Info
Background:
Superparamagnetic iron oxide nanoparticles (SPIONs) are widely utilized as negative contrast agents in magnetic resonance imaging (MRI). However, SPIONs generate negative contrast, resulting in signal hypointensities in the images. This characteristic presents a significant limitation for clinical translation, as the regions that contain SPIONs appear darker and are difficult to distinguish from spaces that exhibit low signal intensity, such as air-filled organs like the lungs or the gastrointestinal tract. At the same time, there is growing interest in the use of SPIONs as theragnostic platforms [1,2]. Consequently, there is a clear need to develop post-analytical techniques that enable reliable differentiation of SPIONs from air. Importantly, such techniques should be straightforward to implement in order to facilitate the widespread clinical translation of SPION-based technologies.

Hypothesis/Objective:
The present study aims to develop open-source post-processing techniques for MRI data to accurately detect and quantify SPION distribution, particularly on in vitro and in vivo models of inflammatory bowel disease (IBD).

Methods:
Agar phantoms with varying SPION concentrations were prepared and imaged using T2-weighted MRI sequences. To mimic increasing physiological complexity, the phantoms were embedded in different environments, ranging from polymer tubes to ex vivo porcine duodenum. Conventional MRI sequences were employed to evaluate signal intensity (I₀) and relaxation rates.

The resulting data were analyzed using multi-parameter mapping techniques (available at https://github.com/yaelsuarez/relaxation_rate_mapping) to enhance the spatial characterization of SPION distribution. These images were subsequently used to train unsupervised machine learning models. The best-performing model was selected and applied to distinguish SPIONs from air, as well as to estimate SPION concentration.

Finally, a Dextran Sulfate Sodium (DSS)-Induced Colitis model in Mice was established, and SPIONs conjugated to antibodies specific to IBD biomarkers were delivered to the mice [3]. The colon of the mice were excised and imaged in MRI. In the presence of IBD, these biomarkers are overexpressed, promoting aggregation of the antibody-conjugated SPIONs and enabling their detection. The selected unsupervised machine learning model was then used to identify the presence of SPIONs, and the results were validated using fluorescence microscopy.

Results:
The post-processing methodologies enabled reliable detection of changes in signal intensity (I₀) and relaxation rates across varying SPION concentrations, demonstrating a clear relationship between SPION presence and MRI signal alterations. These findings were consistent across increasingly complex environments, including ex vivo porcine tissue. Multi-parameter mapping improved spatial characterization and supported unsupervised machine learning models, with the best-performing model successfully distinguishing SPIONs from air and estimating their concentration. In the IBD application, antibody-conjugated SPIONs aggregated in the presence of overexpressed biomarkers, producing detectable signal changes confirmed by fluorescence microscopy. Notably, all methods were compatible with standard MRI protocols, supporting their clinical applicability.

Conclusion:
This study demonstrates the feasibility of using antibody-conjugated SPIONs for the detection of IBD biomarkers and establishes a quantifiable relationship between SPION distribution and MRI signal changes. Furthermore, it highlights the effectiveness of accessible, open-source post-processing techniques combined with standard MRI protocols for reliably characterizing SPIONs in gastrointestinal tissues and supporting their clinical translation.

Author
Yael del Carmen Suarez Lopez
Uppsala University