Discussion

In this study, iodine-stained tumorous lungs from mouse models of pulmonary cancer were inspected with µCT. The need of using a contrast agent in attenuation-based µCT of or-gans is demonstrated in the comparison of an unstained lung (Fig.6.2(a)) with a stained one (Fig.6.2(b)). We tested some staining solutions containing either tungsten or iodine as X-ray absorbing elements, and we finally selected an iodine in ethanol solution (I2E) as the best

Figure 6.5: 2D and 3D images of three different I2E-stained tumorous lungs. (a)–(c) To-mograms of three different types of lung samples carrying tumors show lesions, including some tumors with undefined boundaries. Below, their corresponding three-dimensional ren-derings are shown (d)–(f), with normal lung tissue represented in transparent white to allow the visualization of the tumorous lesions pseudo-colored in pink. Scale bar: 1 mm.

staining agent for our purposes (Fig. 6.2(b) and Fig. 6.3). I2E is known as a good contrast agent to highlight the morphology of organs of mammals [Rasch et al., 2016] and, as we show here, it is also adequate to discriminate areas of healthy tissue from unhealthy tissue in diseased lungs. Due to the small voxel size obtained in the reconstructed I2E-stained lungs, we were able to segment the tumors that grow with defined boundaries using a semi-automated tool of the VGStudio software. However, the growth of tumors does not always occur with well-defined limits and burdens may not have very precise edges, as indicated, for example, by the arrows in the enlarged area in Fig. 6.4(a). These areas are a challenge for the semi-automated segmentation. Therefore, additional manual intervention was taken along with an interpolation over the reconstructed slices to obtain the reconstructed lungs shown in Fig.6.5. To validate our method, the tumorous lungs previously analyzed byµCT were later stained with the well-established H&E-staining protocol and imaged with an op-tical microscope. Routine H&E-staining plays a key role in histopathological studies of tissues for diagnosis and research purposes. It yields excellent contrast to the histological slice images obtained with an optical microscope, thus tissue morphology as well as tissue microscopic changes and abnormalities can be well identified. The H&E-stained slices of three different diseased lungs are compared to similar virtual µCT slices in Fig. 6.6. Al-most identical structures are observed in both the H&E-stained slices (Fig. 6.6(d)–(f)) and the analogous virtual µCT slices (Fig. 6.6(a)–(c), enlarged areas), such as indicated by the arrows that identify tumor burdens (blue) and blood vessels (red). A meticulous analysis of one histological slice and a much comparable virtual µCT slice shows differences in the

6.2 Preclinical studies: stained lungs 77

Figure 6.6: Tomograms and comparable H&E-stained histological sections of three tumor-ous lungs. (a)–(c) Tomograms of tumortumor-ous lungs, with the corresponding framed highlighted areas shown below. The regions of these tomograms are equivalent to the histological slices shown in (d)–(f). Blue arrows point tumor burdens, while red solid arrows indicate blood vessels. Scale bar: 1 mm.

areas identified as tumors. The tumorous regions marked with red lines in histological slices in Fig. 6.7(a) and (b), were delineated by an experienced pathologist. They correspond to 20.5 % (Fig. 6.7(a)) and 3.3 % (Fig. 6.7(b)) of the total surface of the lung in these spe-cific slices. Other areas located in the outer limit of the red contours, although stained in a darker pink tone, do not correspond to tumors, but to abnormal tissue such as inflammation, dead and/or stressed tissue. The segmentation tool used to define the tumorous areas marked in red in the virtualµCT slices give slightly higher numbers of tumorous areas (22.8 % in Fig. 6.7(c) and 10.5 % in Fig. 6.7(d)). Even considering small deviations conditioned by the individual interpretation of suspicious structures in the histological slice, there is a dis-crepancy between the percentage of tumorous areas in the histological (Fig.6.7(a)–(b)) and virtualµCT slices (Fig.6.7(c)–(d)). The stain (Iodine) used here to improve the contrast in µCT is non-specific to tumorous tissue, and therefore stains similarly the inflammation and the dead and/or stressed tissue.

Figure 6.7: Segmentation of tumorous areas in histological sections and their more resem-bling virtualµCT slices. (a)–(b) Histological sections of diseased lungs with tumorous areas circumscribed about red lines. (c)–(d) Virtual µCT tomograms, corresponding to the his-tological slices above. Red marked areas correspond to tumors identified using the semi-automated “region grower” tool of VGstudio Max. Scale bar: 1 mm.

Nowadays, H&E is the gold standard in histopathological studies, providing a detailed representation of tissues with high resolution, thus allowing to identify tissue abnormalities.

However, it is known that even among experienced histopathologists some differences in the evaluation of tumor extents are likely to occur. Moreover, histopathology does not allow the volumetric study of tumors, since evaluating the percentage of tumorous areas based on a few histopathological slices presents a major drawback related to the missing information from the excluded slices. Hence, performingµCT prior to H&E-staining would be advisable in a workflow, as the first does not prevent the latter to be completed. Especially in the case of tumor volume studies,µCT would provide extra volumetric information that is important in the evaluation of potential differences in the macroscopic growth of patterns among tumor types.