A Proof-of-Concept Study of Multitask Learning for Cranial Synthetic CT Generation Across Heterogeneous MRI Field Strengths

2026-04-30

AI Summary

Current methods for synthesizing skull CT from MRI suffer from limited generalizability due to heterogeneity in MRI field strengths and acquisition protocols, hindering their clinical deployment. This work proposes a structurally coupled modular multi-task learning framework that explicitly models anatomical consistency through a deep neural network architecture integrated with cross-domain adaptation strategies to address multi-center heterogeneous MRI data. Evaluated on multiple real-world clinical datasets, the proposed method significantly outperforms existing approaches, consistently generating high-fidelity synthetic skull CT across varying field strengths and imaging protocols. The approach markedly enhances model robustness and cross-domain generalization, thereby facilitating its translation into clinical practice.

Abstract

Accurate synthesis of computed tomography (CT) images from magnetic resonance imaging (MRI) is clinically valuable for cranial applications such as attenuation correction, radiotherapy planning, and image-guided interventions. However, heterogeneity across MRI field strengths and acquisition protocols limits the generalizability of existing methods. In this study, we formulate cranial CT synthesis as a modular, structurally coupled problem and propose a deep learning framework to improve robustness across heterogeneous MRI conditions. The model is designed to adapt to variations in field strength and imaging protocols while preserving anatomical consistency. Experiments on multi-site datasets demonstrate improved performance and generalization compared with conventional approaches. The proposed method enables reliable CT synthesis across heterogeneous MRI settings, supporting broader clinical translation.

Problem

• synthetic CT
• MRI heterogeneity
• cranial imaging
• field strength variation
• image synthesis

Innovation

• multitask learning
• synthetic CT
• heterogeneous MRI
• deep learning
• cranial imaging