Table 1.
Samples found in the literature on preclinical MR sequences for solid tumor volume measurements in mice.
Fig 1.
Schematic representation of a co-clinical trial which utilizes translational imaging.
Flow charts describing the clinical (A) and preclinical (B) sections of a co-clinical trial of RT with or without the addition of PD-1 inhibition in soft tissue sarcoma. Treatment dosing and imaging procedures in the preclinical arm have been designed to mimic the clinical trial as closely as possible.
Table 2.
Sensitivity and homogeneity of a uniform sample scanned with three rf coil configurations for small animal MR at 7.0T.
Table 3.
Clinically-driven scan program for preclinical MR imaging of soft tissue sarcomas of the extremity.
Fig 2.
Comparison of human and mouse MR images acquired for the co-clinical trial.
Micro-MR images of a sarcoma-bearing mouse leg (bottom row) were obtained with a scan program designed to mimic images acquired in patients enrolled in the clinical arm (top row) bearing soft tissue sarcomas of the extremity. T1-weighted (left), T2-weighted (middle), and T1-weighted + contrast agent injection (right) were acquired in both the human and the mouse arms of the co-clinical trial. White scale bars indicate distances of 5 mm.
Fig 3.
T1 and T2 fitting across a preclinical soft tissue sarcoma demonstrate the anticipated T1 signal range of tumors on study.
Histograms of T1 (A), and T2 (D) values measured in sarcoma tissues are shown, including the mean (blue line) and 2 standard deviations (shaded light blue). Bottles containing a dilution series of magnevist (B) and agarose (E) were measured in the 72 mm volume coil and used to mimic the ranges of T1 and T2 values in tumor during construction of a custom study phantom. Linear regressions of T1 or T2 measurements (C and F, respectively) were plotted along a log scale of solution concentration and shown with 95% CI (dotted lines).
Fig 4.
A custom study phantom demonstrates T1 + T2 range and resolution for mouse sarcoma imaging.
A 3D-printed phantom was designed to hold tubes containing a range of magnevist (T1) and agarose (T2) concentrations, as well as a resolution insert, and was loaded into a syringe filled with water (A). The T1 sequence used in the preclinical trial demonstrates the range of T1 signal within predetermined magnevist concentrations (B), where dilution factor refers to the dilution of a 1% solution in deionized water. The T2 sequence used in the preclinical trial demonstrates the range of T2 signal within predetermined agarose gel concentrations (C). The resolution insert confirms sufficient resolution of the sequences down to 100um (D). Images shown have not been altered or corrected for bias.
Fig 5.
Correction of position-dependent bias resulting from use of the surface coil.
A custom phantom containing concentrations of magnevist with T1 values that span those anticipated in tumors were scanned using the surface coil. T1-weighted imaging of the phantom demonstrated a significant bias (A, left), which distorted the intensity of the magnevist tubes due to distance from the coil. Resulting measurements were unable to reflect the concentration-dependent signal appropriately (B). Application of an N4ITK bias correction, masked with a proton density weighted image (TR = 3000) reduced the effects of the coil bias (A, right), restoring linearity in the signal vs concentration curve (B).
Fig 6.
Reliability and consistency of tumor volume measurements in repeated scans with limb repositioning.
T2-weighted images were analyzed to determine reproducibility of tumor volume measurements resulting from three consecutive scans of sarcoma-bearing limbs in three positions. Each mouse was scanned three times in succession, with the tumor-bearing limb positioned under the surface coil in a flexed, relaxed, or extended position to shift the location and shape of the tumor (represented as blue in the diagram shown in (A)). ROIs were hand-drawn slice-by-slice in triplicate in each resulting image (9 total measurements per mouse) (B), and calculated volumes were compared for repeatability (user precision) and reproducibility (consistency with shifting position) (C). ANOVA analysis of volume reproducibility suggested no dependence of volume measurements on leg position (D), and precision of hand-drawn measurements was confirmed using Brown-Forsythe (E).
Fig 7.
Semi-automated tumor segmentation with 3D Slicer is an acceptable alternative to hand-drawn ROIs for sarcoma volume measurement.
T2-weighted images of three tumors of varying size and morphology were analyzed with hand-drawn ROIs and semi-automated segmentation to calculate tumor volume. Examples of hand-drawn (A) and semi-automated (B) segmentation of the same tumor are shown, including an axial slice of the segmented tumor (a.) and the rendered 3D representation of the tumor according to each segmentation (b.). Resulting volume calculations were compared between triplicate measures with each method for each tumor (C), demonstrating similar measurements and comparable precision. Bland-Altman analysis of both methods, when employed in 6 separate tumors, indicated that semi-automated segmentation is a viable alternative to hand-drawn segmentation (D).
Fig 8.
Prospective gating allows for synchronization of image acquisition with breathing patterns.
Micro-CT acquisition monitoring without (A) and with respiratory gating (B). Coronal images of a lung tumor are shown without gating (C), with respiratory gating (D), and post-mortem (E). Tumor is indicated by white arrows, and diaphragm is noted with yellow arrows. Representative images from triplicates are shown.
Fig 9.
Respiratory gating improves precision and accuracy of lung tumor volume assessment in a free-breathing mouse.
Multiple tumors of varying size and location in the lungs of a free-breathing mouse scanned with micro-CT were identified (A, a-d). Hand-drawn ROIs were used to calculate the volume of each lesion in each scan (a total of 9 measurements per tumor). ANOVA showed that in half of the lesions, respiratory gating had a significant impact on the accuracy of volume measurement (B and C). Variance within repeated measures from gated images was approximately double that of post-mortem images, while non-gated image results demonstrated more than five-fold variance compared to post-mortem measurements (D).
Fig 10.
Demonstration of the VoxPort interface for image data sharing.
Images of VoxPort demonstrating archives of multiple forms of imaging data, including images from multiple modalities, protocol documentation and standard operating procedures, scan acquisition information, and segmentation stacks which correspond to archived images. The image gallery provides a thumbnail and metadata about the images. The user can choose to download a given image or stack or examine the data within the interface.