Abstract
<title>Abstract</title> <p>The problem of accurate reconstruction of the mitochondrial shape is important for understanding the life-sustaining mechanisms at the cellular level that are indicative of many illnesses and poor health conditions. Current approaches are restricted to the problem of 2D segmentation of fluorescent images and do not address the inherent 3D structure. We present MiShape, a 3D shape prior for modeling mitochondrial shapes in fluorescence microscopy images. MiShape is representative of the real distribution of mitochondrial shapes as it is trained using shapes sourced from high-resolution Electron Microscopy images. MiShape uses an implicit representation to model shapes, offering a versatile solution for accurately representing the diverse shapes of sub-cellular structures. To our knowledge, MiShape is the first framework to use a learnt 3D mitochondrial shape prior to reconstruct 3D mitochondrial morphology from a single 2D fluorescence microscopy image. Across simulated, external, and real microscopy datasets, MiShape outperforms conventional and state-of-the-art baselines in terms of geometry-based and topological reconstruction metrics.</p>