cs.CVAug 22, 2026

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

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

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 L1L_1 error in the Carney attenuation-coefficient (μμ) 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\mu map [ 3 ] the CT metric is computed in. Its knee at 4747 HU makes one HU of error worth 1.88×1.88\times more μ\mu below the knee than above, which is why we optimize in μ\mu space rather than in HU; the shaded band marks the >300>300 HU region our loss weights ×3\times 3 . (b, c) A 110110 mm window on the ribs and spine of training subject sub-000 , before and after the 44 mm FWHM smoothing the pipeline applies before forward projection. (d) Profile along the red line. Smoothing removes 3.63.6 % of the mean μ\mu in this window and cuts the profile’s peak-to-peak range from 7777 to 37×10−337\times 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.
#ModelConfigurationCT μ\mu -MAE ↓\downarrowSUV MAE ↓\downarrowOrgan ↓\downarrowBrain ↓\downarrow
1Baselineplain L1L_{1} , NAC-PET only0.0066100.05864.560.0541
2MuNet2\text{MuNet}_{2}Lμ+Lgrad+LAnPer\mathcal{L}_{\mu}+\mathcal{L}_{\text{grad}}+\mathcal{L}_{\text{AnPer}}0.0058030.03772.740.0148
3+ LNGF+\ \mathcal{L}_{\text{NGF}}0.0059720.03522.580.0158
4+ LLOR+\ \mathcal{L}_{\text{LOR}}0.0059580.03452.450.0168
5MuNet4\text{MuNet}_{4}Lμ+Lgrad+LAnPer\mathcal{L}_{\mu}+\mathcal{L}_{\text{grad}}+\mathcal{L}_{\text{AnPer}}0.0056790.03832.800.0168
6+ LLOR+\ \mathcal{L}_{\text{LOR}}0.0057030.03392.380.0078
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\mu -MAE is measured, and its effect on all four metrics is visible in row 9.
More in this topic

Medical Image Synthesis

Medical image synthesis creates, translates, reconstructs, and edits clinical images to fill missing data and support analysis.

See all 310 papers

Explore similar work

CardsList

  1. Multimodal pseudo-CT synthesis for PET attenuation correction using separate modality encoding and topogram conditioning

    Aug 21, 2026Rory Bell, Artemis Bouzaki, Jiaming Cao +2Medical Image SynthesisMultimodal Fusion

    Radiotherapy Related Research, Division of Cancer Sciences, School of Medical Sciences, The University of Manchester, Manchester, United Kingdom

  2. Anatomical and Physical Supervision for CT-less PET Attenuation Correction: BIC-MAC 2026 Challenge

    Aug 16, 2026Petros Chatzitoulousis, George K. MatsopoulosPET/CT ImagingMedical Image Synthesis

    Biomedical Engineering Laboratory, National Technical University of Athens, Athens, Greece

  3. Geometry-anchored PET-aware multimodal pseudo-CT synthesis for whole-body attenuation correction: the BIC-MAC Challenge

    Aug 19, 2026Xuan Loc Nguyen, Hoang-Loc Cao, Truong Thanh Hung Nguyen +4Medical Image SynthesisPET/CT Imaging

    University of Science, VNU-HCM, Ho Chi Minh City, Vietnam · Analytics Everywhere Lab, University of New Brunswick, Canada · Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany