Equivariant Multiscale Learned Invertible Reconstruction for Cone Beam CT: From Simulated to Real Data

2025-12-24

AI Summary

Cone-beam CT (CBCT) suffers from low image quality, scarcity of real paired training data, and stringent clinical real-time requirements. Method: We propose LIRE++, an end-to-end rotationally equivariant, multi-scale, invertible primal-dual reconstruction framework. It introduces a novel design paradigm integrating rotational equivariance, multi-scale modeling, and invertible network architecture; employs a quasi-Monte Carlo CBCT projection simulator for high-fidelity synthetic training data generation; and adopts memory-aware training and inference to drastically reduce GPU memory consumption. Contribution/Results: On real clinical CBCT data, LIRE++ achieves the first quantitative improvement over planning CT: average MAE reduction of 10 HU versus the current state-of-the-art hybrid method, and +1 dB PSNR on synthetic data. The framework delivers fast, memory-efficient, and parameter-light deep reconstruction—establishing a new paradigm for clinical CBCT deployment.

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: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Cone Beam CT (CBCT) is an important imaging modality nowadays, however lower image quality of CBCT compared to more conventional Computed Tomography (CT) remains a limiting factor in CBCT applications. Deep learning reconstruction methods are a promising alternative to classical analytical and iterative reconstruction methods, but applying such methods to CBCT is often difficult due to the lack of ground truth data, memory limitations and the need for fast inference at clinically-relevant resolutions. In this work we propose LIRE++, an end-to-end rotationally-equivariant multiscale learned invertible primal-dual scheme for fast and memory-efficient CBCT reconstruction. Memory optimizations and multiscale reconstruction allow for fast training and inference, while rotational equivariance improves parameter efficiency. LIRE++ was trained on simulated projection data from a fast quasi-Monte Carlo CBCT projection simulator that we developed as well. Evaluated on synthetic data, LIRE++ gave an average improvement of 1 dB in Peak Signal-to-Noise Ratio over alternative deep learning baselines. On real clinical data, LIRE++ improved the average Mean Absolute Error between the reconstruction and the corresponding planning CT by 10 Hounsfield Units with respect to current proprietary state-of-the-art hybrid deep-learning/iterative method.

Problem

• Develops an equivariant multiscale invertible network for CBCT reconstruction
• Addresses memory constraints and lack of ground truth in deep learning CBCT methods
• Improves image quality over existing techniques on both simulated and real data

Innovation

• Rotationally-equivariant multiscale learned invertible primal-dual scheme
• Memory optimizations and multiscale reconstruction for fast training
• Trained on simulated data from quasi-Monte Carlo CBCT simulator