Scientists from LMU, TU Berlin, and Charité have created a new AI tool that uses medical images to detect rare diseases in the digestive system. While AI is already used in many medical fields, it usually requires a lot of examples to learn how to spot problems. This is easy for common diseases but harder for rare ones. To solve this, the research team developed a new method that only needs data from common cases to also identify the rare ones. The study was published in New England Journal of Medicine AI.
How It Works:
The AI uses a technique called anomaly detection. It first learns to recognize what normal tissue looks like and what common diseases look like. Then, it can find unusual patterns that might indicate a rare disease, even if it hasn’t seen that exact condition before. The researchers tested the tool with 17 million images from 5,423 tissue samples. Their model accurately detected rare problems in the stomach and colon, including uncommon cancers. The AI even creates color-coded maps to show where the unusual areas are.
Why It Matters:
This tool could greatly help doctors by spotting normal cases and common diseases quickly, which could account for up to 30% of cases. It also helps prioritize more complex cases and reduces the chances of missed diagnoses. Although a doctor still needs to confirm the results, the AI could save time and make diagnosis easier.
Source: news-medical
