Braid Health

Researchers estimate 75,000 lives could be saved every year if everyone with a concerning scan got a second radiology opinion. Early detection. Your right. Your choice. braid.health/www/page/the-l… #SecondOpinion #PatientRights

Standard CNNs use fixed filters. Our Bandelet-DenseNet uses learnable geometry that adapts to how anatomy actually curves. 0.916 AUC, no pretraining. braid.health/www/page/bande… #RadiologyAI #DeepLearning

Our fracture AI: 1.9% miss rate. We published every false negative and false positive. Publishing your failures is what makes clinical AI trustworthy. braid.health/www/page/fract… #RadiologyAI #FractureDetection

Most radiology AI trains on 1-2 institutions. Ours: 360K+ studies, thousands of sites. External validation: 0.933 AUC. Full methodology: braid.health/www/page/xray-… #RadiologyAI

When you get a radiology report, ask for the DICOM files too. That's the actual image data, and it's yours under federal law. braid.health/www/page/dicom… #PatientRights #DICOM

Symptoms don't match your scan report? ER read done in a rush? Report says "cannot exclude"? These are all reasons to consider a second opinion. braid.health/www/page/secon… #RadiologySecondOpinion

Patients Googling their MRI results at 11pm deserve better than a wall of jargon. The imaging report was written for a clinician — not for you. That gap is the whole problem. (And yes, a radiologist should still be accountable for what it says.)

Our AI models don't all fail the same way. That's by design. Ensemble majority voting across architectures: 4.14% auto-clear, zero false negatives. braid.health/www/page/ensem… #RadiologyAI

Patients are Googling their CT and MRI results more than ever — and finding answers that often raise more anxiety than they resolve. A radiologist who can explain what the report actually means for your situation isn't optional. It's the whole point.

"Was something missed on my MRI?" gets searched thousands of times a day. The answer shouldn't be a blog post. It should be a radiologist who actually looked at your scan.

Our AI models achieved 98.1% sensitivity on fracture detection, trained on 359K+ studies. Full methodology + every false negative published. braid.health/www/page/fract… #RadiologyAI #FractureDetection

Everyone is racing to prove their AI is the most accurate in radiology. Nobody talks about what happens when it's wrong. The miss rate matters as much as the hit rate. Publish both. We do: braid.health/www/page/fract… #RadiologyAI

Our AI models cleared 4.14% of chest X-rays with zero false negatives — ~130 hrs/year returned per radiologist. Full methodology: braid.health/www/page/chest… #RadiologyAI

“We don’t release imaging files” If you’ve heard this, you should know: that’s often not true under federal law. What patients are legally entitled to — and what to do if access is delayed: braid.health/www/page/your-…

Radiology AI just hit a turning point at #RSNA2025. Specialized models → fading. Foundation models + agentic workflows → rising. Our breakdown of the shift happening right now in Chicago: braid.health/www/page/rsna-…

Which part of your scan makes you most curious? Ask AI connects multiple advanced models to explain scans in plain language. Explore: braid.health/ask-ai

Tech peek: why a second read finds what a first pass can miss. braid.health/www/page/could…

Which part of your scan makes you most curious? Ask AI connects multiple advanced models to explain scans in plain language. Explore: braid.health/ask-ai

Your scans hold thousands of biomarkers. Ask AI explains them in plain English. Read more at braid.health/www/page/ask-a…

Hundreds of AI perspectives, one clear outcome: plain-language, cross-checked explanations of your scan — private and exploratory, not a diagnosis. Explore your own scan: braid.health/ask-ai