New AI Diagnostic Tool Aletheia Offers Offline Support for African Healthcare

A new clinical decision support system, Aletheia, has been developed to address the severe shortage of specialist medical expertise in sub-Saharan Africa. Unlike most existing AI diagnostic tools, Aletheia is designed to function offline, making it practical for use in rural clinics and district hospitals that lack reliable internet connectivity and high-specification hardware. This innovation is crucial for frontline healthcare workers in regions where physician-to-patient ratios are extremely low, sometimes falling to 1:25,000.

The system is built on the Qwen2.5-3B-Instruct model, fine-tuned using QLoRA on a dataset of 27,000 clinical reasoning samples. This dataset specifically focuses on 50 disease conditions prevalent in East Africa, ensuring its relevance to the local health context. The research highlights the feasibility of deploying advanced large language model-based clinical reasoning directly at the primary care level, removing the dependency on cloud infrastructure.

Initial evaluations show promising results, with a Top-1 diagnostic accuracy of 80% and a Top-3 accuracy of 100% on a small set of ten clinical cases. While these figures are indicative rather than statistically robust due to the limited sample size, the system also successfully met the memory budget constraints of the Africa Deep Tech Challenge 2026. This demonstrates its practical viability for deployment on standard, lower-cost hardware, further enhancing its potential impact in resource-constrained settings.