Quantitative monitoring of mouse lung tumors by magnetic resonance imaging

Abstract

Primary lung cancer remains the leading cause of cancer-related death in the Western world, and the lung is a common site for recurrence of extrathoracic malignancies. Small-animal (rodent) models of cancer can have a very valuable role in the development of improved therapeutic strategies. However, detection of mouse pulmonary tumors and their subsequent response to therapy in situ is challenging. We have recently described MRI as a reliable, reproducible and nondestructive modality for the detection and serial monitoring of pulmonary tumors. By combining respiratory-gated data acquisition methods with manual and automated segmentation algorithms described by our laboratory, pulmonary tumor burden can be quantitatively measured in approximately 1 h (data acquisition plus analysis) per mouse. Quantitative, analytical methods are described for measuring tumor burden in both primary (discrete tumors) and metastatic (diffuse tumors) disease. Thus, small-animal MRI represents a novel and unique research tool for preclinical investigation of therapeutic strategies for treatment of pulmonary malignancies, and it may be valuable in evaluating new compounds targeting lung cancer in vivo.

Figure 1

Anesthesia and RF coil/animal tray setup for MR acquisition: (A) Anesthetized mouse placed inside the nose cone and then inside the RF coil (B); (C) RF coil assembly, including an anesthetized mouse, placed inside the RF shield; (D) Entire coil assembly, placed within the animal tray; (E) Animal tray placed inside the magnet.

Figure 2

Manual segmentation of discrete tumors: (A) Screen shot of a montage containing four segments and the ROI manager. In our experience, this organization of the screen is optimal for segmentation of discrete lung tumors. (B) As described in the text, after selecting the polygon tool, a discrete lung tumor is encircled through a series of left mouse clicks. (C) Clicking the “Add” button in the ROI manager creates an entry within the ROI manager uniquely associated with the encircled ROI (e.g., 0085-0135). (D) A new ROI is created by outlining the tumor on a contiguous image slice and is added to the ROI manager by pressing the “Add” button. (E) After all the ROIs have been added to the ROI Manager, selecting the “Measure” button produces the Results screen shown in (F). Data in Results include the area of the ROI, the mean intensity of the image within the ROI, as well as Minimum (Min) and Maximum (Max) intensity values. As described in the text, these values are subsequently used to estimate the tumor burden.

Figure 3

Graphic, illustrating three different approaches to measuring diffuse tumor burden, as described in Module #3: (A) Monitor progression of disease (as illustrated here) or response to therapy in individual mice via serial imaging; (B) Compare relative lung-tumor burden between mice – Shown here are animals with heavy (left) and light (right) tumor burden; (C) Measure absolute tumor burden (in units of mg), by comparing average, normalized lung-image intensity to that of a cohort of non-tumor-bearing control mice. This measurement requires the development of a calibration curve, as detailed in Tidwell, et. al.

Figure 4

Manual segmentation of diffuse metastatic tumors: (A) The montage is assembled and displayed as described in Figure 2 for segmentation of individual discrete tumors. (B) The ROI manager is opened and the left and right lungs in each image are outlined as a single ROI. (C) The lung ROIs are sequentially added to the ROI Manager, as in Figure 2.

Figure 5

Sample calibration curve for estimation of absolute tumor burden from a test MR image. The calibration curve was generated based on lung weights and automatic lung segmentations of four control mice and ten tumor-bearing mice, following the instructions provided in Step 54.C.

Figure 6

Automated segmentation of diffuse metastatic tumors: (A) After opening Matlab and running the file named Lung_Segmentation_Main.m, the directory containing the files to be segmented is input. (B) If necessary, images oriented horizontally are rotated 90° to a vertical orientation by entering “1”, otherwise enter “0”. (C) The numbers of the image slices containing the most ventral and dorsal portion of the lung are entered into the program. This step is necessary, since not all images will contain lung tissue and some will contain images of spine or anterior chest wall (e.g., slices #1-4, 23 and 24, as illustrated). (D) To initiate automatic segmentation of all the displayed lung fields, the exterior boundary of the lungs in one representative slice is outlined. (E) Following the automated program prompt, the upper (E) and lower (F) center points of the displayed segmented lungs are identified. (G) After outlining the boundary of the lungs and identifying the upper and lower points of the lung, the user is prompted to outline a part of the liver. (H) The lungs are segmented by the program and displayed as a montage plot. (I) Values of the liver-corrected, average lung intensity (Avg_intensity) and total segmented lung volume (Lung_Volume) are output to the Matlab Workspace window.