Target Volume Delineation Using Diffusion-weighted Imaging for MR-guided Radiotherapy: A Case Series of Laryngeal Cancer Validated by Pathology

Abstract

In radiotherapy treatment planning, tumor delineation based on diffusion-weighted imaging (DWI) by magnetic resonance imaging (MRI) is a promising technique. MR-only-based target definition becomes important with the recent development of MRI integrated radiotherapy treatment modalities. In this case series, DWI-based gross tumor volume (GTV) was validated using pathology and compared with a clinical GTV based on computed tomography (CT) imaging and MRI. This case series includes three patients with a laryngeal tumor. Prior to total laryngectomy (TLE), imaging was performed on CT and MRI, including a DWI scan. After TLE, the surgical specimen was processed and cut into 3-mm thick slices. The tumor was delineated on hematoxylin-eosin (HE) stained sections by a pathologist (tumor_HE). This pathological imaging, including the tumor_HE delineation, was three-dimensionally reconstructed and registered to the imaging. The GTV was delineated by a radiation oncologist based on CT and MR imaging (GTV_clinical) and semi-automatically delineated based on DWI (GTV_DWI). The microscopic tumor extent outside the GTV_DWI contour was 3.0 mm, 2.7 mm, and 11.3 mm for cases I, II, and III, respectively. The microscopic tumor extent outside the GTV_clinical was 7.5 mm, 2.1 mm, and 1.5 mm for cases I, II, and III, respectively. The tumor, on histology, was covered by the GTVs for 80%, 74%, and 31% (GTV_DWI) and 73%, 72%, and 89% (GTV_clinical) for the three subsequent cases, respectively. The GTV_DWI resembled the tumor_HE more than the GTV_clinical in case I and case II. In case III, GTV_DWI missed the caudal part of the tumor that was included in the clinical delineation due to a lack of contrast and the heterogeneous signal intensity of the tumor in DWI. In this case series, we showed the potential of DWI for MR-guided radiotherapy treatment if a clear contrast is visible. DWI-based GTV delineation might be a fast alternative to manual delineation, which could speed up the on-table target definition using an MRI-linac system. A larger case series is needed to verify these results.

Keywords: dwi; gtv; larynx; pathology; radiotherapy; target delineation; validation.

Figure 1. HE-stained sections with the larynx tumor of case I. The sections are shown from caudal to cranial direction (from left to right, top to bottom). Both GTV delineations are registered to the HE-stained sections. From these sections, it can be observed that at some locations, the tumor extent is overestimated and at other locations, it is underestimated. In this case, the underestimation is larger in the anterior direction, especially for the GTVclinica. The extralaryngeal growth between the thyroid and cricoid cartilage (indicated with *) and the anterior extent over the midline was not observed in both GTV delineations. The GTV delineations can extend into the air cavity due to small deformations of the specimen and registration errors.

GTV: gross tumor volume; HE: hemotoxylin-eosin; tumor_HE: microscopic tumor delineated by pathologist; GTV_clinical: GTV delineated by radiation oncologist; GTV_DWI: GTV semi-automatically delineated based on DWI

Figure 2. The transverse plane of the larynx tumor of case I for the various images. The tumor growth in the anterior direction is enhanced in the DWI image, while the extent is less visible on the other imaging. The GTVDWI was manually adapted when the DWI image and the ADC map both were enhanced (see arrowhead) because this is indicating edema instead of tumor tissue. The delineations differ slightly from the delineations shown in the HE section due to an angulation difference.

GTV: gross tumor volume; HE: hemotoxylin-eosin; ADC: apparent diffusion coefficient; tumor_HE: microscopic tumor delineated by pathologist; GTV_clinical: GTV delineated by radiation oncologist; GTV_DWI: GTV semi-automatically delineated based on DWI

Figure 3. The transverse plane of the larynx tumor of case II for the various images. The GTVclinical overestimated the tumorHE, whereas the GTVDWI more closely resembles the tumorHE, especially in the posterior part of the larynx where tissue was excluded from the tumorHE and from the GTVDWI but included in the GTVclinical. This region was enhanced in the T2-weighted image, which indicates that these regions can be excluded from the GTVclinical based on these enhancements.

GTV: gross tumor volume; HE: hemotoxylin-eosin; tumor_HE: microscopic tumor delineated by pathologist; GTV_clinical: GTV delineated by radiation oncologist; GTV_DWI: GTV semi-automatically delineated based on DWI

Figure 4. The transverse plane at the cricoid level of the larynx tumor of case III for the various images. The tumor is heterogeneous as can be seen from the HE section, with a mixture of cartilage, fibrotic tissue, and cohesively growing squamous tumor cells centrally located in the tumorHE. And, predominantly, cohesively growing squamous tumor cells at the sides, especially within the GTVDWI. This is reflected in the other images, where the enhancement within the tumor is inhomogeneous. The GTVDWI does not overlap with the whole tumorHE because of the higher ADC values in a part of the tumor. The GTVclinical shows a good overlap with the tumorHE. The delineations on the various images do vary in shape due to a difference in slice thickness, and the angulation of the HE sections compared to the imaging plane of the CT and MRI.

GTV: gross tumor volume; HE: hemotoxylin-eosin; ADC: apparent diffusion coefficient; tumor_HE: microscopic tumor delineated by pathologist; GTV_clinical: GTV delineated by radiation oncologist; GTV_DWI: GTV semi-automatically delineated based on DWI