Automated Histological Remission Assessment in Ulcerative Colitis via Single-Wavelength Endoscopy

Overview

A deep learning-based computer-aided diagnosis (CAD) system using single-wavelength endoscopy (SWE) demonstrated superior accuracy in detecting histological remission in ulcerative colitis (UC) compared to white light endoscopy (WLE). The SWE-CAD model achieved 95.2% diagnostic accuracy, significantly outperforming the WLE-CAD model.

Background

Ulcerative colitis management increasingly targets histological remission as it better predicts disease relapse than endoscopic activity alone. Traditional endoscopic scoring systems like the Mayo Endoscopic Score (MES) have limitations due to inter- and intra-rater variability and do not capture subtle microvascular changes associated with histological inflammation. Single-wavelength endoscopy (SWE) at 410 nm enhances visualization of mucosal microvasculature, potentially allowing better in vivo assessment of histological disease activity. This study developed and compared deep learning CAD models based on SWE and WLE imaging to automate histological remission detection.

Data Highlights

Key Findings

• The SWE-CAD model significantly outperformed the WLE-CAD model in sensitivity (88.0% vs 73.9%, p < 0.001) and diagnostic accuracy (83.3% vs 67.5%, p < 0.005) in initial training on 42 patients.
• Specificity differences between SWE-CAD and WLE-CAD were not statistically significant in the initial training (71.7% vs 65.6%, p = 0.45).
• Training the SWE-CAD model on the full dataset of 112 patients further improved performance, achieving 96.4% sensitivity, 92.9% specificity, and 95.2% accuracy.
• SWE imaging at 410 nm enhances visualization of mucosal microvasculature, enabling detection of subtle histological inflammation not captured by WLE.
• The automated SWE-CAD system provides objective, operator-independent assessment, reducing inter-reader variability in histological remission evaluation.

Clinical Implications

The SWE-CAD system offers a reliable, non-invasive tool for real-time assessment of histological remission in UC, potentially guiding treatment decisions without the need for repeated biopsies. Its high accuracy and objectivity may improve disease monitoring and patient outcomes by enabling precise evaluation of mucosal healing. Integration of SWE-CAD into clinical practice could standardize endoscopic assessments and reduce variability among endoscopists.

Conclusion

This study demonstrates that a deep learning CAD model based on single-wavelength endoscopy significantly enhances the accuracy of histological remission detection in ulcerative colitis compared to conventional white light endoscopy. SWE-CAD represents a promising advancement for objective, real-time disease activity assessment.

References

1. Author/Source/2024 -- Creation of an Automated System for Assessing Histological Remission in Ulcerative Colitis Utilizing Single-Wavelength Endoscopy Techniques