Are Pixel-Wise Metrics Reliable for Computerized Tomography Reconstruction?

Code, dataset, and models: https://github.com/MrGiovanni/CARE NeRF R 2 -GS +CARE Ground Truth Gall bladder Spleen Celiac trunk Aorta Colon Postcava Left Kidney R 2 -GS NeRF +CARE +29.0% +30.0% +31.7% +8.7% +8.6% +20.3% +21.5% +3.7% +35.7% +22.3% +14.0% +2.5% +21.9% +31.0% +40.2% +17.2%

Figure 1: CARE Improves Structural Completeness in Sparse-View CT Reconstruction. Top: Qualitative comparison of reconstructed CT scans from NeRF and R 2 -GS, with and without our proposed CARE on clinically important structures. Bottom: Quantitative gains in structural completeness across four reconstruction methods. CARE consistently improves results, confirming its model-agnostic and plug-and-play nature.