In partnership with
When it all clicks.
Why does business news feel like it’s written for people who already get it?
Morning Brew changes that.
It’s a free newsletter that breaks down what’s going on in business, finance, and tech — clearly, quickly, and with enough personality to keep things interesting. The result? You don’t just skim headlines. You actually understand what’s going on.
Try it yourself and join over 4 million professionals reading daily.
Hi!
Welcome back.
Lately, big tech is moving fast into healthcare AI.
Perplexity is starting to connect real patient data into its platform to provide personalized health insights. Oracle is introducing AI agents directly into clinical workflows. Microsoft continues to expand tools like Copilot and DAX into clinical documentation.
Companies like NVIDIA are building the infrastructure that will power much of what comes next.
It feels like we’re entering a different phase. What are your thoughts on this?
Let’s get into this week’s issue
🤖 AIBytes
Researchers tested an AI model that predicts leukemia type using routine blood tests. They validated it across multiple countries and improved it to work better in real clinical settings.
🔬 Methods
Design: Retrospective, multicenter validation study
Participants: 6,206 patients with acute leukemia from 20 centers (16 countries)
Adults: 4,460
Pediatric: 1,746
Routine labs:
White blood cells (WBC)
Platelets
Monocytes, lymphocytes
Lactate dehydrogenase (LDH)
Coagulation markers
Age, sex
AI model predicted:
Acute myeloid leukemia (AML)
Acute promyelocytic leukemia (APL)
Acute lymphoblastic leukemia (ALL)
Improvements:
Added a step to flag uncertain cases
Built a separate model for children
📊 Results
The model correctly identified leukemia type in most cases.
High-confidence mode:
Very accurate
But excluded up to 90% of patients
After improvement:
Accuracy stayed strong
Only 12% of patients excluded
Detection improved, especially for harder cases
Children:
Initial performance was lower
After retraining → accuracy became high
Limitation:
Results varied between hospitals
Lower accuracy in complex cases
🔑 Key Takeaways
AI can use basic blood tests to help identify leukemia early.
Models do not perform the same across all hospitals.
Separate models may be needed for different populations.
Flagging uncertain cases improves safety.
This supports clinicians, it does not replace diagnosis.
🔗 Turki AT, Fan Y, Hernández-Sánchez A, et al. International testing and refinement of AI algorithms predicting acute leukemia subtypes from routine laboratory data. Nat Commun. 2026;17:2649.
https://doi.org/10.1038/s41467-026-70584-z
Researchers tested an AI tool during real CT scan readings to see if it helps doctors find lung nodules faster and better.
🔬 Methods
Design: Prospective randomized controlled trial
Participants: 911 adults undergoing low-dose CT (LDCT) screening
Setting: Real clinical workflow (PACS system)
Groups:
AI-assisted reading
Standard reading (no AI)
Readers: 10 thoracic radiologists
AI tool:
Detected, measured, and classified lung nodules.
Results shown directly in the radiology system.
📊 Results
Reading time:
AI: 187 seconds
No AI: 172 seconds
Clinically important nodules:
AI: 16.9%
No AI: 10.3%
All nodules detected:
AI: 52.9%
No AI: 32.6%
Follow-up CT recommended:
AI: 15.3%
No AI: 7.4%
No lung cancers diagnosed during follow-up.
🔑 Key Takeaways
AI helps doctors find more nodules, including important ones.
It does not save time in real-world practice.
More detection leads to more follow-up scans.
Many detected nodules were not cancer.
Integration into workflow is a major limitation.
🔗 Hwang EJ, Lee T, Lim WH, et al. Artificial Intelligence–Assisted Lung Nodule Evaluation on Low-Dose Chest CT in Asymptomatic Individuals: A Prospective Randomized Controlled Trial. AJR. 2026.
https://doi.org/10.2214/AJR.26.34552
🦾TechTools
Connects health data, including medical records, labs, wearables, and Apple Health.
Answers health questions with more personal context than a standard web search.
Built for emergency and inpatient physicians.
Uses voice-enabled AI to draft clinical documentation during care.
Useful because it targets one of the biggest pain points in medicine: documentation burden.
Non-medical
Works directly inside your computer files, not just chat
Can read, edit, and create documents across folders autonomously
Represents the shift from chatbots → AI that actually does the work
🧬AIMedily Snaps
These medical X-rays are all deepfakes — and they fool even radiologists (Link).
Introducing Perplexity Health (Link).
UpToDate Expert AI now awards Continuing Medical Education Credits (Link).
AMA: Physician Survey on Augmented Intelligence (Link).
The Guardian: Google scraps AI search feature that crowdsourced amateur medical advice (Link).
OpenAI is throwing everything into building a fully automated researcher (Link).
Google is adding medical records to Fitbit personal health coach (Link).
🧪Research Signals
NEJM: Integrating Human–AI Collaboration in ECG Analysis for Clinical Advancement (Link).
Nature: The regulation of AI in intensive care units from narrow tools to generalist systems (Link).
Nature: Precision cardiovascular medicine with big data and AI (Link).
Nature: Large language models in healthcare (Link).
JMIR: The Right to Understand in Health Care AI (Link).
Nature: High-sensitivity pan-cancer AI assessment of lymph node metastasis via uncertainty quantification (Link).
🧩TriviaRX
Which condition has AI been shown to detect on CT scans months to years before clinical diagnosis?
A) Alzheimer’s disease
B) Pancreatic cancer
C) Multiple sclerosis
D) Rheumatoid arthritis
Now, time for the answer from last week TriviaRX.
✅ C. The AI may give slightly different answers each time.
AI models can produce slightly different answers even with the same input because they are probabilistic, not deterministic.
That’s it for today.
As always, thank you for taking the time to read.
You’re already ahead of the curve in medical AI — don’t keep it to yourself. Help another clinician get there—share AIMedily.
Until next week.
Itzel Fer, MD PM&R
Forwarded this email? Sign up here
P.S. Quick favor—if AIMedily has helped you, would you leave a short review? It takes less than a minute. 👉 Write a quick review.
