Deep Learning for Improved Community-based Acute Flaccid Paralysis Surveillance in Ethiopia

Abstract

Acute flaccid paralysis (AFP) case surveillance is pivotal for early detection of potential poliovirus cases, particularly in endemic countries such as Ethiopia. Community-based surveillance system implemented by CORE Group Partners Project (CGPP) Ethiopia have significantly improved AFP surveillance. However, challenges like delayed detection and disorganized communication persist. This work proposes a simple deep learning method for AFP surveillance, leveraging transfer learning on images collected from CGPP Ethiopia's community key informants. The transfer learning approach is implemented using vision transformer model pretrained on ImageNet dataset. The proposed model outperforms traditional convolutional neural network-based methods and vision transformers trained from scratch, achieving superior accuracy, F1-score, precision, recall, and area under receiver operating curve (AUC). ViTB-16 emerges as the optimal architecture, demonstrating the highest average AUC of 0.870±0.01. Statistical analysis confirms the significant superiority of the proposed method over alternative approaches (P < 0.001). By bridging community reporting with health system response, this study offers a scalable solution for enhancing disease surveillance in low-resource settings. This study is limited in terms of the quality of image data collected, necessitating future work on improving data quality. The establishment of a dedicated platform that facilitates data storage, analysis, and future learning can strengthen global health security and disease eradication efforts. Nonetheless, this work represents a significant step towards leveraging artificial intelligence for community-based rare disease surveillance from images, with implications for addressing global health challenges and disease eradication strategies.

Keywords: Acute flaccid paralysis, surveillance, community, transfer learning, computer vision

JEL Classification: C00, I00, O00