Domain-Agnostic Stroke Lesion Segmentation Using Physics-Constrained Synthetic Data

2024-12-04

arXiv.org

AI Summary

Multi-center MRI acquisition protocol variations severely impair the generalizability of stroke lesion segmentation models. To address this, we propose two quantitative MRI (qMRI)-guided synthetic data generation methods: (1) estimating qMRI parameters from MPRAGE images and synthesizing multi-sequence MRI volumes via forward simulation; and (2) inverting a qMRI atlas from tissue-label maps. Both methods explicitly embed the underlying qMRI biophysical signal model into the synthesis pipeline—marking the first such integration for domain-agnostic feature learning. We integrate these synthetic data with a U-Net variant for image-to-image mapping and the nnUNet benchmark framework. Extensive evaluation across multiple out-of-distribution test sets demonstrates consistent superiority over the nnUNet baseline. Notably, Method 2 achieves Dice score improvements of 3.2–5.7 percentage points over prior synthesis-based approaches, empirically validating the efficacy of physics-informed synthesis for robust, multi-center stroke lesion segmentation.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Web Mining and Content Analysis: Web data generation and simulation
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Segmenting stroke lesions in Magnetic Resonance Imaging (MRI) is challenging due to diverse clinical imaging domains, with existing models struggling to generalise across different MRI acquisition parameters and sequences. In this work, we propose two novel physics-constrained approaches using synthetic quantitative MRI (qMRI) images to enhance the robustness and generalisability of segmentation models. We trained a qMRI estimation model to predict qMRI maps from MPRAGE images, which were used to simulate diverse MRI sequences for segmentation training. A second approach built upon prior work in synthetic data for stroke lesion segmentation, generating qMRI maps from a dataset of tissue labels. The proposed approaches improved over the baseline nnUNet on a variety of out-of-distribution datasets, with the second approach outperforming the prior synthetic data method.

Problem

• Improving stroke lesion segmentation across diverse MRI protocols
• Generating synthetic qMRI images with physics constraints
• Enhancing model generalizability for heterogeneous domains

Innovation

• Physics-constrained synthetic qMRI generation
• Neural network estimates qMRI from MPRAGE
• Label-conditioned Gaussian mixture models