U-Net Guided Digitization of 12-Lead Printed ECGs

Team Berru's Widows presents a solution for the Digitization of ECG Images in The George B. Moody Physionet Challenge 2024.Although we were not ranked on the official test data, our approach integrates traditional image processing techniques with deep learning methods, which we consider relevant for the state of the art.Our pipeline consisted on a preprocessing stage, then a single lead isolation process and the conversion from image to signal.The preprocessing included shadow removal, using classic computer vision methods, and segmentation using a U-Net.The other stages included single lead isolation and signal digitization.