[Paper Review] MPRAD: A Multiparametric Radiomics Framework

Vishwa S. Parekh, Michael A. Jacobs|arXiv (Cornell University)|Sep 25, 2018

Radiomics and Machine Learning in Medical ImagingMedicine56 references17 citations

TL;DR

MPRAD is a multiparametric radiomics framework that integrates multiple MRI sequences to extract high-dimensional textural features for improved tissue characterization. It achieved 87% sensitivity and 80.5% specificity (AUC 0.88) in distinguishing malignant from benign breast lesions—9–28% higher AUC than single-parameter radiomics—while showing consistent feature distributions across normal tissue types in both breast and brain stroke applications.

ABSTRACT

Multiparametric radiological imaging is vital for detection, characterization and diagnosis of many different diseases. The use of radiomics for quantitative extraction of textural features from radiological imaging is increasing moving towards clinical decision support. However, current methods in radiomics are limited to using single images for the extraction of these textural features and may limit the applicable scope of radiomics in different clinical settings. Thus, in the current form, they are not capable of capturing the true underlying tissue characteristics in high dimensional multiparametric imaging space. To overcome this challenge, we have developed a multiparametric imaging radiomic framework termed MPRAD for extraction of radiomic features from high dimensional datasets. MPRAD was tested on two different organs and diseases; breast cancer and cerebrovascular accidents in brain, commonly referred to as stroke. The MPRAD framework classified malignant from benign breast lesions with excellent sensitivity and specificity of 87% and 80.5% respectively with an AUC of 0.88 providing a 9%-28% increase in AUC over single radiomic parameters. More importantly, in breast, the glandular tissue MPRAD were similar between each group with no significance differences. Similarly, the MPRAD features in brain stroke demonstrated increased performance in distinguishing the perfusion-diffusion mismatch compared to single parameter radiomics and there were no differences within the white and gray matter tissue. In conclusion, we have introduced the use of multiparametric radiomics into a clinical setting

Motivation & Objective

• To address the limitation of single-parameter radiomics in capturing complex tissue heterogeneity in multiparametric imaging.
• To develop a unified framework that enables joint analysis of multiple MRI sequences for improved radiomic feature extraction.
• To evaluate the performance of multiparametric radiomics in distinguishing malignant from benign breast lesions.
• To assess the framework’s ability to differentiate perfusion-diffusion mismatch in ischemic stroke.
• To ensure robustness by verifying consistent feature distributions across normal tissue types (e.g., glandular tissue, white and gray matter).

Proposed method

• The MPRAD framework fuses multiple MRI sequences (e.g., T1, T2, DWI, perfusion) into a high-dimensional radiomic feature space.
• It extracts textural features from the combined multiparametric dataset using established radiomic methods such as first- and second-order statistics, shape, and wavelet-based features.
• Feature selection and dimensionality reduction are applied to identify the most discriminative features across disease states.
• Machine learning models (e.g., SVM or random forest) are trained on the multiparametric feature set to classify tissue types.
• The framework ensures tissue-specific consistency by comparing feature distributions in normal tissues (e.g., glandular tissue in breast, white/gray matter in brain).
• Performance is evaluated using AUC, sensitivity, and specificity, with comparison to single-sequence radiomic approaches.

Experimental results

Research questions

• RQ1 Can a multiparametric radiomics framework improve diagnostic accuracy in breast cancer detection compared to single-sequence radiomics?
• RQ2 How does MPRAD perform in identifying perfusion-diffusion mismatch in acute ischemic stroke?
• RQ3 Are MPRAD features consistent across normal tissue types (e.g., glandular tissue, white and gray matter) in both organs?
• RQ4 Does integrating multiple MRI sequences lead to a statistically significant improvement in AUC compared to single-sequence radiomics?
• RQ5 Can MPRAD features reliably distinguish malignant from benign breast lesions while maintaining stability in normal tissue regions?

Key findings

• MPRAD achieved an AUC of 0.88 in distinguishing malignant from benign breast lesions, representing a 9–28% improvement over single-parameter radiomic approaches.
• The framework demonstrated 87% sensitivity and 80.5% specificity in classifying breast lesions, indicating strong diagnostic performance.
• MPRAD features in normal glandular tissue showed no significant differences between malignant and benign groups, confirming robustness in normal tissue.
• In brain stroke, MPRAD outperformed single-parameter radiomics in identifying perfusion-diffusion mismatch, enhancing diagnostic precision.
• No significant differences in MPRAD features were observed within white or gray matter in the brain, indicating consistent behavior across normal tissue types.
• The multiparametric approach successfully captured complex tissue characteristics in high-dimensional imaging space, overcoming limitations of single-sequence radiomics.