AI-based diagnostics company Noul said it presented its latest research on device-embedded artificial intelligence for blood and cancer pathology at the 28th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2025), held from Sept. 23 to 27 at the Daejeon Convention Center.

MICCAI is regarded as one of the world’s most prestigious academic gatherings in medical imaging, attracting global experts in analysis and clinical applications. At this year’s meeting, Noul showcased how its on-device diagnostic platform miLab overcomes longstanding limitations in pathology such as data scarcity and imbalance, which have hindered diagnostic performance.

During a workshop session, the company unveiled two core studies. The first demonstrated how a cascaded classification framework -- combining lightweight convolutional neural networks with compact vision transformers -- improves sensitivity and efficiency for malaria detection on miLab.

Clinical validation in Malawi and Ethiopia showed the system achieved 98 percent sensitivity and 99.3 percent specificity, about 11 percentage points higher in sensitivity compared with manual microscopy.

The second presentation detailed how generative augmentation using advanced AI models produced high-quality synthetic cell images to strengthen training datasets.

This approach significantly improved detection of malaria-infected red blood cells, including rare gametocyte forms that are difficult to capture in real-world samples.

Noul CEO David Lim said the company’s platform, powered by NVIDIA’s Jetson board, integrates AI inference and automated imaging to deliver faster and more accurate diagnostics in both high- and low-resource settings.

“Our research confirms that Noul’s medical imaging and AI technologies surpass conventional methods, with broad potential for blood and cancer pathology,” he said.

The company emphasized that the findings are not limited to malaria but can be applied to other infectious diseases, tuberculosis, rare hematologic disorders, and oncology where sample scarcity or morphological variability pose challenges.

Noul has already commercialized miLab, the world’s first on-device AI pathology platform integrating automated staining, imaging, and analysis. By simply switching reagents in the cartridge, the system can adapt to different diagnostic areas. It has so far been deployed for malaria detection, blood analysis, and cervical cancer screening, and the firm plans to expand applications further.