Multimodal Radiomics Model for Predicting Gold Nanoparticles Accumulation in Mouse Tumors

Jiajia Tang, Jie Zhang, Jiulou Zhang, Yuxia Tang, Hao Ni, Shouju WangarXiv 2024 · DOI · arXiv

Abstract

Nanoparticles (NPs) hold significant promise as diagnostic and therapeutic agents in oncology, yet their clinical translation is hindered by highly variable accumulation in solid tumors. This study addresses this heterogeneity by developing a multimodal radiomics model to predict gold nanoparticle (GNP) uptake. Using 183 mice across seven tumor models, researchers combined CT, B-mode ultrasound, Shear Wave Elastography (SWE), and Contrast-Enhanced Ultrasound (CEUS) to extract 3,299 radiomic features. A composite SVM-based model, incorporating a radiomics signature, tumor type, and imaging means, significantly outperformed basic biological models, achieving an AUC of 0.78 in the test set. These findings suggest that non-invasive imaging can effectively stratify patients for nanomedicine by identifying tumors with high accumulation potential.

Introduction

The "Enhanced Permeability and Retention" (EPR) effect is a cornerstone of nanomedicine, describing the tendency of nanoparticles to accumulate in tumor tissue due to leaky vasculature and impaired lymphatic drainage. However, the clinical efficacy of this effect is often inconsistent. Delivery efficiency can vary by more than 14-fold between individuals and tumor types, influenced by factors such as interstitial fluid pressure, tissue stiffness, and vascular perfusion. To move toward personalized nanomedicine, clinicians require non-invasive tools to predict whether a specific tumor will successfully take up a therapeutic dose before treatment begins.

Methodology

Experimental Design

The researchers synthesized PEGylated gold nanoparticles (GNPs) in three sizes (15nm, 40nm, and 70nm). These were tested across seven distinct subcutaneous mouse tumor models to ensure biological diversity: B16 (Melanoma), E0771 (Breast), Hepa1-6 (Liver), LLC (Lung), MB49 (Bladder), MC38 (Colon), and PAN02 (Pancreatic). A total of 183 tumors were analyzed, split into a training set (n=121) and a test set (n=62).

Multimodal Imaging and Radiomics

Prior to GNP injection, tumors were characterized using four imaging modalities:

  1. Computed Tomography (CT): To assess density and structure.
  2. B-mode Ultrasound: To evaluate morphology.
  3. Shear Wave Elastography (SWE): To measure tissue stiffness.
  4. Contrast-Enhanced Ultrasound (CEUS): To assess blood perfusion.

Quantitative features were extracted using the Pyradiomics package. The researchers employed LASSO regression and Recursive Feature Elimination (RFE) to identify the most predictive features from a pool of over 3,000 candidates.

Model Construction

The team built a composite model using a Support Vector Machine (SVM) classifier to categorize tumors into "high" or "low" uptake groups based on the median accumulation (3.37% ID/g). Additionally, a K-Nearest Neighbors (KNN) regression model was developed to predict the exact percentage of injected dose per gram (%ID/g).

Key Insights

The "Soil" Over the "Seed"

One of the study's most striking findings was that nanoparticle size (within the 15–70nm range) was not a primary determinant of accumulation. Instead, the tumor microenvironment—the "soil"—played a far more critical role than the properties of the nanoparticle "seed."

The Impact of Tissue Stiffness and Collagen

There was a strong negative correlation ($r = -0.563$) between collagen content and GNP accumulation. High collagen density increases tissue stiffness and interstitial fluid pressure, creating a physical barrier to nanoparticle entry. SWE emerged as a vital predictive tool, as stiffer tumors consistently demonstrated lower uptake. Histological validation confirmed that high-uptake tumors had significantly lower collagen levels ($P < 0.001$).

Results

The composite radiomics model demonstrated superior performance over basic models that relied solely on tumor type or nanoparticle size:

  • Classification: The composite model achieved an AUC of 0.93 in the training set and 0.78 in the test set.
  • Regression: The KNN-based model predicted precise accumulation levels with an $R^2$ of 0.58 in the test set.
  • Accuracy: The best-performing iteration reached an accuracy of 83% in distinguishing high-uptake tumors.

Discussion and Limitations

While the study provides a robust framework for predicting nanoparticle delivery, several limitations remain. The use of clinical-grade CT and ultrasound equipment, while beneficial for translational relevance, offers lower resolution than dedicated small-animal scanners. This may have limited the sensitivity of certain radiomic texture features. Furthermore, the CEUS analysis was based on static peak intensity; future work incorporating dynamic "wash-in/wash-out" curves could further refine predictions of vascular permeability.

Conclusion

This research demonstrates that multimodal radiomics can successfully identify tumor heterogeneity and predict the accumulation of gold nanoparticles. By leveraging non-invasive, clinical-grade imaging, this approach offers a path toward better patient stratification and the optimization of nanotherapeutic interventions in oncology.

Summary updated

Abstract

Background: Nanoparticles can accumulate in solid tumors, serving as diagnostic or therapeutic agents for cancer. Clinical translation is challenging due to low accumulation in tumors and heterogeneity between tumor types and individuals. Tools to identify this heterogeneity and predict nanoparticle accumulation are needed. Advanced imaging techniques combined with radiomics and AI may offer a solution. Methods: 183 mice were used to create seven subcutaneous tumor models, with three sizes (15nm, 40nm, 70nm) of gold nanoparticles injected via the tail vein. Accumulation was measured using ICP-OES. Data were divided into training and test sets (7:3). Tumors were categorized into high and low uptake groups based on the median value of the training set. Before injection, multimodal imaging data (CT, B-mode ultrasound, SWE, CEUS) were acquired, and radiomics features extracted. LASSO and RFE algorithms built a radiomics signature. This, along with tumor type and mean values from CT and SWE, constructed the best model using SVM. For each tumor in the test set, the radiomics signature predicted gold nanoparticle uptake. Model performance was evaluated by AUC. Results: Significant variability in gold nanoparticle accumulation was observed among tumors (P &lt; 0.001). The median accumulation in the training set was 3.37% ID/g. Nanoparticle size was not a main determinant of uptake (P &gt; 0.05). The composite model based on radiomics signature outperformed the basic model in both training (AUC 0.93 vs. 0.68) and testing (0.78 vs. 0.61) datasets. Conclusion: The composite model identifies tumor heterogeneity and predicts high uptake of gold nanoparticles, improving patient stratification and supporting nanomedicine's clinical application.
BibTeX
@misc{tang2024multimodal,
  author = {Jiajia Tang and Jie Zhang and Jiulou Zhang and Yuxia Tang and Hao Ni and Shouju Wang},
  title = {{Multimodal Radiomics Model for Predicting Gold Nanoparticles Accumulation in Mouse Tumors}},
  year = {2024},
  eprint = {2406.10146},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2406.10146},
}
Figure