In partnership with
Talk to your AI tools the way you'd talk to a colleague.
You don't send a colleague a three-word brief. You explain the context, the constraints, what you've already tried. But typing all that into ChatGPT takes forever — so you don't.
Wispr Flow lets you speak your prompts instead. Talk through your thinking naturally and get clean, paste-ready text. No filler words. No cleanup. Just detailed prompts that actually get you useful answers on the first try.
Millions of users worldwide. Works system-wide on Mac, Windows, and iPhone.
Today, I’m publishing issue 50! 🥳 I honestly can’t believe I’ve been writing this newsletter for this long.
I want to say thank you.
Thank you for opening these emails, reading them, sharing them, and being part of this with me. I’m grateful for your support.
Almost every week, right before I click send — usually very late at night like today — I still worry.
Did I miss something important? Is my grammar correct? Is this useful? Do the links work properly?
But I also love it!
I love learning week by week. I love watching AI in healthcare unfold in real time. I love trying to understand it a little better and then sharing it with you.
My hope is that this newsletter helps you save time, stay informed, and feel a little more prepared for where medicine is going.
And since this is issue 50, I would love your help celebrating it 🎉
Please consider sharing it with a friend or healthcare professional who may enjoy it too 👉Share AIMedily (that would mean a lot to me).
Now, let’s get into this week’s issue.
🤖 AIBytes
Researchers tested whether an AI electrocardiogram (AI-ECG) could identify left ventricular systolic dysfunction (LVSD) in Kenyan outpatient clinics, where access to echocardiography may be limited.
🔬 Methods
Study design: Multicenter cross-sectional study
Setting: 8 outpatient clinics in Kenya
Participants: 6,678 adults enrolled
AI-ECG cohort: 5,992 participants had ECGs the AI could process
Paired test cohort: 1,444 participants had both ECG and echocardiography
AI tool: AiTiALVSD, a model that reads standard 12-lead ECG waveforms
Reference standard: Echocardiography within 7 days of ECG
📊 Results
The AI model could not process ECGs from 592 participants, mostly due to ECG quality or technical issues.
Echocardiography confirmed LVSD in 204 of 1,444 participants (14.1%).
AI-ECG detected LVSD with:
Sensitivity: 95.6%
Specificity: 79.4%
Performance remained strong across cardiovascular risk groups.
In the full AI-ECG cohort, 1,096 of 5,992 participants (18.3%) were flagged as high probability for LVSD.
A positive AI-ECG result was also associated with left ventricular hypertrophy and diastolic dysfunction.
Image from https://medicalai.com/
🔑 Key Takeaways
AI-ECG may help triage echo access. In clinics with limited echocardiography, it could help identify patients who need confirmatory imaging.
The strongest value was rule-out. The negative predictive value was 99.1%, making a negative AI-ECG result reassuring in this cohort.
Positive results still need echo. The positive predictive value was 43.2%, so AI-ECG should not be treated as diagnostic.
Signal quality matters. The model failed to process some ECGs, which is important for real-world deployment.
This is not outcomes evidence yet. The study did not test whether AI-ECG screening improved treatment initiation, heart failure outcomes, or survival.
🔗 Pandey A, Keshvani N, Segar MW, et al. Artificial intelligence electrocardiogram and left ventricular systolic dysfunction in Kenya. JAMA Cardiol. Published online May 6, 2026. doi:10.1001/jamacardio.2026.0908
Researchers tested whether Paige Prostate, an AI pathology tool, could help pathologists review prostate biopsies.
🔬 Methods
Study design: Prospective, multicenter service evaluation
Setting: 3 NHS specialist centers in England
Cases: 1,613 prostate biopsy cases
AI-assisted cases: 1,049 cases were read with AI support
AI tool: Paige Prostate, used to flag suspicious areas and help grade prostate cancer on digital slides
Readers: Specialist urological pathologists
Workflow: AI was tested in phases:
usual reporting without AI
AI as a second read
AI available during the full review
Main outcomes: Report changes, possible impact on management, reporting time, extra staining tests, and pathologist experience
📊 Results
When AI was used as a second read, it changed the final diagnosis or Grade Group in 21 of 386 cases (5.4%).
