Healthcare
AI That Reads Heart Scans Faster Than a Doctor
mlj.solutions
mlj.solutions Kardiologické vyšetření v Bangladéši. Stejná přesnost jako v Curychu. Zadarmo. ODySSeI je open-source AI, která analyzuje snímky srdečních tepen během sekund. Detekuje léze, odhaduje závažnost zúžení. Testovaná na 2149 pacientech z Evropy, Ameriky i Asie. Cíl není nahradit lékaře. Cíl je dát mu nástroj, který vidí vždy stejně. 👉 Kde si myslíš, že AI v medicíně pomáhá nejvíc?
Zobrazit na Instagramumlj.solutions
mlj.solutions Kardiologické vyšetření v Bangladéši. Stejná přesnost jako v Curychu. Zadarmo. 🫀 Nová AI dokáže analyzovat EKG se stejnou přesností jako špičkový kardiolog a to kdekoliv na světě. Bez čekacích lhůt. Bez nákladů. Bez rozdílu, kde žiješ. Je tohle budoucnost zdravotnictví? Celé video ve čtvrtek. 🎬 👉Zažili jste někdy situaci, kdy jste neměli přístup ke správné péči?
Zobrazit na InstagramuWhat the study is about
Interpretation of coronary angiographies is the clinical gold standard for diagnosing coronary artery disease, but results vary depending on who reads the scan. ODySSeI is an open-source framework that automates the entire process: it detects lesions, segments them, and estimates their severity in real time, without the subjective judgment of a physician.
Key numbers
2,149
patients from Europe, North America, and Asia
The framework was trained and tested on diverse patient groups to demonstrate its ability to generalize across populations.
2.5×
higher lesion detection performance compared to baseline
The novel Pyramidal Augmentation Scheme delivered a 2.5-fold improvement in lesion detection, while segmentation improved by only 1 to 3%.
2–3 pixels
accuracy in estimating minimum lumen diameter
The deviation between predicted and actual values is small enough for the system to achieve clinically usable accuracy without the need for standard quantitative coronary angiography.
What this means
- Physician subjectivity is a real problem, the same scan can be interpreted differently by different doctors — AI eliminates this variability.
- Open-source lowers the barrier to deployment, the framework is available as a plug-and-play web interface.
- Speed is clinically relevant, processing a scan takes at most a few seconds.
- Generalization across populations is key, the model performs well on data from geographic regions outside its training set.
- Automation without accuracy is not enough, the study shows that for clinical deployment, AI must match expert-level precision.
Our take
AI in healthcare is not about replacing doctors. It's about ensuring every patient receives the same quality diagnosis — whether they are treated in Bangladesh or here in the Czech Republic.
Michal Dobrovolný
Zakladatel & CTO
Zdroj
Choudhary, A., Sun, X., Mahendiran, T., Senouf, O., Auberson, D., De Bruyne, B., Fournier, S., Muller, O., Abbé, E., Frossard, P., & Thanou, D. (2026). ODySSeI: An Open-Source End-to-End Framework for Automated Detection, Segmentation, and Severity Estimation of Lesions in Invasive Coronary Angiography Images. arXiv:2603.20021. https://arxiv.org/abs/2603.20021
Keep reading
Don't miss these posts
Want to implement AI in your company?
We'll go through your needs together and propose an approach that delivers measurable results.
Write to us
