Abstract
<title>Abstract</title> <p> <bold>Background:</bold> Undifferentiated pleomorphic sarcoma (UPS) is the most common subtype of soft-tissue sarcoma (STS). At our institution, preoperative management typically consists of neoadjuvant chemotherapy plus radiotherapy (CT-RT) or radiotherapy (RT) alone. This study aimed to assess (1) the impact of treatment strategies on therapeutic response, and (2) the utility of multiparametric MRI (mp-MRI), qualitative and morphologic features from contrast-enhanced susceptibility-weighted imaging (CE-SWI) and dynamic contrast-enhanced perfusion-weighted imaging (PWI/DCE), for response evaluation. <bold>Methods:</bold> In this IRB-approved retrospective study, 36 UPS patients underwent presurgical mp-MRI between 02/2021 and 03/2025. Unlike our prior STS research, we intentionally excluded clinically unavailable software-based semiquantitative, quantitative, and radiomic analyses, focusing instead on morphologic and qualitative features readily accessible to clinical radiologists. Lesions were morphologically classified by CE-SWI-T2* and PWI/DCE in six patterns, respectively. PWI/DCE Time-intensity curves (TICs) types III—V, characterized by rapid arterial upstroke, were considered “aggressive,” while TIC I–II as non-aggressive. Patients with ≥90% pathology-assessed treatment effect (PATE) on the surgical specimen were designated as responders (n=20), >30 and <90% PATE as partial-responders (n=10), and ≤30% as non-responders (n=6). <90% patients were combined into a single group (partial/non-responders n=16). <bold>Results:</bold> Of 36 patients, 23 received CT-RT and 13 received RT alone. The CT-RT group demonstrated a higher responder rate (61% vs. 45%) and a higher mean PATE (81% vs. 66%; P=0.054) than the RT group. Among CT-RT responders, 86% showed a CE-SWI-complete-ring pattern (P=2.72×10⁻⁹), 86% a PWI/DCE-capsular-pattern (P=1.05×10⁻⁷), and 93% TIC-type-II (P=1.35×10⁻⁵). A predictive model combining PWI/DCE-capsular-enhancement, TIC-II, and favorable CE-SWI-patterns (“full blooming” or “complete ring”) achieved excellent performance (AUC=0.98) for distinguishing responders from non-responders, outperforming RECIST (AUC=0.52). In the RT group, all responders demonstrated TIC-II (P=0.014), and a model based solely on TIC-II achieved perfect classification (AUC=1.0). <bold>Conclusion:</bold> The CT-RT group shows a non-statistically significant higher proportion of responders and PATE-average than patients receiving RT-alone. MP-MRI-based predictive models using PWI/DCE and CE-SWI morphologic and qualitative clinically available features substantially outperform RECIST in predicting UPS pathology response. While the CT-RT model closely resembled a previously published treatment-agnostic UPS response model, pathology response in RT-alone patients was reliably predicted by the presence of presurgical TIC-II alone, indicating that optimal classification models for distinguishing responders vary by both tumor histology and treatment regimen, underscoring the importance of personalized imaging approaches. </p>