5 of 386 cases (1.3%) had changes that could have affected management options.
Most changes were Grade Group changes, not cancer vs no cancer.
Two cases had diagnostic changes:
one changed from benign to Grade Group 1 prostate cancer
one changed to a suspicious but not clearly cancerous finding
Reporting time improved at one site:
total time decreased by 30.1 hours
pathologist time decreased by 24.7 hours
Reporting time did not improve significantly at the other two sites.
Extra staining tests, called immunohistochemistry (IHC), decreased at all three sites.
At the end of the study, all surveyed pathologists said they would use AI if it were available.
🔑 Key Takeaways
AI changed some reports. It changed diagnosis or Grade Group in 5.4% of second-read cases.
Most changes did not change care. Only 1.3% of cases had changes that could affect management options.
The biggest benefit may be workflow. AI reduced extra staining tests at all three centers.
Speed gains were mixed. One center reported faster turnaround time; the other two did not.
Pathologists stayed in control. AI supported review, but pathologists made the final diagnosis.
Local setup matters. Benefits depended on digital workflow, lab process, and how well AI fit into daily work.
🔗 Browning L, Colling RT, Oxley J, et al. An evaluation of artificial intelligence assisted prostate biopsy reporting in the Articulate Pro study. npj Digit Med. 2026. doi:10.1038/s41746-026-02592-8
🦾TechTools
SeamlessMD recently launched Seamless Answers, a conversational AI feature for patients during their care journey.
Patients can ask questions at any time, and the answers come from care-team-approved content.
It focuses on the space between visits, when patients often have questions but clinical teams do not have unlimited time.
AI-supported symptom intake and triage for patient navigation.
Helps collect symptoms before the visit and guide patients toward the right level of care.
📈 Productivity Tool of the Week:
Brings AI into your computer workflow so you can search, summarize, and move faster across apps.
Helps with quick questions, writing, files, and repeated tasks without opening a million tabs.
🧬AIMedily Snaps
This is what’s been happening in AI in medicine this week:
OpenEvidence partnered with Cedars-Sinai to bring patient-aware clinical intelligence into the workflow (Link).
OpenAI is adding more safety context to ChatGPT for sensitive conversations and possible harm risk (Link).
University of Michigan is deploying Avenda Health’s AI-guided prostate cancer platform for 3D cancer mapping and treatment planning. (Link).
The Lancet published an audit of fabricated citations across 2.5 million biomedical papers (Link).
Mayo Clinic and Bayesian Health are using AI to help identify patients who may need palliative care earlier (Link).
A new report found that 71% of U.S. adults use health-related apps and 64% use health-related devices (Link).
🧪Research Signals
Papers worth your attention this week:
Nature: A multi-expert AI model was developed to support kidney biopsy triage (Link).
Nature: A vision-language model was tested for seizure detection across hospital, home, and public video recordings (Link).
JAMA: An AI-enabled CPR instructor was tested for bystander cardiopulmonary resuscitation guidance. (Link).
JMIR: Protocol for improving triage accuracy through real-time evaluation and AI (Link).
JAMA: A clinical decision support system for chronic kidney disease was tested in primary care (Link).
JAMA: A clinical decision support system increased antihypertensive treatment intensification, but did not improve overall control (Link).
🧩TriviaRX
In 1929, Werner Forssmann performed a risky self-experiment that later helped earn him a Nobel Prize. What did he do?
A) Injected himself with a heart medication
B) Advanced a catheter from his arm into his heart
C) Stopped his own heartbeat under anesthesia
D) Recorded the first human electrocardiogram
Now, the answer from last week: ✅ ✅ Answer: B) X-rays
In 1895, Wilhelm Röntgen noticed that a covered cathode-ray tube made a nearby screen glow. That observation led to the discovery of X-rays.
That’s it for today.
Thank you for being part of AIMedily.
Until next week (I’ll be writing from Stanford).
Itzel Fer, MD PM&R
Forwarded this email? Sign up here
P.S. Enjoying AIMedily? 👉 Write a review here (it takes less than a minute).
