Evaluation of an improved computer-aided detection system for Barrett’s neoplasia on real-world imaging conditions

Background: Computer-aided detection (CADe) systems may improve detection of Barrett’s neoplasia. Most CADe systems are developed with data from expert centers, unrepresentative of heterogeneous imaging conditions in community hospitals. Therefore, they may underperform in routine practice. We aimed to develop a robust CADe system (CADe 2.0) and compare its performance to a previously published system (CADe 1.0) under real-world imaging conditions. Method: CADe 2.0 was improved through a larger and more diverse training dataset, optimized pretraining, data augmentation and ground truth use, and architectural adjustments. CADe systems were evaluated using three prospective test sets. Test set 1 comprised 429 Barrett’s videos from 114 patients collected across five referral centers. Test set 2 addressed endoscopist-dependent variation (e.g. mucosal cleaning and esophageal expansion), with paired subsets of high, moderate, and low-quality images from 122 patients. Test set 3 addressed endoscopist-independent variation. It contained 16 paired subsets of 396 images (122 patients), each being based on a different software image-enhancement setting. Results: CADe 2.0 outperformed CADe 1.0 on all three test sets. For test set 1, sensitivity increased significantly from 86.9% to 96.4% (p=0.021), with on-par specificity scores. In test set 2, CADe 2.0 consistently surpassed CADe 1.0 across all image quality levels, with a larger performance gap on lower-quality images. In test set 3, CADe 2.0 showed improved performance and displayed reduced performance variability across enhancement settings. Conclusion: Based on several key improvements, CADe 2.0 demonstrated increased detection rates and better robustness to data heterogeneity, making it ready for clinical implementation.

Publication History

Received: 28 January 2025

Accepted after revision: 22 June 2025

Accepted Manuscript online:
25 June 2025