AI-Based Diagnosis System Identifies Malaria Parasites from Blood Smear Images

Factors such as fatigue, a shortage of skilled professionals, and the varying appearance of the parasite at different life stages often complicate accurate diagnoses. The application of artificial intelligence (AI) in healthcare continues to expand, including its potential to help diagnose tropical diseases like malaria, which remains a significant health threat in several regions worldwide.

Researchers at the National Research and Innovation Agency (BRIN, Jakarta, Indonesia) have developed an AI-based diagnostic tool to assist healthcare workers in identifying malaria parasites. This system analyzes microscopic images of thin and thick blood smears to detect signs of infection. To develop this tool, the researchers used a dataset of 1,388 blood smear microphotos collected from malaria-endemic areas in Indonesia. The dataset includes various malaria parasite types, such as Plasmodium falciparum, P. vivax, P. malariae, and P. ovale, along with one case of mixed infection and one negative sample.

Early testing of the AI-based diagnostic system has yielded promising results. The system was tested using 35 micrographs from real cases in malaria-endemic areas of Indonesia, covering 3,362 cells. The AI tool demonstrated a strong ability to identify malaria parasites, with a sensitivity of 84.37% in distinguishing between healthy and infected cells. The system achieved an accuracy value (F1-score) of 80.60% and a positive predictive value (PPV) of 77.14% in correctly identifying the parasite species and their stages. These results suggest that the system is highly reliable in distinguishing infected blood cells from healthy ones. This diagnostic system is also designed to facilitate mass blood surveys in the field, where a single smear may require observation of 500 to 1,000 erythrocytes or 200 leukocytes. AI can accelerate this process while maintaining accuracy.

Beyond improving efficiency, this system also opens up the possibility of remote diagnostics, making it especially relevant for use in underserved areas. Additionally, the system retains microscopic knowledge and expertise, aiding health workers with limited training. The researchers highlight the importance of addressing factors like dataset characteristics, data quality, model selection, and proper performance evaluation methods in the development of AI for biomedical applications. AI alone cannot function effectively—collaboration between computing experts and biomedical researchers is crucial for ensuring the reliability of such technologies. With the potential to significantly enhance diagnostic accuracy and improve healthcare delivery in malaria-endemic areas, the researchers are optimistic that AI will become a valuable partner in national malaria control efforts. The team is committed to further refining the system through extensive collaborative research and field trials.

Related Links:
BRIN