The study included the analysis of 5,135 MRI scans performed for extremity STS in our institutional 29-magnet fleet (consisting of 1.5T and 3.0T scanners from two manufacturers) between February 2021 and May 2023. Within this time range, 643 UPS studies were completed, including pre-operative primary tumor assessment and post-operative surveillance studies. We excluded all post-operative surveillance and primary myxoid-UPS cases, yielding 33 extremity cases. All 33 cases underwent surgical resection preceded by at least one pre-operative MRI study within the defined time range mentioned above. As our study aims to evaluate the effectiveness of PWI/DCE metrics in characterizing treatment response, we established advanced/functional imaging MRI protocols to evaluate STS of the extremities. PWI/DCE metrics were compared at different time points against the standard references, including: (1) Pathology-assessed treatment effect (PATE) on surgical specimens, and (2) conventional tumor treatment size-based metrics such as RECIST, WHO, and volume.UPS patient populationThis retrospective study included 33 patients ranging in age from 36 to 85 years. Twenty were male (61%) and 13 were female (39%). The average age of the patients was 64 years (Table 1). Out of the 33 patients, 10 underwent BL MRI studies outside our institution without PWI/DCE, while their subsequent PC and PRT MRIs were obtained at our institution. These 10 patients were grouped under the “Conventional Imaging” category. The remaining 23 patients had a complete set of advanced MRI studies, including BL, PC, and PRT, performed at our institution. This latter group was classified as the “Functional Imaging” group. Eighteen patients did not receive neoadjuvant chemotherapy, and 3 did not receive neoadjuvant radiotherapy before surgical excision. Therefore, they were excluded from the PC and PRT analyses, respectively. Overall, 23 patients were included in the BL group, 15 in the PC group, and 30 in the PRT group (Table 1).Table 1 UPS patient population characteristics.Reference standard PATEThe 33 patients were categorized into three groups based on the surgical specimen’s PATE percentage. Tumors demonstrating equal to or greater than 90% PATE were classified as responders (R), tumors with a PATE in the 31–89% range were labeled as partial responders (PR), and tumors with a PATE of 30% or less were considered non-responders (NR). The three groups were compared, including 16 subjects in the R group, 10 in the PR group, and 7 in the NR group (Table 1). The combined PR/NR group had 17 subjects.Reference standard RECIST, WHO, and volumeAll 33 patients were analyzed for RECIST, WHO, and volumetric measurements. Maximum diameter, area, and volume were measured at PC and PRT with respect to BL, comparing responders and partial/non-responders. RECIST, WHO, and volume criteria for partial response (PR) threshold were set at 30%, 50%, and 50% decrease, respectively. Progressive disease (PD) threshold was set at 20%, 25%, and 25% increase, respectively14. When compared with BL, all R, and PR/NR patients at PRT displayed size changes that would fall within the range of stability, namely between + 20% and − 30% for RECIST and + 25% and − 50% for WHO and volume criteria. All R and PR/NR cases demonstrated pseudo-progression at PC, crossing the threshold of + 20% for RECIST and + 25% for WHO and volume assessments. (Fig. 2).Fig. 2Scatter plots of RECIST, WHO, and volumetric measurements comparing responders vs. partial/non-responders across different time points in their treatment. BL: Baseline; PC: Post-Chemotherapy; PRT: Post-Radiation; PD: Progressive Disease; PR: Partial Response.MR image acquisitionStandardized PWI/DCE sequence2 parameters were tailored according to MRI vendor and field strength. Following a power-injector delivered intravenous dose of a gadolinium contrast agent at a 3 cc/sec rate, a fast volumetric T1-shortening sensitive three-dimensional (3D) gradient echo acquisition lasting 4 min was performed, with a temporal resolution of 7–12 s. Sequence parameters of the dynamic sequence were: FOV = 38–42 cm, matrix = 256 × 160 (in-plane resolution 1.5–1.65 mm), TR = 3.4-3.8ms, TE = 1.3-1.6ms, flip angle = 12, slice thickness = 4 mm. During the pre-operative treatment, multiple MR studies were acquired for each patient and compiled into three classes: Baseline (BL, pre-therapy), post-systemic chemotherapy (PC), and pre-operative/post-radiation (PRT) time points. For STS, we conducted one pre-therapy baseline study, one to three studies during systemic therapy, and at least one post-radiation study one to two months after radiation therapy and immediately before surgical resection. While the BL is a single time point, multiple time points were classified as PC or PRT for