AI in diagnostics: 11 real deployments

xR RNA-sequencing assay for cancer genomic profiling and identification of targetable gene fusions

Reported: pairing xR with Tempus's DNA testing identified 19% more targetable gene fusions through RNA alone, raising matched-therapy eligibility by 24%

Used Caristo's CaRi-Heart AI to detect coronary artery inflammation on routine coronary CT angiography in asymptomatic adults

Reported: 42.9% of asymptomatic adults showed elevated coronary artery inflammation; 17.9% had very high inflammation; 47% of participants with no detectable plaque still had elevated perivascular inflammation

AI-based BioIntegrated FFR-CT model for non-invasive coronary ischemia estimation, FDA 510(k) cleared

Reported: 92.2% sensitivity and 93.1% specificity versus invasive FFR across 491 vessels in 347 patients in the ELUCIDATE trial

FAHR-FaceSurvival foundation model used to flag 90-day cancer mortality risk from a single face photo as a complement to standard prognostic workups

Reported: Of 122 patients flagged as high-risk, 52.5% died within 90 days, versus about 2% of 975 patients in the lowest-risk group

Deploying an AI ECG system to detect cardiac conditions in patients

Reported: Spotted cardiac amyloidosis in patients whose symptoms could have pointed elsewhere

The UK government is investing £20 million to extend AI-powered chest X-ray analysis for lung cancer diagnosis, currently available in half of England's NHS Trusts, to all NHS Trusts by 2029, with the AI acting as a virtual second pair of eyes for radiologists.

Reported: Early deployments have already helped over 4 million patients receive faster lung cancer diagnoses or all-clears, cutting analysis time for complex cases from 8 days to approximately 4 days.

Five Johns Hopkins hospitals prospectively validated Bayesian Health's TREWS algorithm, which runs continuously against live EHR streams to flag sepsis before clinicians do.

Reported: 82% sensitivity; documented 18% relative mortality reduction when alerts were acted on within an hour.

Mayo Clinic's REDMOD radiomics model analyzed ~2,000 routine CTs to detect pancreatic cancer up to 3 years before diagnosis, with the prospective AI-PACED trial now enrolling elevated-risk cohorts.

Reported: flagged 73% of prediagnostic cancers at a median 16 months pre-diagnosis, roughly double unaided specialist sensitivity; 19% false-positive rate (81/430 controls flagged)

Applies SCOPE IO AI pathology platform to H&E whole-slide images for tumor microenvironment analysis and biomarker discovery in Summit Therapeutics' ivonescimab trials across NSCLC, colorectal, urothelial, head-and-neck, and biliary tract cancers

Deployed AI vision system in CameraJet toothbrush trained on 470,000 dental images that analyzes 28 live images per second to detect gaps between teeth and direct a water jet within 100ms

Reported: analyzes 28 live images per second, detects gaps within 100ms

Taiwan's NHIA runs Google's open-weight MedGemma model on 30,000+ pathology reports for lung-surgery planning.