Enhancing Malaria RDT Interpretations with AI in Kano State, Nigeria

HealthPulse AI achieved a high overall interpretation accuracy, supporting and enhancing FHW performance which showed a 95.3% accuracy.

AI outperformed FHWs by over 14% in detecting faint positives, ensuring early and accurate malaria diagnosis.

Strong alignment between AI and FHWs in positive mRDT results, indicating that the majority of FHWs are accurately interpreting RDTs.

Tackling Malaria with Advanced Diagnostics

Malaria remains a major health threat in sub-Saharan Africa, causing significant illness and death. To reduce malaria mortality, morbidity, and disease spread, early and accurate diagnosis leading to effective treatment is critical. To support timely diagnosis, malaria rapid diagnostic tests (mRDTs) are widely used for their simplicity and speed. However, misinterpretation by Frontline Health Workers (FHWs) remains a challenge, especially with faint positive results.

Innovative AI-Driven Solution

To understand these challenges, health workers in Kano State, Nigeria used Audere’s HealthPulse AI, which was integrated into the workflow of ThinkMD’s mobile clinical risk assessment platform. Forty-four health workers  captured mRDT images and assessed over 2,800 mRDTs during the study, which ran from August to December 2020. AI algorithms provided an interpretation of the test result based on captured images.

Breakthroughs in AI-Assisted Malaria Diagnostics

The AI achieved a weighted F1 score of 96.4%, surpassing the FHWs' score of 95.3% and demonstrating the AI's reliability in supporting frontline health workers. AI correctly interpreted 97.12% of negative mRDTs and 96.38% of positive mRDTs.

One of the critical areas where AI showed remarkable value was in the identification of faint positives. A crucial capability for early and accurate diagnosis of malaria. AI correctly classified 90.2% of faint positive mRDTs compared to only 76.1% by FHWs.

The AI performed as well as or better than experienced and well trained FHWs, indicating AI’s strength in elevating FHW performance.  AI matched FHWs with 97.52% agreement on positive results and 93.38% on negative results.

Conclusion

As healthcare demands rise and access to highly trained personnel becomes limited, integrating AI solutions like HealthPulse can improve diagnostic accuracy. This study demonstrated that AI algorithms can perform as well as, or better than, experienced FHWs, particularly in challenging cases such as faint positive results. The synergy between AI and FHWs offers a promising solution to enhance malaria diagnostics and patient care in high-burden areas.