11/05/2025 updated
100 % available
Lead Data Scientist & AI Solution Architect | Healthcare & AI Expert
Saint Petersburg, Russian Federation
Worldwide
PhD About me
Data Scientist and AI Leader with 20+ years of experience and a PhD Specializes in end-to-end AI solution design and implementation for healthcare—from EHR data extraction to deploying machine learning models for CDS . DL and predictive modeling, with successful project delivery in cardiology
Deep Learning
- Languages & Technologies: Python, R, SQL, PySpark, TensorFlow, PyTorch, Keras, Scikit-learn, SpaCy, NLTK, Gensim, BERT, Hadoop/Spark, Cursor AI
- ML/DL Methods: Machine Learning, Deep Learning (RNN, LSTM, GRU, CNN, Transformer), NLP, Feature Engineering, Statistical Analysis, Fuzzy Logic
- Domains: Clinical Decision Support Systems (CDSS), Telemedicine, Remote Patient Monitoring, EHR/EMR, Cardiology, Medical Device Development
Experience
Professional experience10+ years
Freelancer since1990
Languages
EnglishGoodRussianNative speaker
Project history
Ejection Fraction (EF) Prediction from ECG — Web Service
URL: http://ikorsakov.ech:3333/
What the Service DoesThe service predicts left ventricular ejection fraction (EF) from standard 12-lead ECG. Ejection fraction is a key indicator of heart pumping function, typically measured by echocardiography. The service uses a machine learning model (Stanford ResNet-50 v2) trained on ECG and echocardiography data to estimate EF directly from the electrocardiogram.
What It’s For- Screening: rapid preliminary assessment of reduced EF risk without mandatory echocardiography
- Referral support: helps identify patients who need further cardiac evaluation
- Research: analysis of ECG–heart function relationships
- Education: demonstration of AI applications in cardiology
- Open http://ikorsakov.ech:3333/
- Sign in (or register)
- Upload ECG in a supported format:
- WFDB (.hea + .dat pair)
- CSV with 12 leads
- NumPy (.npy)
- EDF, MATLAB
- INKART/Bridge (.dat)
- Check quality — the service automatically assesses the ECG (length, amplitude, noise, baseline drift, etc.)
- Enter patient data (age, gender, reference EF from Echo if available)
- Run prediction — receive the EF category (Reduced / Mildly reduced / Normal) and class probabilities
- Results are saved in history; optional email delivery with full ECG as WFDB (.dat + .hea) in a zip archive
- 12 leads: I, II, III, aVR, aVL, aVF, V1–V6
- Sampling rate: typically 500 Hz; other rates supported
I led the end-to-end development of an AI-based software for predicting LVEF from standard 12-lead ECGs and successfully secured a Medical Device Certificate from the Ministry of Health.
Create a python web base scrypt to prognose a Left ventricular ejection fraction (LVEF)