Brain MRI Intensity Normalization Using Dual Tissue References
Poster No:
1503
Submission Type:
Abstract Submission
Authors:
Silu Zhang1, Qing Ji1, Joseph Holtrop1, Asim Bag1, Nicholas Phillips1, Matthew Scoggins1
Institutions:
1St. Jude Children's Research Hospital, Memphis, TN
First Author:
Co-Author(s):
Introduction:
Intensity normalization is crucial for quantitative analysis of brain MRI, particularly for radiomic analysis, where it is important to ensure that the extracted features reflect true differences in tumor characteristics rather than variations due to MRI acquisition parameters. Traditional normalization methods, such as z-score normalization (Moore and McCabe, 1989), standardize intensity values based on the mean and standard deviation of the entire image or brain-extracted image. However, this approach assumes uniform global statistics across subjects, which may not hold true due to regional variability, especially in pathological cases. Another approach, white-stripe normalization (Shinohara et al.), uses white matter (WM) intensity statistics to ensure consistent normalization in WM, but it may introduce variability in other tissue types that have intensity values different from WM. To address these limitations, we propose a novel dual-tissue reference approach using both WM and cerebrospinal fluid (CSF) as reference points to improve the robustness and accuracy of intensity normalization in brain MRI.
Methods:
We used a normalization method involving a linear transformation using both WM and CSF as reference points. Specifically, the transformation is defined by the following equations: I'i = ai⋅Ii + bi, ai = (ÎWM - ÎCSF)/(µiWM - µiCSF), bi = (ÎCSF·µiWM - ÎWM·µiCSF)/(µiWM - µiCSF), where I'i is the normalized intensity for subject i, Ii is the original intensity for subject i, ÎWM and ÎCSF represent the median WM and CSF intensities calculated across all subjects, while µiWM and µiCSF represent the median WM and CSF intensities for each individual subject. This linear transformation ensures that both WM and CSF intensities are mapped to consistent values across subjects.
We tested the performance of different normalization methods on preoperative brain MRI scans from the SJMB12 protocol. The dataset has a total number of 13733 MRI images including T1-weighted pre- and post-contrast, T2-weighted, Fluid-Attenuated Inversion Recovery (FLAIR), and Apparent Diffusion Coefficient (ADC), acquired with various scanning parameters from multiple institutions. WM and CSF masks were generated using a knowledge-based brain tissue segmentation algorithm (Zhang et al., 2021). For comparison, we implemented single-reference normalization using WM only and z-score normalization applied to the whole brain or brain-extracted images.
Given that the tumor is typically the region of interest for radiomic analysis, we evaluated the performance of the different normalization methods by calculating the signal-to-noise ratio (SNR) of the tumor, which is defined by SNRdB = 10log10(3·σ2tumor/(σ2CSF+σ2WM+σ2GM)). Tumor masks were first generated by a cascaded CNN model (Wang et al., 2018) and then manually edited by radiologists.
We tested the performance of different normalization methods on preoperative brain MRI scans from the SJMB12 protocol. The dataset has a total number of 13733 MRI images including T1-weighted pre- and post-contrast, T2-weighted, Fluid-Attenuated Inversion Recovery (FLAIR), and Apparent Diffusion Coefficient (ADC), acquired with various scanning parameters from multiple institutions. WM and CSF masks were generated using a knowledge-based brain tissue segmentation algorithm (Zhang et al., 2021). For comparison, we implemented single-reference normalization using WM only and z-score normalization applied to the whole brain or brain-extracted images.
Given that the tumor is typically the region of interest for radiomic analysis, we evaluated the performance of the different normalization methods by calculating the signal-to-noise ratio (SNR) of the tumor, which is defined by SNRdB = 10log10(3·σ2tumor/(σ2CSF+σ2WM+σ2GM)). Tumor masks were first generated by a cascaded CNN model (Wang et al., 2018) and then manually edited by radiologists.
Results:
As shown in Table 1, the dual tissue reference method demonstrated superior SNR of the tumor compared to other normalization methods for all five MRI contrasts. Figure 1 illustrates the histograms of the images before and after normalization, showing reduced variability and more consistent intensity distribution after applying the dual tissue reference normalization. These findings suggest that incorporating CSF as an additional reference point provides a more stable normalization framework, particularly in datasets with high anatomical variability.
·Table 1. Signal to Noise Ratio (dB) of Tumor
·Figure 1. Histograms of image before and after normalization.
Conclusions:
The proposed dual-tissue reference normalization method enhances the robustness and reliability of brain MRI intensity normalization.
Modeling and Analysis Methods:
Methods Development 1
Other Methods 2
Keywords:
Data analysis
MRI
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Shinohara, R. T., Sweeney, E. M., Goldsmith, J., Shiee, N., Mateen, F. J., Calabresi, P. A., ... & Alzheimer's Disease Neuroimaging Initiative. (2014). Statistical normalization techniques for magnetic resonance imaging. NeuroImage: Clinical, 6, 9-19.
Wang, G., Li, W., Ourselin, S., & Vercauteren, T. (2018). Automatic brain tumor segmentation using cascaded anisotropic convolutional neural networks. In Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries: Third International Workshop, BrainLes 2017, Held in Conjunction with MICCAI 2017, Quebec City, QC, Canada, September 14, 2017, Revised Selected Papers 3 (pp. 178-190). Springer International Publishing.
Zhang, S., Edwards, A., Wang, S., Patay, Z., Bag, A., & Scoggins, M. A. (2021). A prior knowledge based tumor and tumoral subregion segmentation tool for pediatric brain tumors. arXiv preprint arXiv:2109.14775.