🤖 AI Summary
This study addresses the significant anisotropy in non-contrast head CT scans caused by large inter-slice spacing (2–5 mm), which adversely affects multiplanar reconstruction, hematoma volume estimation, and downstream algorithm performance. The authors propose the first deep slice interpolation approach tailored for head CT, leveraging a neural network to synthesize intermediate axial slices from neighboring ones, thereby simultaneously mitigating anisotropy and noise in a single inference pass. Through systematic evaluation of various loss function combinations, they identify MS-SSIM+L1 as optimal and uncover training instabilities associated with SSIM-based losses along with effective mitigation strategies. Experiments demonstrate that the proposed method substantially outperforms conventional interpolation techniques and state-of-the-art video frame interpolation models—such as RIFE and FILM—on both internal test sets and external datasets, while also enhancing 3D visualization quality and achieving implicit denoising.
📝 Abstract
Head computed tomography (CT) typically uses sub-millimeter in-plane resolution but 2-5 mm through-plane spacing, creating substantial anisotropy that degrades multiplanar reconstructions, volumetric measurements such as hematoma volume estimation, and downstream algorithms that assume near-isotropic voxels. We present a deep learning system that synthesizes intermediate CT slices from pairs of neighboring axial slices, halving the effective through-plane spacing. The system improves three-dimensional visualization while simultaneously producing inherently denoised outputs, yielding two complementary benefits from a single inference pass.
To build a reliable system, we systematically evaluate pixel-wise losses, namely mean squared error (MSE) and mean absolute error (L1); structural-similarity losses, namely the structural similarity index (SSIM) and its multi-scale variant (MS-SSIM); and hybrid combinations. On a held-out test set, all converged models outperform classical interpolation baselines and pretrained video frame interpolation methods (RIFE, FILM) on all structural measures, with MS-SSIM+L1 offering the strongest balanced profile. We also document training instability in SSIM-family losses and identify partial remedies: the standard numerical fixes eliminate the dominant failure mode but leave residual divergence at smaller batch sizes. All results are reported with patient-level bootstrap confidence intervals and paired statistical tests.
As an illustration, we apply the system to an out-of-distribution head CT series from Hospital Universitario Virgen del Rocío: the model synthesizes intermediate slices and exhibits on the real slices the implicit-denoising signature predicted by our theoretical analysis, supporting in a single external case that interpolation quality and implicit denoising are not confined to the training distribution.