We present a novel method for volumetric left ventricle mesh personalisation from cardiac MR images. The proposed method does not require any ground-truth mesh training data. Additionally, it corrects for slice misalignment and can propagate these correction back to the original image data. The method is expressive enough to capture diverse morphology, and is also differentiable, allowing for direct inclusion in deep-learning pipelines. We demonstrate that our mesh personalisation approach works robustly on both healthy and pathological anatomy.
This abstract and the presentation materials are available to members only; a login is required.