Imaging Patterns of Response in Desmoid Tumors

• 

  Fig 1: Representative responding desmoid tumor. Superior row display response evolution using CE-SWI with a higher demonstration of 90th -percentile voxels. The lower row displays the response evolution of the same desmoid tumor using T2-WI, where 10th-percentile voxels are slightly better represented.
  Figure 1 presents a comparison between CE-SWI and T2-STIR images and their respective histograms of a desmoid fibromatosis tumor across 4 time points undergoing response to treatment from left to right demonstrating the process of reduced enhancement and increased T2 hypointensity during response. The increase of T2 hypointense elements is typically considered a sign of positive response irrespective of tumor size. In contrast, an increase in the T2 hyperintense and T1 shortened enhancing components is often seen as a sign of progression that can precede enlargement.

Expected Trends in Responding Desmoid Tumors

• 

  Fig 2: Expected radiomic trends in responding desmoid tumor patients.
  Figure 2 shows plots of RECIST, Volume, mChoi, and first-order radiomic features of a representative patient across treatment demonstrate expected response trends, including decreased mean, increased skewness, increased 10th percentile voxel proportion, and decreased 90th percentile voxel proportion. Thresholds for progression and response were set at 20% and 30% for RECIST and volumetric mChoi assessments. An increase of 25% and a decrease of 50% were considered thresholds for progression and response, respectively, for 3D volumetric assessment.

T2-Based RECIST Response Categorization

• 

  Fig 3: RECIST, volume, and mChoi plots from 6 representative desmoid fibromatosis patients and their corresponding response categorization.
  On Figure 3 the left panels show true progression by T2-based RECIST, also detected by CE-SWI-based-volume and CE-SWI-based-mChoi with an average of 4.5 months earlier. Middle panels show stable T2-based RECIST with discrepancy assessment of progression by CE-SWI volume in 2 representative patients. Clinical radiologists are often insensitive to detect progression by volume change. Right panels show stable RECIST with discrepancy assessment of response detected by CE-SWI mChoi in 2 representative patients. Clinical radiologists are very sensitive to changes in T2 signal and enhancement in correlation with the variation of mChoi values. 

CE-SWI Sensitivity for Detection of Hyperintense Voxels in Progression

• 

  Fig 4: Representative progressive (left) and responding (right) desmoid tumors with their corresponding voxel intensity histograms. Responding tumors are densely collagenized with T2* hypo intense fibrosis causing an increase of 10th-percentile voxels and skewness. Immature, enhancing and progressive tumors have a higher proportion of 90th-percentle voxels.
  On Figure 4 the left panels show a higher percentage of voxels above the 90th percentile in CE-SWI versus T2-STIR in an active progressing lesion. Right panels show a higher percentage of voxels below the 10th percentile in T2 versus CE-SWI in a highly collagenized responding tumor. On average, CE-SWI imaging captured 23% more 90th percentile voxels than T2.

First-Order Radiomic Trends in Desmoid Fibromatosis

• 

  Fig 5: First-order radiomic trends in 6 representative patients for each of the three response categories.
  Figure 5 presents respresenative patients for each response category. The red boxes highlight the trends of first-order radiomic features concordant with the expected response category trend. The expected trends associated to responding patients, including increasing 10th percentile hypointense voxel proportion, decreasing 90th percentile hyperintense voxel proportion, decreasing mean, and increasing skewness, and the opposite in progressive patients, were present in at least one of these features in 90% of patients, and all four combined 30% of them.