Causal PETS: Causality-Informed PET Synthesis from Multi-modal Data

Yujia Li, Han Li, S Kevin Zhou

Keywords: medical image synthesis, causality

TL;DR: We propose a model that analyzes the causality between different modality data and seamlessly integrates such causality into PET image generation.

Abstract: The synthesis of medical images is particularly important when certain modality data are difffcult to obtain, for example, Positron emission tomography (PET). PET is crucial for diagnosing and monitoring neurological disorders. However, the availability is limited due to factors such as high costs, radiation exposure risks, and other constraints. In this study, we propose Causal PETS, a novel causality-informed synthesis model for synthesizing PET images from multi-modal data including MRI, demographic information, and cerebrospinal fluid (CSF) biomarkers. Unlike conventional approaches that involve a straightforward conversion from T1 to PET, our model analyzes the causality between different modality data and seamlessly integrates such causality into PET image generation. Through comprehensive evaluations, we demonstrate that our Causal PETS model outperforms existing non-causal methods in terms of image clarity and accuracy, particularly in identifying regions of interest critical for neurological disorders such as Alzheimer’s Disease (AD). This work underscores the importance of causal reasoning in medical image synthesis and highlights the potential of multimodal integration to advance clinical decision making.

Primary Subject Area: Image Synthesis

Secondary Subject Area: Causality

Paper Type: Validation or Application

Registration Requirement: Yes

Visa & Travel: Yes

Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Created a single midl25_NNN.zip file with midl25_NNN.tex, midl25_NNN.bib, all necessary figures and files., Includes \documentclass{midl}, \jmlryear{2025}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package, Did not use the times package., All authors and co-authors are correctly listed with proper spelling and avoid Unicode characters., Author and institution details are de-anonymized where needed. All author names, affiliations, and paper title are correctly spelled and capitalized in the biography section., References must use the .bib file. Did not override the bibliographystyle defined in midl.cls. Did not use \begin{thebibliography} directly to insert references., Tables and figures do not overflow margins; avoid using \scalebox; used \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., Appendices and supplementary material are included in the same PDF after references., Main paper does not exceed 9 pages; acknowledgements, references, and appendix start on page 10 or later.

Submission Number: 5