Pilot Study on Evaluating Multiparametric MRI (mp-MRI) Including Contrast-Enhanced Susceptibility Imaging (CE-SWI) in Predicting Pathology Treatment Response in Rhabdomyosarcoma

Abstract

Background and purpose: Rhabdomyosarcoma (RMS) is the most common soft-tissue sarcoma in the pediatric population and an aggressive cancer subtyped as embryonal, alveolar, pleomorphic, or not otherwise specified (NOS). It demonstrates post-therapeutic hemosiderin deposition, granulation tissue formation, fibrosis, and calcification. This pilot study aims to provide initial evidence for developing a multiparametric MRI-feature-derived (mp-MRI) predictive response model to outperform RECIST.
Methods: A UT MD Anderson Cancer Center IRB-approved retrospective pilot study of 11 extremity and pelvic RMS patients with presurgical mp-MRI, including diffusion-weighted imaging (DWI), contrast-enhanced susceptibility-weighted imaging (CE-SWI), and perfusion-weighted imaging with dynamic-contrast-enhancement (PWI/DCE), with surgical resection between 02/2021-06/2024. Lesions were categorized into 6 CE-SWI and 6 PWI/DCE morphologic patterns. Time-intensity curves (TICs) were classified as types I-V. Patients were categorized by the percentage of pathology-assessed treatment effect (PATE) in the surgical specimen as responders (PATE> 90%, n = 3) and partial/non-responders (PATE < 90%, n = 8).
Results: The ADC-mean for the 6 extremity RMS increased minimally from 1,425 ± 476 x 10-6 mm2/s at baseline (BL) to 1,494 ± 386 x 10-6 mm2/s at the presurgical time point (PS). The ADC-mean for the 5 pelvic RMS increased from 1,093 ± 342 x 10-6 mm2/s at BL to 1,677 ± 313 x 10-6 mm2/s at PS. All responders and partial/non-responders presented presurgical RECIST, WHO, and volume stability. At PS, 50% of responders displayed CE-SWI Complete Ring pattern (p = 0.5578), PWI/DCE Capsular pattern (p = 0.6065), and TIC Type-2 (p = 0.6065). No statistically significant differences were observed at PS in ADC or CE-SWI first- or high-order radiomics and PWI/DCE semi-quantitative parameters comparing responders vs. partial/non-responders at PS. PS ROC Analysis: The model based on the CE-SWI textural radiomic GLSZM Large-Area-High-Gray-Level-Emphasis yielded perfect classification performance (AUC = 1.0) separating responders vs. partial/non-responders, outperforming other radiomic, morphologic, and qualitative features such as ADC-Maximum-2D-Diameter-Slice (0.83), CE-SWI Complete Ring (AUC = 0.67), PWI/DCE Capsular (0.67), RECIST (0.67), and TIC type-2 (0.6).
Conclusion: Our pilot study provides initial evidence supporting a model utilizing a single CE-SWI-derived high-order texture GLSZM radiomic feature that, in our small sample, obtained a perfect classification performance (AUC = 1.0), further building on the body of evidence supporting the use of mp-MRI with CE-SWI in soft-tissue sarcoma response assessment and suggesting its potential in predicting RMS response, outperforming RECIST (AUC = 0.67). These promising early results are hindered by a limited statistical power (10%) inherent to the small sample size and the retrospective nature of an exploratory pilot study, highlighting the necessity for further validation through a larger, more representative, and balanced prospective study.