each patient.Imaging storage and post-processingThe acquired MRI data sets were transferred to the institutional Picture Archiving and Communication System (PACS) (IntelliSpace PACS, Philips, Amsterdam, Netherlands). MR images were retrieved from PACS, and each tumor was manually contoured in three dimensions by an MD-trained research assistant to create tumor volume of interest (VOI) segmentation. Conventional tumor size metrics for all three x, y, and z planes were registered for all time points and used to estimate RECIST, WHO, and volumetric assessment metrics. MIM version 7.1.4 (MIM Software Inc., Cleveland, USA) was used for post-processing to outline, process, and generate VOIs from PWI/DCE images. The segmented tumor volume files were subsequently exported to an institutional network storage drive as DICOM RT-Struct files for further analysis. An in-house developed Cancer Radiomic and Perfusion Imaging (CARPI) automated framework15, capable of intensity histogram-based first-order and high-order radiomic feature extraction from advanced MRI sequences, processed all the collected RT-Struct files containing segmented VOI data. CARPI also extracted 7 semiquantitative kinetic parameters and maps: Wash-in rate (WiR), wash-out-rate (WoR), peak enhancement (PE), the wash-in area under the curve (WiAUC), time-to-peak (TTP), wash-out area under the curve (WoAUC), and total area under the curve (AUC) from TICs15,16. One-hundred and seven radiomic features were then derived, including shape (14 features), first-order statistics (18 features), and texture (75 features), which were automatically recorded in a database management system (DBMS) based on PostgreSQL (The PostgreSQL Global Development Group, Berkeley, USA). Statistical analyses compared responders vs. partial/non-responders.Qualitative PWI/DCE analysisQualitative analysis refers to the subjective analysis of the time-signal intensity curve (TIC), which visually represents the net contrast resulting from inflow and clearance during the passage of the gadolinium bolus over time. In PWI/DCE, the TIC usually displays three stages of perfusion: (1) upslope, which reflects contrast wash-in; (2) plateau, which represents the steady state of contrast within the interstitial fluid but may not be visible in all lesions; and (3) downslope, which reflects contrast wash-out as gadolinium passes out of the tissues under examination. The TIC shape was subjectively assigned to one of five curve types based on their morphology by an experienced radiologist: Type I, II, III, IV, and V. Lesions were categorized into TIC III, IV, and V when displaying a rapid wash-in/early upstroke curve9,13.Semiquantitative PWI/DCE analysisThe semiquantitative analysis measures empirical perfusion parameters derived from the TIC morphology. It includes maximum/peak enhancement (Emax, PE – curve peak), rate of enhancement/wash-in rate (Eslope/WiR -upslope gradient), area under the curve (AUC), and time to peak enhancement (TTP -time between base of upslope and Emax). Additional semiquantitative metrics include wash-out rate (WoR), wash-out area under the curve (WoAUC), and wash-in area under the curve (WiAUC)17,18,19,20.Quantitative PWI/DCE analysisIn the field of musculoskeletal (MSK) imaging, quantitative analysis involves pharmacokinetic modeling, which requires arterial input function to measure gadolinium concentration changes over time. The most common pharmacokinetic model used in MSK imaging is the extended Tofts model, which generates four perfusion parameters. These parameters include the transfer constant (Ktrans), exchange rate constant (Kep), extravascular extracellular space fractional volume (Ve), and plasma fractional volume (VP)21,22,23. Given that most quantitative analyses occur primarily in the research setting, the relatively high number of available overlapping publications, and the limited availability of FDA-approved clinically applicable vendor-neutral solutions for MSK analysis, we decided to focus mainly on assessing the value of morphologic, qualitative TIC-related and semiquantitative mapping-derived features that can be easily assessable in the daily clinical setup.Statistical analysisThe PWI/DCE morphology patterns and TIC curve types in R, PR, and NR were compared using one-way chi-squared tests. The PWI/DCE semiquantitative parameters and radiomic features in R vs. PR/NR were compared using two-tailed non-parametric Wilcoxon rank-sum tests. Finally, receiver operating characteristic (ROC) analysis of the most relevant PWI/DCE features discriminating R from PR/NR was performed. All statistical analyses were implemented in Python 3.10.13 using the SciPy library version 1.12.0 and the Scikit-Learn library version 1.4.1. Statistical significance was assessed at 5% (P < 0.05).
