The evidence is already in.
These aren’t projections. Across randomized trials and multi-site studies in American hospitals and community clinics, AI-assisted screening and detection are catching disease earlier and saving lives.
- +21.6%
Higher breast cancer detection
A US study across 109 community imaging sites in California, Delaware, Maryland, and New York found that a multistage AI-assisted breast cancer screening workflow was associated with a 21.6 percent higher cancer detection rate than the prior standard-of-care 3D mammography workflow,. Among women with dense breasts, the detection rate was 22.7 percent higher, with no significant differences in the workflow's benefit across the racial, ethnic, or breast-density groups studied.
- +29%
More cancers found in a 105,000-woman trial
In a randomized trial of more than 105,000 women an AI-supported mammography workflow increased the cancer detection rate by 29 percent compared with standard double reading, without a statistically significant increase in false positives.[The Lancet]
- 81% vs 74%
More cancers caught at screening
Full trial results found 81 percent of cancers were caught at screening in the AI group versus 74 percent in the control group, with false positive rates essentially unchanged at 1.5 percent and 1.4 percent. This matters because cancers diagnosed between routine screening exams are often the more aggressive ones and advanced types.[EurekAlert]
- −55%
Fewer missed adenomas in colonoscopy
An analysis of 28 randomized trials covering 23,861 patients found AI-assisted colonoscopy raised the adenoma detection rate by 20 percent and cut the adenoma miss rate by 55 percent.[GIE Journal]
- 5 per 1,000
New low-ejection-fraction diagnoses
In a randomized trial conducted by the Mayo Clinic across 45 sites in Minnesota and Wisconsin, including rural clinics, AI-guided EKG screening resulted in five new diagnoses of low ejection fraction per 1,000 patients screened, compared to routine care. According to the study's clinician, AI-guided screening enabled earlier detection of patients with previously unrecognized low ejection fraction who might have gone undiagnosed during standard care.
- −44%
Faster stroke transfers
At a regional primary stroke center, a quality-improvement initiative combining AI-assisted stroke detection and care coordination with standardized transfer protocols and a comprehensive-stroke-center partnership reduced average door-in-door-out transfer time for large-vessel-occlusion stroke patients from 202 minutes to 113 minutes, a 44 percent reduction. Reported care team notification time also fell from 45 minutes to 7 minutes.[Business Wire]The results show how AI-supported workflows can help regional and community hospitals connect stroke patients with potentially life-saving treatment faster.
- −31 min
Faster time to stroke treatment
A multicenter retrospective analysis of 474 patients found that implementing an AI-supported stroke detection and communication platform was associated with a 31 minute reduction in the adjusted time from initial hospital arrival to arterial puncture. The finding is especially meaningful because previous research estimates that every minute of faster endovascular treatment can preserve roughly four days of healthy, disability-free life.[Endovascular Today]
- 100% vs 22%
Completed diabetic eye exams
In a randomized trial of young people with diabetes, autonomous AI eye exams delivered at the point of care achieved a 100 percent completion rate for diabetic eye exams, compared with 22 percent among participants referred for standard eye care. Of the 25 participants whose AI exam identified possible diabetic eye disease, 64 percent completed follow-up with an eye care specialist. Diabetic retinopathy is the leading cause of blindness among working-age American adults, making timely screening and follow-up especially important.
- 2.75 days
Earlier COVID-19 warning signs
During the COVID-19 pandemic, researchers found that a machine-learning algorithm using physiological data from a consumer wearable could identify patterns associated with COVID-19 an average of 2.75 days before participants underwent diagnostic testing.
- −38%
Lower all-cause mortality with digital tools
A review covering more than 11.4 million participants found that digital cardiovascular tools, including wearable devices, smartphone applications, and AI and machine-learning models, generally showed better diagnostic accuracy than traditional risk scores. Wearables performed particularly well in detecting arrhythmias. Across studies that assessed clinical outcomes, digital health interventions were also associated with a 38 percent lower risk of all-cause mortality, although results varied substantially among the included studies.
- ~87%
Accuracy detecting sleep apnea
At a time when approximately 80% of obstructive sleep apnea cases go undiagnosed, a review of 38 studies found that wearable AI showed promising performance in detecting the condition. The systems achieved a pooled accuracy of about 87 percent for identifying people with sleep apnea. Researchers concluded that the technology does not replace standard diagnostic testing but supported the rate of detection.