Summary
In this article, we present a novel breast specimen orientation approach with 3D-visualization in a small cohort. The goal was to test feasibility and evaluate to what extent the 3D-visualization supports the relocation of inadequate resection margins. In this novel approach, the surgical clips embedded in the wound bed were also labeled on the specimen. Histologic findings were subsequently reported relative to the clips, instead of relative to the original anatomical orientation in-vivo. We evaluated the feasibility through three outcome measures.
Interpretation results
Outcome measurement 1
For outcome measure 1, a pathologist (RS) and surgeon (HT) scored the face validity of the 3D-visualizations for intuitivity. These scores suggest the 3D-visualizations are predominantly of added value for the surgeons. Experienced pathologists are trained to mentally generate a 3D perspective of the resection margins similar to the 3D-visualization, termed spatial cognition[21]. It neither improves nor confuses the 3D perspective of the pathologist. Hence, the pathologist’s face validity score ranges from ‘+/-’ to ‘+’. On the other hand, communicating this 3D perspective of the resection margin to interdisciplinary colleagues is challenging and the visualization method might be helpful for this purpose. Therefore, pathologists might welcome the 3D-visualization for the interdisciplinary communication of their 3D perspective.
In all cases, the 3D-visualization provided the surgeon with a 3D perspective of the resection margin in the specimen on a level that couldn’t be obtained from conventional reports. The scores ‘+/-’ to ‘++’ were primarily based on the specific case. The 3D-visualization was extremely helpful (++) in cases where the minimal resection margin was situated close to a cliplabel, or in a direct line between two cliplabels, which enables easy relocating in case of re-excision. We hypothesize that the 3D-visualization becomes less effective when the cliplabels are situated further from the resection margin, since the uncertainty of the inadequate margin location would be larger, which might be the underlying factor for the difference in scores.
Outcome measurement 2
We compared the histologic margins to both the 2D and 3D margins computed from the 3D model. The found differences can possibly be attributed to rounding errors to whole numbers by pathology, or the occurrences that the specimen is a bit squeezed during slicing, resulting in a smaller histologic resection margin ventral and dorsal and vice versa for the rest. Moreover, some cases contained tears in the histologic slice, which could not be digitally repaired, overestimating the margin. The largest difference of 3mm can be attributed to the fact that pathologic assessment was not determined in the 2D plane of the histologic slice, but perpendicular to it as it was the most outer slice of the specimen. Here, the unrestricted 3D-visualization resection margin is close to the histologic margin. This highlights the added value of the ability to determine the resection margin in all directions. A mean difference of around 1 mm for the reported resection margin between pathologic assessment and the 2D plane from the 3D-visualization is within expected range since the pathologist rounds the numbers to whole diameters. Given that the deformations during slicing could be partially responsible for the differences, the results suggest that our 3D-visualization is rather accurate.
Outcome measurement 3
For outcome measure 3, we investigated the feasibility of this clip-based approach for breast specimen orientation. The approach has the potential to directly link the histologic findings to the wound bed surface. However, for this approach to work, the locations of the cliplabels on the specimen have to correlate directly to the locations of the clips on the wound bed. Some problems were observed during the course of this study. First, in a selection of cases, the surgeon chose to place a part of the clips and cliplabels ex vivo directly after excision, which possibly distorts the correlation between both locations. We suspect that sufficient training and experience overcomes this problem, but the presence of this learning curve does complicate broad implementation. Second, we observed in two cases (#3 and 7) that two cliplabels were situated very close to each other. We suspect that in these cases the surgeon chose to increase the margin around the tumor, resulting in both the clip and the cliplabel being located on the specimen. Third, a cliplabel appeared missing in several cases. Multiple reasons can contribute to a missing cliplabel: (1) The surgeon chose to decrease the margin, meaning the cliplabel is embedded in the wound bed. (2) A cliplabel detached during excision. (3) A cliplabel gets enveloped by the specimen tissue, in which case the label was later found during slicing. Originally, anatomical orientations were assigned to each clip at pathology. This proved to be difficult when a clip (with unknown orientation) was missing or appeared double. For future research, we are looking for a more robust approach to prevent these problems[8].
