Image accuracy does not guarantee correct 3D physical reconstructions
Image Fidelity is Not Field Fidelity: Joint Thermodynamic Reconstruction and Error Localization in Neural Tomography
Machine Learning
Summary
This paper shows that when creating 3D models of physical objects from 2D images, getting very accurate images doesn’t always mean the 3D model is correct. The researchers developed a method called CoroNeRF that uses multiple views and multiple types of light to reconstruct certain physical properties like electron density and temperature in the solar corona. They discovered that sometimes a model can look right in 2D but still be wrong in 3D. They also found ways to estimate errors in the 3D models without needing ground truth data by comparing multiple runs of the method.
What this means in practice
- •For solar physicists: Improve accuracy assessment of 3D solar corona reconstructions from line-of-sight images by jointly reconstructing multiple physical fields and detecting errors without ground truth.
- •For medical imaging specialists: Adopt joint multi-parameter reconstruction and error localization strategies from this work to improve 3D tomography quality assurance in complex multi-modal imaging applications.
Authors
Alan Hsu, Jenna Samra, Alin Razvan Paraschiv, Liam Connor
Abstract
Neural fields for scientific tomography are optimized from 2D images, but the actual quantity of interest is often a latent 3D physical field. Because the forward map is many-to-one, low 2D image error need not certify a correct 3D field. Moreover, the latent field is not directly supervised during training, and its error cannot be evaluated against truth at deployment. We develop CoroNeRF to jointly optimize 3D electron density and temperature fields directly from multiview, multiline intensities through a differentiable atomic-emission renderer. Using solar coronal tomography as a controlled testbed, we evaluate physical-field recovery and test whether cross-seed instability provides a ground-truth-free-at-inference indicator of local physical-field error. We underscore the following two observations. (i) Image fidelity is not field fidelity: spectral ablations show that limited-channel reconstructions can fit their available observations well while recovering substantially worse fields, whereas evaluation on a common richer probe exposes the discrepancy. (ii) Cross-seed instability ranks local physical-field error across tested matched-model conditions, supported by sparsification and physical signal-strength controls. Seed-deviation projections provide complementary directional validation, but shared forward-model mismatch can still produce incorrect cross-seed consensus. These results characterize joint thermodynamic recovery and the usefulness and limits of seed-based error localization in a controlled, single-scene solar tomography testbed.