Auto-cropping of phone camera color images to segment cardiac signals in ECG printouts

Digital analysis of the bio-electrical signals allows us to obtain valuable information for clinical and research purposes. Despite current devices record and store the digital electrical signal, in some cases only a printout version is available. Although previous research focused on algorithms to digitize a scanned version of the printed signal, nowadays a color image of the printout can be immediately obtained with a cell phone. Most of these algorithms digitize the signals after manually selecting from the whole image one sub-image per signal to digitize. Here, propose a procedure to automatically crop the image to get as many sub-images as signals are in the ECG printout. First, the perspective distortion is corrected. Then, an image I mainly containing the grid, is obtained from a transformation of the RGB space to another colorspace taking advantage of the grid chromatic characteristics. The grayscale image, G, and I are morphologically processed to emphasize both signal and grid in G, and just grid in I. Boundaries between contiguous signals are set by means of projections and morphological operations. Thirty one phone camera color images of 12-lead-ECG printouts were used for evaluation. Experiments were conducted with images with different spatial resolutions and quality factors according to the jpeg compressor. Our procedure worked correctly when the tracing of consecutive leads did not overlap in the amplitude axis and the spatial resolution and the quality factor were higher than 0.25 (one-fourth of the original size) and 40%, respectively.