Introduction: Myocardial strain imaging provides a complex dataset that is often reduced to a single parameter, global longitudinal peak strain (GLPS). Raw temporal strain data is largely unexplored and restricted as studies often only save images of graphs lacking numerical data. While certain strain imaging packages do allow for the prospective export of raw data, the task is slow, rarely done, and cannot be done retrospectively. The troves of retrospective strain studies saved as images can be unlocked by reconstructing the numerical values contained in strain graphs. Such a dataset would be critical to future artificial intelligence strain analysis and morphological detection of disease modalities. This study aims to investigate the methods of extracting and reconstructing strain curve graphs to raw numerical outputs. Methods: By using vendor-specific color recognition algorithms, individual curves can be isolated as scattergram type plots of pixels. These pixel locations are converted to temporal strain values by defining scale factors for the graph. Intelligence-guided filtering, interpolation, and curve overlap filling can convert the pixel scattergrams into clean and complete temporal strain data. Using a set of 21 studies each containing 3 views (2Ch, 4Ch, APLAx) with each view containing 6 regional curves, 378 unique curves were reconstructed and compared to prospectively exported temporal strain data. Results: Results demonstrate that this method reconstructs the numerical data with high accuracy and precision. The reconstructed data was compared with the exported numerical data, yielding an average R 2 of .996 with an average runtime per study of 7.32s. The minute deviation from fidelity can be attributed to the finite resolution of the pixels in the analyzed graph. Discussion: Highly accurate temporal data can be extracted from routine strain graphs in clinical studies, which may allow for more nuanced assessment and deep learning of strain.