Application Value of Intelligent Quick Magnetic Resonance for Accelerating Brain MR Scanning and Improving Image Quality in Acute Ischemic Stroke

Bo Xue, Dewen Duan, Junbang Feng, Zhenjun Zhao, Jennifer Tan, Jinrui Zhang, Chao Peng, Chang Li, Chuanming LiCurrent 2025 · DOI

TL;DR

This paper introduces an AI-powered post-processing technique called Intelligent Quick Magnetic Resonance (IQMR) that significantly reduces brain MRI scan times for acute ischemic stroke patients by almost half, while simultaneously improving image quality. This advancement holds promise for faster diagnosis and more effective treatment in critical stroke cases.

What Problem Does This Paper Address?

Acute ischemic stroke is a time-sensitive medical emergency where rapid diagnosis and treatment are crucial to minimize disability and mortality. Magnetic Resonance Imaging (MRI) is a powerful tool for detecting and assessing these strokes, offering superior soft-tissue resolution compared to CT scans. However, conventional MRI sequences are notoriously slow, leading to prolonged examination times. This slowness poses a significant challenge, especially for patients who may have difficulty remaining still for extended periods, increasing the risk of motion artifacts and potentially delaying critical treatment decisions. Existing methods to speed up MRI scans often compromise image quality or require specialized hardware, creating a need for a solution that can accelerate scans without sacrificing diagnostic accuracy.

Methods

The core innovation presented is the Intelligent Quick Magnetic Resonance (IQMR) system. IQMR is an AI-driven post-processing algorithm that works on existing MRI data without requiring any changes to MRI hardware or software. Its methodology involves several key steps:

  • Iterative Image Reconstruction (IIR): This is a volumetric algorithm that retrospectively reconstructs MRI scans. It breaks down the image data into small 3D patches, separates signal from noise, and iteratively refines the image until optimal quality is achieved. This process is designed to significantly reduce noise and restore fine details.
  • Edge Enhancement: IQMR algorithms sharpen image edges by increasing contrast around them, improving the perceived sharpness of the image without affecting uniform areas.
  • AI-based Enhancement (CNN): The system utilizes a convolutional neural network (CNN), specifically a modified SRGAN architecture, to enhance small image details and improve overall image sharpness. This CNN is trained on a massive dataset of MRI scans from various vendors and clinical settings, allowing it to learn a wide range of image qualities and anatomical variations.
  • Weighted Averaging: To ensure the denoised images appear natural to radiologists, IQMR employs a weighted average of the processed and original images, incorporating user-defined preferences for sharpness.

The IQMR processing is applied to accelerated MRI sequences (T1WI, T2WI, and T2-FLAIR) that were acquired with reduced averaging.

Datasets

The study utilized data from 58 patients diagnosed with acute ischemic stroke. These patients underwent head MRI examinations between July 2023 and January 2024. The paper does not mention the use of any external or publicly available datasets for training the IQMR algorithm; it implies the algorithm was pre-trained by the developers (Medic Vision Imaging Solutions) on a large, proprietary dataset.

Key Insights and Ideas

  • AI as a Post-Processing Solution: The study highlights the power of AI not just for image acquisition but also for post-processing to overcome limitations of existing hardware. IQMR's ability to improve image quality and speed without hardware modifications makes it highly adaptable.
  • Balancing Speed and Quality: The research demonstrates that by combining accelerated acquisition with intelligent post-processing, it's possible to achieve both significantly reduced scan times and improved image quality, a balance that has been challenging with previous methods.
  • Clinical Relevance for Stroke: The findings are particularly impactful for acute ischemic stroke, where every minute counts. The substantial reduction in scan time directly translates to faster diagnosis and potentially better patient outcomes.
  • Generalizability: The IQMR algorithm's training on a diverse dataset suggests its potential applicability across different MRI scanners and clinical scenarios, which is a significant advantage for widespread adoption.

Key Results

  • Scan Time Reduction: The total acquisition time for routine T1WI, T2WI, and T2-FLAIR sequences was 5 minutes and 9 seconds. With IQMR-accelerated sequences, this was reduced to 2 minutes and 40 seconds, a 48.22% reduction. Individual sequence reductions were: T1WI (55.84%), T2WI (52.17%), and T2-FLAIR (42.95%).
  • Image Quality Improvement:
    • Quantitative: IQMR significantly improved the Signal-to-Noise Ratio (SNR) and Contrast-to-Noise Ratio (CNR) of lesions and surrounding tissues in accelerated sequences (P < .05). Crucially, these improved metrics achieved parity with routine sequences (P > .05).
  • Qualitative: IQMR significantly improved lesion conspicuity, lesion internal display, and overall image quality (P < .05). Artifacts were not significantly improved by IQMR processing (P > .05).
  • Diagnostic Consistency: The Alberta Stroke Program Early CT Score (ASPECTS), used to evaluate ischemia severity, showed no significant difference between IQMR and routine images (P = 0.79), indicating that IQMR does not compromise diagnostic assessment of stroke severity. Intra-reader reliability for ASPECTS was strong (ICC: 0.91-0.93).

Limitations and Future Work

The authors acknowledge several limitations:

  • Limited Sequences: The study only evaluated DWI, T1WI, T2WI, and T2-FLAIR sequences. Specialized functional sequences like SWI and PWI were not assessed.
  • Small Sample Size: The study included 58 patients, which is considered a relatively small sample size.
  • No Dynamic Tracking: Dynamic tracking of patient movement was not conducted.
  • Single Scanner Field Strength: The analysis was limited to images from a 1.5T MRI system, excluding evaluations on lower-field-strength images.

Future work suggested includes:

  • Incorporating a larger sample size.
  • Evaluating additional MRI sequences.
  • Conducting dynamic follow-up studies.
  • Assessing the performance of IQMR on MRI systems with different field strengths.
  • Exploring the clinical utility of IQMR in a broader range of neurological conditions.

Summary updated

Abstract

IQMR could significantly reduce brain MRI scanning time and enhance image quality in patients with acute ischemic stroke.
BibTeX
@inproceedings{xue2025application,
  author = {Bo Xue and Dewen Duan and Junbang Feng and Zhenjun Zhao and Jennifer Tan and Jinrui Zhang and Chao Peng and Chang Li and Chuanming Li},
  title = {{Application Value of Intelligent Quick Magnetic Resonance for Accelerating Brain MR Scanning and Improving Image Quality in Acute Ischemic Stroke}},
  booktitle = {Current},
  year = {2025},
  doi = {10.2174/0115734056418245250912095159},
}
Figure