ALI B. KARIMI

Classification of Cardiac Rhythms from ECG Signals with Deep Learning Models

focused on implementing and comparing four different deep learning (DL) models to classify cardiac abnormalities (CA) from 12-Lead ECG recordings. These CA's are associated with irregular cardiac rhythms. algorithms based on recently popular architectures, CNN, LSTM, and ResNet, to perform multi-label classification. The proposed models were calibrated on the available 40,000+ ECGs and the test set consist of 10,000+ recordings. The motivation came from the PhysioNet 2021 Challenge focusing on the automatic classification of cardiac abnormalities from 12-lead ECG signals.

Check out the project and codebase here

ECG Signal Conditioning by Morphological Filtering

Clinically collected electrocardiographic (ECG) signals are often contaminated with a lot of noise and to obtain information from the noisy signals is always challenging. In order to facilitate automated ECG alanysis, Signal Conditioning is a necessity. In this project I implemented a morphological filter technique for the baseline correction and noise suppression with minimum signal distortion.
Data set is gathered from physionet.org
In order to keep eye on erros, we generated a synthesised ECG signal with specific known characteristics and then by knowing the result and mean square error on that better produce reults for real data sets.

Synthesising Human Iris by Means of Generative Models (GAN)

Research and implementation of generating human iris by Generative Adversarial Networks in an unsuprevised way. The list of papers I gathered which are helping me the way are listed on my github.
Also I summerized each of the papers in one or two pages so might be easier to go through. You can find it also along with other papers, called: summary

Top Rated Upwork Freelancer

LinkedIn has the most updated version of my study/career CV