A new AI model reads a routine biopsy slide and predicts cancer subtype, genetic mutations and survival outcomes across 32 solid tumor types at once, potentially shortcutting the wait for separate genomic testing.
UpDoc's FDA-cleared V1.0 platform uses a large language model to talk with type 2 diabetes patients and adjust insulin doses within physician-approved bounds, and it is now deployed at Cleveland Clinic, AHN, and UCSF, raising questions about how far generative AI should go in direct patient care.
RadNet's DeepHealth won FDA clearance for an AI system that standardizes breast ultrasound reads, cutting radiologist interpretation time by 37% and boosting cancer-detection sensitivity by 8% in clinical validation testing.
Andromeda Robotics' companion robot Abi has expanded to 22 Australian aged-care facilities and is now opening a US waitlist after a $23 million Series A round, even as University of Houston researchers warn that surveillance fears remain a major barrier to older adults trusting robots like it.
Noom has expanded AI photo, voice, and text food logging in 2026 as rivals like MyFitnessPal (via its Cal AI acquisition) push into the same space, even as benchmark testing shows AI food recognition still misidentifies about 1 in 4 meals.
Cellens, a Boston cancer diagnostics startup, raised a $6.5M Series Seed round led by SOSV for its AI and atomic-force-microscopy-based BioFeel test, which identified every recurrence case in a nearly 100-patient study versus 60-65% sensitivity for standard molecular assays.
Owkin announced on June 5, 2026 a multi-year expansion of its Sanofi partnership, licensing its K Pro 'AI Scientist' platform for five years to build autonomous AI agents that independently carry out pharmaceutical research tasks.
Aidoc won FDA clearance in January 2026 for a foundation-model AI workflow that triages 14 acute conditions from a single abdominal CT scan, hitting 97% mean sensitivity and 98% mean specificity in its pivotal study.
NIH researchers presenting at NUTRITION 2026 found four popular AI photo-based calorie-tracking apps underestimated meals by 250 to 345 calories on average, using a controlled metabolic kitchen to test MyFitnessPal, LoseIt!, CalAI, and Appediet against precisely measured meals.
A Mayo Clinic AI model called REDMOD detected 73% of pancreatic cancers a median of 16 months before diagnosis by re-analyzing CT scans radiologists had called normal, according to a study published in Gut.