BALTIMORE, MD, August 23, 2025 – A new artificial intelligence (AI) method is showing promise in making early cancer detection more accurate by distinguishing between cancer-related signals and those triggered by inflammation in blood tests. This innovation could tackle a major hurdle in diagnostic tests that rely on circulating free DNA patterns, wich can mimic cancer signals in conditions like autoimmune and vascular diseases.
AI Tool Boosts Cancer Detection Accuracy
Researchers have developed an AI-powered approach to improve the precision of blood-based cancer tests.
The system, named Might, uses tens of thousands of decision trees to analyze complex biological data. This allows it to measure uncertainty and adjust its predictions, making it especially useful for biomedical datasets with many variables and limited patient samples.The tool, created by researchers at Johns Hopkins Medicine in Baltimore, Maryland, is designed to build the high level of trust needed for clinical decision-making.
- A new AI method, Might, aims to improve early cancer detection from blood samples.
- It distinguishes cancer signals from those caused by inflammation,reducing false positives.
- Studies show promising sensitivity and specificity in tests involving cancer patients.
- A complementary tool, Comight, combines biological signals for enhanced detection.
- The technology is now available for broader testing.
Might analyzes various biological characteristics in blood, such as DNA fragment lengths and chromosomal anomalies. It has been extended into a complementary tool called Comight, which merges sets of variables for even better detection. By training the algorithm with data related to inflammation, it can more precisely differentiate between cancer signals and those from other diseases.
published research indicates meaningful potential. In one study with 1,000 participants, including 352 cancer patients, Might demonstrated 72% sensitivity and 98% specificity using aneuploidy-based characteristics. Comight, tested on 125 breast cancer patients, 125 pancreatic cancer patients, and 500 controls, showed that combining biological signals improved early breast cancer detection and made pancreatic cancers easier to identify.
These findings highlight AI’s potential in diagnostics. beyond enhancing cancer detection,Might could also lead to new diagnostic tests for autoimmune and vascular diseases by addressing the issue of inflammation-induced false positives. While further clinical trials are necesary before widespread adoption, Might and Comight are currently accessible for broader testing via TreEPLE.AI.
“Confidence in the result is essential, and now that Might has a reliable quantitative tool, we and other researchers can use it and focus our efforts to study more patients and add statistically significant characteristics to our tests for the early detection of cancer,” said Dr. Bert Vogelstein, a study collaborator.
Researchers have developed an AI algorithm called Might for medical decisions and Might-based liquid biopsies to distinguish inflammatory di
