Research

Arkangel AI: multi-LLM retrieval system gives evidence-based medical answers with 90% accuracy.

LLM auto-extracted ICU/outpatient (~100%) and applied PUMA: 94% scoring accuracy; 100% sensitivity

Minimal-data ML screened CKD in T2D across 6 LMICs: 90% sensitivity, AUC 0.63.

Review: LLMs score highly on QA but need human-in-loop, real-world evaluation for safe clinical use.

Vitruvius: multi-LLM, retrieval-augmented chat answers USMLE-style queries with 90.3% accuracy.

PANDORA used GPT-4 to extract clinical notes and apply PUMA: >90% extraction, 95-98% COPD scoring.

MedSearch: real‑time, evidence‑based AI cut clinicians' answer time 79% and searches 34% in pilot.

AI-CNN reads OCT to detect intraretinal fluid/macular edema with ~97% accuracy, aiding retina care.

GPT-4o conversational agent, trained on 17 AD guidelines, achieved near-perfect accuracy.

In COVID inpatients, AI chest X-ray + clinical data predicted ICU AUC 0.92; death AUC 0.81.

Latin America: ensemble AI using routine clinical data flags CKD - 91% sensitivity (T2D), 92% (NT2D)

Latin America: ensemble AI using routine clinical data flags CKD - 91% sensitivity (T2D), 92% (NT2D)

AI models across 68M patients predicted CKD, HF, diabetes risks (AUC>0.85), cutting admissions ≈45%.

Automated AI reading of chest X‑rays and low‑dose CT raised nodule sensitivity from 47% to 60% (+13%), cut false positives ~11%, and achieved ~94% diagnostic accuracy, aiding...

Hippocrates AutoML lets clinicians build no-code, HIPAA-compliant AI (auto data prep, model selection) 10x faster, cutting costs and enabling prospective validation—bias...

Arkangel AI used non-invasive ML on routine clinical data in Latin America to boost early CKD detection - high sensitivity/specificity, identifying up to 90% more at-risk patients.

In Latin America, Arkangel AI's mammogram + clinical-data AI decision support improved detection (AUC up to 0.90) and cut radiologist reading load by up to 88%, enabling earlier,...