Binary Screening Gate NORMAL/ABNORMAL in Brazilian Upper Gastrointestinal Endoscopy: Safety Analysis Guided by Target Sensitivity

• Willgnner Ferreira Santos UFG
• Paulo Victor dos Santos UFG
• Daniela Medeiros Milhomem Cardoso HGG
• Marcella Scoczynski Ribeiro Martins UTFPR
• Ronaldo Martins da Costa UFG

DOI: https://doi.org/10.5753/eniac.2026.30843

Resumo

Automatic screening of Upper Gastrointestinal Endoscopy (UGE) images can reduce the specialized reading workload, but its adoption depends on quantifying how many pathologies would be missed when discarding images as normal. We evaluated a binary NORMAL/ABNORMAL gate on 1,984 UGE images (image-level analysis; patient and exam identifiers are unavailable) from two Brazilian centers, comparing five CNN architectures in 5-fold cross-validation. DenseNet-121 achieved the best performance (AUROC 0.887; AUPRC 0.903). At the 95% target sensitivity, the gate filters 24.2% of images (NPV 0.892), missing 28 of the 616 images with pathology (4.5%); relative missrates are higher for rare classes, and a global threshold does not guarantee the target in both centers.

Referências

Al Hayek, M., Barberio, B., Al Hayek, O., Rugge, M., Di Pietro, M., Noviello, D., et al. (2026). Efficacy of artificial intelligence–assisted upper gastrointestinal endoscopy for neoplasm detection: a systematic review and meta-analysis of randomized controlled trials. Gastrointestinal Endoscopy.

Ali, H., Yasmin, M., Sharif, M., and Rehmani, M. H. (2018). Computer assisted gastric abnormalities detection using hybrid texture descriptors for chromoendoscopy images. Computer Methods and Programs in Biomedicine.

Arribas, J., Antonelli, G., Frazzoni, L., Fuccio, L., Ebigbo, A., van der Sommen, F., Ghatwary, N., et al. (2021). Standalone performance of artificial intelligence for upper GI neoplasia: a meta-analysis. Gut.

Chang, Y. H., Shin, C. M., Lee, H. D., Park, J., Jeon, J., Cho, S.-J., Kang, S. J., et al. (2024). Real-world application of artificial intelligence for detecting pathologic gastric atypia and neoplastic lesions. Journal of Gastric Cancer.

Choi, S. J., Khan, M. A., Choi, H. S., Choo, J., Lee, J. M., Kwon, S., Keum, B., and Chun, H. J. (2022). Development of artificial intelligence system for quality control of photo documentation in esophagogastroduodenoscopy. Surgical Endoscopy.

Fujishiro, M., Kakushima, N., Abe, S., Yip, H. C., Liu, X., et al. (2026). World Endoscopy Organization Position Statements for Artificial Intelligence in Endoscopic Diagnosis of Gastric Neoplasia. Digestive Endoscopy.

He, Y., Chen, Q., Liu, B., and Cao, Y. (2025). Foundational multi-task multimodal model for upper GI endoscopy. In 2025 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW).

Krenzer, A., Friedetzki, T., Dietsch, M., and Puppe, F. (2026). Classification of gastroscopy images under extreme class imbalance: A deep learning pipeline. In Lecture Notes in Computer Science.

Lui, T. K. L., Tsui, V. W. M., and Leung, W. K. (2020). Accuracy of artificial intelligence-assisted detection of upper GI lesions: a systematic review and meta-analysis. Gastrointestinal Endoscopy.

Osawa, S., Yamada, T., Inui, W., Niwa, T., Takahashi, K., Egami, T., et al. (2026). Real-world evaluation of an AI-assisted diagnostic support system for early gastric cancer: Diagnostic performance, confidence stratification, and determinants of false-positive diagnosis. Journal of Clinical Medicine.

Su, X., Liu, Q., Gao, X., and Ma, L. (2023). Evaluation of deep learning methods for early gastric cancer detection using gastroscopic images. Technology and Health Care.

Teramoto, A., Shibata, T., Yamada, H., Hirooka, Y., Saito, K., and Fujita, H. (2022). Detection and characterization of gastric cancer using cascade deep learning model in endoscopic images. Diagnostics.

Truong, V. D., Nguyen, D. A., Manh, X. H., Nguyen, T. H., Pham, T. T., Nguyen, T. T., et al. (2024). Automatic classifying informative frames for upper gastrointestinal endoscopy diagnosis using deep learning method. In Lecture Notes in Networks and Systems.

Yang, H., Wu, Y., Yang, B., Wu, M., Zhou, J., Liu, Q., Lin, Y., Li, S., Li, X., Zhang, J., Wang, R., Xie, T., et al. (2022). Identification of upper GI diseases during screening gastroscopy using a deep convolutional neural network algorithm. Gastrointestinal Endoscopy.

Yin, M., Zhang, R., Lin, J., Zhu, S., Liu, L., Liu, X., Lu, J., Xu, C., et al. (2023). Identification of gastric signet ring cell carcinoma based on endoscopic images using few-shot learning. Digestive and Liver Disease.

You, H., Tao, X., Dong, Z., Li, Y., Huang, L., Li, J., Du, H., Deng, M., Wu, Z., et al. (2026). Artificial intelligence assistance narrows the experience gap in endoscopic reporting of gastric lesions: a prospective study. Surgical Endoscopy.

Zhang, L., Yao, L., Lu, Z., and Yu, H. (2024). Current status of quality control in screening esophagogastroduodenoscopy and the emerging role of artificial intelligence. Digestive Endoscopy.