Strengths and limitations
Research design
The strengths of our study design included: (1) it is a prospective study, (2) the cohort contained multiple types of cancers (DCIS, LCIS, IDC, and ILC), (3) both multiple quantitative and qualitative outcomes were measured, and (4) multiple surgeons participated, suggesting the results are more generalizable towards other hospitals. A study design limitation was the small sample size (n = 15), with only one case with inadequate margins.
Research setup
The strengths of the research setup included: (1) an affordable 3D camera that handles wet and reflective surfaces well, and (2) a compact and mobile turntable setup to capture the 3D model of the specimen, which allowed coloring of the clips to mark them as well. Limitations included: (1) the relatively low detail of the camera (1 mm) leading to a smoothed 3D model, (2) the turntable setup left the bottom of the specimen uncaptured, (3) the procedure was time-consuming and laborious, (4) missing data points such as disappeared clips, or torn histologic images, (5) digital slicing was equally distributed over the specimen and not based on the real thickness of each slice, and (6) the deformable registration algorithm did not account for varying stiffness of different tissue types within the specimen.
There were two comparative studies, one about the label-based relocation approach[19], and one describing a 3D-visualization method of the specimen[18]. First, the study by van Lanschot et al. demonstrated a label-based relocation approach during oral cavity oncology surgery. The main difference with our approach was, aside from the specimen type, that they performed an intraoperative margin assessment. Therefore, they were able to number the labels and remove them before closing the wound. Second, the study by Koivulhoma et al. presented a 3D-visualization method of a tongue specimen including histology. Besides specimen type, there were several differences with our setup. First, their scanning method needed optic spray to handle wet reflective surfaces. Second, they implemented a net to capture the bottom of the specimen, where we employed a transparent turntable. Third, their method can cope with slices in multiple directions, where our method assumes parallel slicing, which is the standard at our pathology department. Fourth, they interpolated between tumor slices to create one solid tumor, where we chose not to because multiple solids (DCIS) can be present and interpolation could result in a misrepresentation of tumor tissue. Last, their method did not provide an estimation of the resection margins.
Future perspectives
Further research should be directed at improving the clip labeling technique or finding an alternative to link the location in the histologic margins to in vivo wound bed surface. Targeted treatment of inadequate margins based on this approach is only feasible when the cliplabels are an accurate representation of the surgical clips embedded in the wound bed.
An alternative strategy for follow-up research is to focus solely on the 3D-visualization benefits and disregard the cliplabels, as the 3D-visualization provides an intuitive 3D perspective in interdisciplinary communications.
A CT scan of the specimen can function as an alternative to the 3D imaging setup. Although a CT scan is more expensive, an advantage is that it provides interior information, which can be linked to histology and enable deformation based on tissue stiffness.
When an accurate clip labeling technique is available and validated, it is interesting to extract the geodesic distance between clips and inadequate margin. The geodesic distance is the shortest distance following the curvature of the specimens’ surface, which represents the maximum distance between the potential inadequate margin and the concerning clip. In addition to which clips are in the proximity of the resection margin for re-excision, the surgeon also acquires the geodesic distance from each clip with this method.
These distances could prove even more beneficial in radiotherapy, where it could indicate the area at risk for residual tumor. In radiotherapy, the geodesic distances can be interpreted as radius for a spherical area around each clip. Where these spheres overlap can be considered as the area at risk.
Although it is out of the scope of this article, the algorithm to compute the geodesic distance is developed and available in supplementary files. Although not accurate due to suboptimal clip labeling discussed in outcome measure 1, the distances were plotted on the 3D-visualizations to provide visual understanding (see supplementary files).
As mentioned in the introduction, an efficient method to avoid inadequate margins is through intraoperative margin assessment. However, for as long pathologic assessment is necessary, research for improved breast specimen orientation methods remains relevant.