Region-Weighted Losses and Model Fusion for Cross-Modal PET Attenuation Correction
Authors: Khoa Tuan Nguyen, Joris Vankerschaver, Wesley De Neve
Organizations: IDLab, ELIS, Ghent University, Belgium · Center for Biosystems and Biotech Data Science, Ghent University Global Campus, Korea · Dept. of Mathematics, Computer Science and Statistics, Ghent University, Belgium
We describe our approach to the Big Cross-Modal Attenuation Correction (BIC-MAC) challenge, which asks for a pseudo-CT in Hounsfield Units to be synthesized from Non-Attenuation-Corrected PET (NAC-PET), DIXON MRI and a topogram, and scores both the pseudo-CT and the Attenuation-Corrected PET (AC-PET) reconstructed from it. Three ideas carried our improvements over the organizers' 3D U-Net baseline. The loss matters more than the architecture: we compute the L1 error in the Carney attenuation-coefficient (μ) space that the CT metric itself uses, weighted by anatomical region. Only once that loss was in place did the unregistered DIXON MRI work as extra input channels. A fixed convex combination of two independently trained models then beat both of its members on three of the four metrics and ranks first overall on the public validation leaderboard.
Figures & tables
Figure 1: The two properties of the fixed reconstruction pipeline that shaped our loss. (a) The Carney bilinear HU-to- μ map [ 3 ] the CT metric is computed in. Its knee at 47 HU makes one HU of error worth 1.88× more μ below the knee than above, which is why we optimize in μ space rather than in HU; the shaded band marks the >300 HU region our loss weights ×3 . (b, c) A 110 mm window on the ribs and spine of training subject sub-000 , before and after the 4 mm FWHM smoothing the pipeline applies before forward projection. (d) Profile along the red line. Smoothing removes 3.6 % of the mean μ in this window and cuts the profile’s peak-to-peak range from 77 to 37×10−3 cm -1 , halving the narrow rib peaks. Anything a model gets right at this scale is invisible to the three AC-PET metrics and visible only to the CT metric, which is the tension every design choice below addresses. Both the conversion and the blur are computed with the organizers’ own code.
#
Model
Configuration
CT μ -MAE ↓
SUV MAE ↓
Organ ↓
Brain ↓
1
Baseline
plain L1 , NAC-PET only
0.006610
0.0586
4.56
0.0541
2
MuNet2
Lμ+Lgrad+LAnPer
0.005803
0.0377
2.74
0.0148
3
+LNGF
0.005972
0.0352
2.58
0.0158
4
+LLOR
0.005958
0.0345
2.45
0.0168
5
MuNet4
Lμ+Lgrad+LAnPer
0.005679
0.0383
2.80
0.0168
6
+LLOR
0.005703
0.0339
2.38
0.0078
Table 1: Validation-set progression, at the precision the leaderboard reports. Each row adds the change in the Configuration column to the row above it within its block. Best per column in bold . Row 9 is the model we submit. Row 7’s three AC-PET entries are marked -- because the challenge’s validation phase closed before we scored that model on its own, and those metrics can only be produced by the organizers’ service. Its CT μ -MAE is measured, and its effect on all four metrics is visible in row 9.
We participated in the BIC-MAC Challenge with a multimodal 3D patch-based U-Net for pseudo-CT generation from NAC-PET, MRI, and 2D topograms. By using separate PET and MR encoders, multi-scale feature fusion, and FiLM-based topogram conditioning at the bottleneck, we obtain a model that integrates complementary cross-modal information while reducing reliance on precise voxel-wise correspondence between modalities. Our final submission can be found: https://github.com/rrr-uom-projects/BIC-MAC-MICCAI2026
Rory Bell, Artemis Bouzaki, Jiaming Cao +2
Radiotherapy Related Research, Division of Cancer Sciences, School of Medical Sciences, The University of Manchester, Manchester, United Kingdom
This report describes our submission to the Big Cross-Modal Attenuation Correction (BIC-MAC) 2026 Challenge for CT-less PET attenuation correction through multimodal pseudo-CT synthesis. We build upon a standard nnU-Net architecture and combine anatomical and physical supervision to improve both pseudo-CT quality and downstream PET reconstruction. Anatomical supervision is introduced through a frozen TotalSegmentator feature extractor, anatomy-guided structural constraints and patch sampling, while physical supervision is achieved using a differentiable attenuation correction factor projection loss based on multi-angle attenuation projections. Furthermore, the network is initialized with pretrained weights obtained from training on the SynthRAD Challenge MR-to-CT dataset. Minimal architectural modifications are applied, while performance improvements are pursued across the nnU-Net pipeline, including preprocessing, plans, and supervision design, among other components. Our final submission demonstrates the effectiveness of combining anatomical supervision, attenuation physics, and efficient nnU-Net scaling for CT-less PET attenuation correction.
Petros Chatzitoulousis, George K. Matsopoulos
Biomedical Engineering Laboratory, National Technical University of Athens, Athens, Greece
The BIC-MAC challenge targets whole-body pseudo-CT synthesis from NAC-PET, Dixon MRI, and a 2D topogram for CT-less PET attenuation correction. We propose GeoPACT, a geometry-anchored multimodal framework that uses NAC-PET as the spatial reference and incorporates topogram and MRI features through gated residual fusion. Absolute coordinates and whole-body conditioning support anatomically consistent patch-based prediction. Training combines attenuation-map supervision with a differentiable PET-response surrogate to reduce errors relevant to downstream PET reconstruction. Full-resolution pseudo-CT volumes are generated using sliding-window inference without requiring CT or PET labels at test time.
Xuan Loc Nguyen, Hoang-Loc Cao, Truong Thanh Hung Nguyen +4
University of Science, VNU-HCM, Ho Chi Minh City, Vietnam · Analytics Everywhere Lab, University of New Brunswick, Canada · Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany