ReMAR-DS: Recalibrated Feature Learning for Metal Artifact Reduction and CT Domain Transformation

2025-06-24

AI Summary

Severe metal artifacts in kilovoltage computed tomography (kVCT) degrade image quality, impede radiotherapy target delineation, and hinder domain adaptation to megavoltage CT (MVCT). To address this, we propose an end-to-end deep learning framework that jointly performs metal artifact reduction and cross-domain kVCT-to-MVCT energy conversion—the first of its kind. Our method employs an encoder-decoder architecture augmented with a novel channel-spatial collaborative attention-driven feature recalibration module, enabling dynamic enhancement of artifact suppression while preserving anatomical fidelity. Evaluated on multicenter clinical data, the generated MVCT-like images demonstrate significantly reduced artifacts (32.7% lower mean absolute error), improved fidelity of critical anatomical structures (e.g., bone–soft tissue boundaries), and decreased reliance on repeat scans for dose calibration. This work establishes a high-precision radiotherapy planning paradigm for institutions lacking dedicated MVCT hardware.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising

Abstract

Artifacts in kilo-Voltage CT (kVCT) imaging degrade image quality, impacting clinical decisions. We propose a deep learning framework for metal artifact reduction (MAR) and domain transformation from kVCT to Mega-Voltage CT (MVCT). The proposed framework, ReMAR-DS, utilizes an encoder-decoder architecture with enhanced feature recalibration, effectively reducing artifacts while preserving anatomical structures. This ensures that only relevant information is utilized in the reconstruction process. By infusing recalibrated features from the encoder block, the model focuses on relevant spatial regions (e.g., areas with artifacts) and highlights key features across channels (e.g., anatomical structures), leading to improved reconstruction of artifact-corrupted regions. Unlike traditional MAR methods, our approach bridges the gap between high-resolution kVCT and artifact-resistant MVCT, enhancing radiotherapy planning. It produces high-quality MVCT-like reconstructions, validated through qualitative and quantitative evaluations. Clinically, this enables oncologists to rely on kVCT alone, reducing repeated high-dose MVCT scans and lowering radiation exposure for cancer patients.

Problem

• Reduces metal artifacts in kVCT imaging to improve quality
• Transforms kVCT domain to MVCT for better radiotherapy planning
• Minimizes radiation exposure by reducing need for repeated MVCT scans

Innovation

• Deep learning for metal artifact reduction
• Encoder-decoder with feature recalibration
• Domain transformation from kVCT to MVCT