Abstract
Cholangiocarcinoma (CCA) is a prevalent malignancy in countries along Mekong basin, closely linked to chronic infections caused by Opisthorchis viverrini (OV). Early detection of OV-infected individuals holds significant promise for screening at-risk populations in endemic regions. Recent advancements in immunochromatographic methods have led to the development of a rapid diagnostic test (RDT) based on urinary antigens. However, the current interpretation relies on visual assessment of T-band color intensity, which can be subjective and prone to variability. Furthermore, aggregating data at the regional/country level demands data digitization, a time and resource intense task that introduces further errors. To address this limitation, we introduce the OV-RDT platform, a cloud-based system incorporating artificial intelligence (AI) designed to standardize the reading and interpretation of OV-RDT results while facilitating mass screening campaigns for opisthorchiasis. This cross-platform solution, available on Android and iOS devices, consists of three key components: a mobile application, an intelligent dashboard, and a cloud server cluster. The server cluster has two main components the data processing server and the AI server. The AI server operates two AI-based models systematically developed and validated for image quality assessment and T-band grading of OV-RDT test kit images. The data processing server periodically retrieves and processes data from the cloud database, enabling comprehensive daily visualization through the intelligent dashboard. Validation through extensive field testing was conducted specifically in the northeastern region of Thailand, where opisthorchiasis prevalence is among the highest globally, demonstrating remarkable effectiveness by processing over 100,000 samples. While our platform shows excellent performance in this endemic region, external validation in other geographical areas would be necessary to establish broader generalizability. The EfficientNet-B5-based deep learning model used in the platform exhibited impressive performance in both image quality assessment (98% accuracy) and infection grading classification (95% accuracy in detecting OV infection status). The platform's user-friendly interface has achieved high satisfaction rates (4.41/5.00) among healthcare workers, while its intelligent dashboard offers real-time analytics and geospatial visualization capabilities. This integrated approach marks a significant advancement in mass screening for opisthorchiasis, potentially enhancing early detection rates and supporting more effective public health interventions in Southeast Asia and the Mekong Basin countries. This study addresses the critical need for mass screening in northeastern Thailand, where liver fluke infection rates are particularly severe; however, the platform's performance in other regions requires future validation studies.