers suggest that future integration with automated WSI platforms and real-time deep learning could support applications in large-scale cancer screening, precision pathology, and AI-assisted clinical workflows.

For KFBIO, this direction aligns with a broader vision.

We believe digital pathology should connect imaging, software, AI, and clinical workflows.

As imaging technologies continue to evolve, KFBIO will continue to explore new ways to support more efficient and intelligent pathology workflows.

Research Reference

Huang, R., Zhang, R., Chen, X., Shu, Y., et al.
Learning from Hybrid Bright-Dark Field Imaging for Resolution-Enhanced Digital Pathology.

Opto-Electronic Advances, 2026.

DOI: 10.29026/oea.2026.260060

KFBIO’s GUI KUN, and Kehui Wang are co-author of the study.