Transilvania Digital Innovation Summit β Oct 8, 2026, Cluj-Napoca π·π΄ 40+ speakers, AI & digital transformation, public sector innovation, leadership, investment, and matchmaking β all in one day. π
ποΈ Tickets: https://t.co/V1zoK7onzD #TransilvaniaDIH#AI#Innovation
MiWear consortium paper is out in @CellCellPress, mid-infrared wearable sensor for metabolic syndrome.
Non-invasive. Continuous. From interstitial fluid in the skin.
AIE leads the AI layer β mid-IR spectral data -> clinically actionable biomarker readings. π§¬
Link in threadπ
"Innovation without rigor is dangerous in healthcare β you're dealing with people's lives."
β Dr. Alexandru Floares, CEO of AIE, in our @Insightscare cover feature.
Every model explainable. Every result traceable. festina lente. π§¬
#HealthcareAI#ExplainableAI
We're the cover story of @Insightscare Magazine's May 2026 edition on AI-Driven Biomedical Data Analysis Services. 𧬠The feature traces it back to 2000 β when our CEO founded the first AI department of Cluj-Napoca, long before "healthcare AI" was a thing. Full story π below
90 days until the EU AI Act's major enforcement date. If your clinical AI model was built for accuracy but not transparency β no documentation, explainability as afterthought β August 2nd is a problem. Fixable. But not in July. #EUAIAct#ClinicalAI
Multi-cancer classifier: 95β99% accuracy. 5 cancer types. 15K+ serum samples. First reaction: what's leaking? Strict splitting. Imbalance handling. SHAP explainability. The number held up. Peer-reviewed and published. #CancerDetection#ClinicalAI
Thread: patients with depression, single wearable EEG session before first dose. 556 features computed. Most contributed nothing. Signal lived in non-linear dynamics. https://t.co/DyQusWOCo4
556 features from wearable EEG. Less than 3% drove half the predictive power. The features that mattered weren't obvious β traditional spectral power wasn't where the signal lived. More features β more signal. Especially in small clinical datasets. #EEG#OPADE#HealthAI#Horizon
"Can you do this with our data?" That's the question we kept hearing. So we built a page that answers it. 4 service areas. Real methods. Validated across 40+ sites in 15+ countries through Horizon Europe. Have a data challenge?
https://t.co/DyQusWOCo4
#BiomedicalAI#HorizonEU
"Your model says this patient won't respond. Why?"
If you can't answer that, your model won't be used.
Clinicians reject AI because they need reasoning before changing treatment. Not just hype.
"The neural network said so" isn't reasoning.
EU AI Act is coming. Ready?
Your collaborators need your patient data.
GDPR says no.
Synthetic data is one way through β preserve statistical properties, share nothing real.
Easy to say. Hard to do well.
We've been working on this with multi-modal clinical data across EU projects. Details coming π
4 hospitals. 4 countries. 4 completely different ways of recording the same clinical data.
Column names don't match. Lab ranges differ. Someone entered age as a date.
Data harmonization isn't preprocessing β it's a research question.
We've spent 3 years building solutions.
30-50% of patients don't respond to their first antidepressant.
What if you could know before treatment starts?
We're working on it β wearable EEG, single recording, before the first dose.
The data is messy. The signals are there. π§
what 556 brain features taught us, soon.
@OpadeProject The real challenge wasn't algorithms.
It was harmonizing data that was never designed to be harmonized.
Making sure the model doesn't just memorize one centre's patterns.
More from this work coming soon β
#BiomedicalAI#HorizonEurope#OPADE
Hardest problem in clinical ML isn't building the model but making it work across multiple hospitals.
In @OpadeProject, we tackled this: 350 patients, 4 countries, 9 data modalities, 789K features β 200 meaningful ones.
Real-world clinical data is messy. Worth it.π§¬
We're on X.
Artificial Intelligence Expert β Romanian biomedical data science company. GenAI & Explainable AI for precision healthcare.
4 EU Horizon projects: oncology, mental health, wearables, digital twins.
Follow for what we're learning 𧬠https://t.co/rEzMhyffNr