Introduction
Monitoring Multiple Sclerosis (MS) progression typically relies on conventional MRI metrics, such as counting new lesions or measuring total lesion volume. However, these macroscopic measures often correlate poorly with a patient's actual clinical disability. This discrepancy arises from "hidden" pathologies—subtle structural damage occurring within what appears to be Normal Appearing White Matter (NAWM) and internal changes within existing lesions that are invisible to the naked eye.
This study investigates the use of texture analysis on high-field (3T) MR images to detect these subtle pathological changes in MS patients treated with minocycline. By applying mathematical algorithms to quantify the spatial distribution of pixel intensities, researchers aimed to develop more sensitive, objective, and automated biomarkers for monitoring treatment response.
Methodology
The longitudinal pilot study followed five patients with relapsing-remitting MS (RRMS) over a seven-month period, including one month of baseline observation and six months of minocycline treatment. Minocycline, an antibiotic, was selected for its potential anti-inflammatory and neuroprotective properties, specifically its ability to inhibit matrix metalloproteinases (MMPs).
Imaging and ROI Selection
Images were acquired using a 3T MRI scanner, which offers a higher signal-to-noise ratio than standard 1.5T clinical scanners. Radiologists identified three distinct types of Regions of Interest (ROIs) for analysis:
- Active lesions
- Inactive lesions
- Normal Appearing White Matter (NAWM)
Texture Extraction
The researchers employed Haralick’s method, a statistical approach to texture analysis that utilizes Gray-Level Co-occurrence Matrices (GLCM). These matrices describe the frequency with which pairs of pixels with specific values occur at defined distances and orientations. The study calculated several key features:
- Angular Second Moment (ASM): A measure of local homogeneity.
- Contrast (CON): A measure of local variations.
- Entropy (ENT): A measure of image complexity or randomness.
- Correlation (COR), Mean Gray Level (MGL), and Variance of Gray Level (VGL).
Results and Discussion
The analysis revealed that texture features changed across all ROI categories, confirming that tissue appearing "normal" to a radiologist undergoes measurable structural shifts during treatment.
Sensitivity of Active Lesions and NAWM
The most dynamic changes were observed in active lesions. For instance, the mean Contrast (CON) in active lesions increased by approximately 33% following treatment. Interestingly, the magnitude of texture change in NAWM was greater than that in inactive lesions, suggesting that NAWM is a critical, yet often overlooked, indicator of global brain health and treatment efficacy.
Structural Stabilization
Following minocycline treatment, features like ASM (homogeneity) generally increased, while ENT (complexity) decreased in active lesions. These shifts suggest a transition toward a more organized and stable tissue structure, potentially reflecting reduced inflammation or early repair processes. The close relationship between texture changes in active lesions and NAWM suggests that the treatment's effect is not merely localized but represents a systemic improvement in brain tissue integrity.
Conclusion
Texture analysis of 3T MR images provides a sensitive means of quantifying subtle tissue changes in MS that are not captured by conventional imaging. While this pilot study was limited by a small sample size ($n=5$), the results demonstrate the feasibility of using mathematical texture features as lead indicators for treatment response. Future work involving larger cohorts and histological validation could establish these features as robust biomarkers for clinical trials, potentially allowing for faster and more accurate assessment of new MS therapies.
Summary updated