A Multi-Modal AI-Driven Framework for Early Gastric Cancer Detection in Low-Income Populations of Developing Countries

Background: Gastric cancer remains a major public health concern in developing countries, especially in East Asia. In China, over 70% of gastric cancer cases are diagnosed at advanced stages due to the lack of affordable early detection tools in low-income populations. Methods: We propose a conceptual three-stage AI-driven screening framework that integrates (1) a deep-learning-based image analysis stage inspired by the publicly published GRAPE model developed by Alibaba DAMO Academy, (2) a hypothetical serum biomarker panel (TriOx), and (3) structured clinical risk factors including Helicobacter pylori infection. No patient-level data were used in this study. Monte Carlo simulations were conducted to evaluate theoretical detection performance and cost-effectiveness in urban and rural scenarios. Results: In urban settings, simulation results suggest that the proposed AI framework could potentially improve early cancer detection rates from 0.06% (endoscopy) to 1.8%, and from 0.02% to 0.15% in rural regions. Simulated screening costs were reduced by 60% compared to endoscopy. Offline capability and modular design enable deployment in resource-limited hospitals. Conclusion: This AI-assisted, multimodal screening framework offers a cost-effective, scalable, and ethically responsible pathway for early gastric cancer detection in low-income populations, aligned with global and national health initiatives.