Artificial Intelligence Technology Accelerates Analysis of 3D Medical Images

Advanced artificial intelligence models face a significant challenge: their need for vast quantities of precisely labeled medical images to train them to diagnose diseases. This difficulty is compounded by three-dimensional scans that comprise hundreds of slices, requiring review by specialized physicians. In a study published in the journal Nature Biomedical Engineering, researchers led by Dr. Charles Wyckoff presented a deep learning framework called “SLIViT,” which can accurately analyze volumetric medical imaging data while requiring far fewer training examples than traditional methods. It achieves this by leveraging knowledge derived from two-dimensional images and applying it to more complex three-dimensional scans. The system demonstrated outstanding performance in identifying biomarkers associated with retinal diseases, such as age-related macular degeneration, outperforming specialized AI models and achieving results comparable to those of clinical specialists, but at a faster speed. Researchers also successfully applied it to cardiac, hepatic, and pulmonary imaging, suggesting its potential as a foundational tool across multiple medical specialties rather than requiring a separate model for each imaging modality. Wyckoff believes this technology could accelerate biomarker research for diseases, reducing both time and cost, while emphasizing that its aim is not to replace physicians but to enhance their capacity to handle increasingly complex imaging data.